Structural Divides Shape the Nonlinear Nature of Human Mobility in COVID-19

Structural Divides Shape the Nonlinear Nature of Human Mobility in COVID-19 | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Structural Divides Shape the Nonlinear Nature of Human Mobility in COVID-19 Madhumini Kulathunge, Lelumi Edirisinghe, Isuri Mapa, Roshan Godaliyadda, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8482833/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract Mobility reflects collective human behaviour, revealing how people interact with their physical and social environments. The COVID-19 pandemic provided a unique opportunity to identify the hidden behavioural and structural properties of society through its mobility responses, with its unprecedented impulse. Data-driven modelling was applied to a nationwide survey dataset collected from Sri Lanka in 2021, to identify 4 distinct archetypes, namely, Privileged Adaptors, Cautious Traditionalists, Stable Minimalists, and Resilient Rebounders, showing diverse adaptations across 13 key factors. These inductively derived clusters from their crisis-driven mobility adaptations subsequently revealed distinct ethnic, educational, and socioeconomic patterns, showing that human behaviour and socioeconomic status are linked not linearly but through structurally conditioned, nonlinear trajectories of risk absorption. These trajectories are governed by the interplay of substitution capacity, exposure necessity, and support infrastructures of the archetypes, offering a generalizable framework for understanding human behavioural inequality in a global disruption. Humanities/Complex networks Social science/Complex networks Physical sciences/Mathematics and computing Physical sciences/Physics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction Mobility is a core attribute of human behavior that extends beyond physical movements to represent the integrated behavioral outcome of cognitive, emotional, and social processes that define human functioning, reflecting bounded rationality under real-world constraints (Simon, H., 1955 ). Social connectivity, economic exchange, and cultural diffusion shape mobility patterns, ultimately determining the very structure of societies. Similar to an impulse function in a dynamic system that reveals its intrinsic characteristics, unexpected disruptions such as pandemics may cause changes and reset the equilibrium states of collective behaviour in the societal structural dynamics, exposing latent properties. These distortions to the usual mobility patterns can reveal useful insights into behavioral adaptation, resilience, and vulnerability, reflecting the broader socioecological systems (Bronfenbrenner, U., 1977 ; Bronfenbrenner, U. 1979 ) to understand how people respond in the perception of risk, stress, need, and opportunity. The coronavirus disease 2019 (COVID-19) pandemic was such a strong driving factor that revealed this underlying behaviour of human societies all over the globe, by offering a unique opportunity to gauge how different civilisations around the world respond, adapt, and falter under extreme pressure in a sudden disturbance (Pangallo, M. et al., 2023 ; Petherick, A. et al., 2021 ; Santana, C. et al., 2023 ). The effects of the outbreak were widespread, including health, economy, psychology, education, and many more. There were over 772 million confirmed cases and approximately 7 million deaths in 2023, with accurate figures most likely to be significantly higher due to underreporting and limited testing (WHO COVID-19 Dashboard, 2025 ). Beyond the immediate health crisis, the pandemic caused the worst global economic downfall since World War II. The International Monetary Fund (IMF) recorded around 3% drop in the world's GDP in 2020 (IMF, 2021 ), and the International Labor Organization (ILO) reported that working hours equivalent to about 255 million full-time jobs were lost, destroying livelihoods and worsening poverty (ILO, 2021 ). In response, governments implemented strict containment measures, including lockdowns, curfews, and social distancing protocols (Onyeaka, H. et al., 2021 ; Andronico, A. et al., 2021 ). These critical public health constraints redefined human mobility, transforming it from a routine aspect of life into a regulated and often restricted privilege (Jay, J. et al., 2020 ; Gozzi, N. et al., 2021 ). This had a cumulative effect on access to essential services, employment, and social interactions, with the burden of these disruptions falling disproportionately along socioeconomic disparities (Xu, F. et al. , 2025 ; Xi, H. et al., 2023 ), exposing the nonlinear dynamics of the civilizations. Existing inequalities in wealth, education, and digital access created great diversity in adaptation and resilience across the society (Jay, J. et al., 2020 ; Wang, J. et al., 2022 ; van Deursen, A. J., 2020 ). Due to the South Asian region's dense populations and large informal economic sectors, more than 80% of jobs fall outside formal protections (World Bank Open Data, 2025 ; Ahmed, S. A. et al., 2023 ). This structural characteristic associated with lockdown measures led to a direct threat to the workforce that relied on daily mobility and physical presence, thereby intensifying the trade-off between public health and economic survival (World Bank, Washington, DC, 2020 ). In Sri Lanka, the Ministry of Health recorded an official death toll of over 15,000 by February 2022, posing a significant public health burden for a population of approximately 22 million (Galgamuwa, L. S. et al., 2023 ). This crisis, compounded by stringent and prolonged control measures, including strictly enforced, district-wide lockdowns and curfews, brought social and economic life to a near standstill (World Bank Group, 2020 ; Agampodi, S., 2021 ). With its distinctive sociodemographic structure marked by high educational attainment yet limited economic resilience, a diverse ethnic composition, and variable quality of health services, Sri Lanka offers a compelling lens through which to examine how such severe constraints reshaped daily lifestyles, behavioural patterns, and adaptive capacities (World Bank Group, 2020 ). The pandemic’s economic repercussions in Sri Lanka were equally profound, with the output sharply contracting due to prolonged lockdowns (Central Bank of Sri Lanka, 2020 ). While Sri Lanka’s economic fragility stemmed from longstanding structural imbalances, the pandemic likely accelerated the trajectory toward collapse by amplifying fiscal pressures and disrupting key income sectors. GDP in Sri Lanka fell by 3.6% in 2020, with sectors such as tourism witnessing income losses exceeding 80% relative to 2019 (Central Bank of Sri Lanka, 2020 ; Central Bank of Sri Lanka, 2022 ; World Bank Open Data, 2025 ). This downturn pushed an estimated half a million people into poverty, while unemployment rose from 4.7% in the first quarter of 2019 to 5.7% a year later, severely constraining households’ ability to sustain their mobility and livelihoods under ongoing restrictions (World Bank Group, 2020 ). The significance of this study is its unique, high-resolution dataset collected through a door-to-door field survey of 3020 households conducted across Sri Lanka during a period of eased restrictions, i.e., from November to December 2021(Ilangarathna, G. A. et al., 2024 ). It captured various social effects of the pandemic, including its impact on health, economy, education, psychology, culture, and mobility. Building on earlier research based on this survey (Senarath, N. et al., 2024 ; Thilakasiri, I. et al., 2025 ), this study analyzes human behavior on household mobility during the time of the crisis. This data collection methodology was critical in capturing comprehensive behavioral and socioeconomic data from groups that are often underrepresented in online surveys, such as those from rural areas and lower socioeconomic classes with limited internet access. A longitudinal perspective that is uncommon in the literature was provided by the survey, which specifically recorded retrospective, self-reported changes in mobility patterns across the three separate waves of the pandemic, together with the pre-pandemic period, allowing for a nuanced and more person-centric analysis to uncover the diverse experiences of different societal groups. While numerous studies have investigated the more general changes in human mobility during the pandemic using digital trace data(Andronico, A. et al., 2021 ; Jay, J. et al., 2020 ; Pan, Y. et al., 2020 ), a scarcity remains in nuanced, data-driven analysis that connects these behavioural changes to the socioeconomic backgrounds of individuals. Most existing literature generally focuses only on what changed and lacks the specificity to reveal why and how different socioeconomic groups exhibited vastly different capabilities in terms of adapting to the unexpected situation. Studies also suggest that risk perception is conditioned by livelihood pressures(Nisa, C. F. et al., 2021 ) and demonstrate that collective resources such as community support can assist in adaptation when formal resources are limited(Durante, R. et al., 2021 ; Ferwana, I. et al, 2021 ; Fraser, T. et al., 2022 ). This study addresses this gap with the use of Factor Analysis together with unsupervised machine learning. We identified distinct behavioural clusters within the Sri Lankan population, revealing how pre-existing socioeconomic disparities dictated the behavioural capacity to adapt to a prolonged crisis. This leads to the revelation of latent properties and uncovers a deeper understanding of the nonlinear characteristics of human society in a global epidemic, highlighting its potential to aid in formulating more informed public health strategies in the future. Our multi-step data-driven approach first quantified the behavioural changes from the pre-pandemic to the pandemic period using factor analysis to identify 13 principal factors related to human mobility. Certain survey questions aligned with them, hence appropriate descriptors were selected for each factor in the subsequent analysis. The dataset was then denoised using Principal Component Analysis (PCA), leaving only the most significant information related to the 13 factors. Then, unsupervised machine learning determined the optimal number of clusters best representing the underlying behavioural distribution of the survey and then revealed 4 distinct archetypes, materializing a complete data-driven approach. In the analysis, we observed non-linear asymmetric outcomes, translating into a K-shaped behavioral and economic recovery, where some groups recover or improve while others fall further behind(Bonaccorsi, G. et al., 2020 ; U.S. Bureau of Labor Statistics, 2025 ; Egger, D. et al., 2021 ; Decerf, B. et al., 2021 ). Hence, we propose a conceptual framework of behavioral adaptation under three structural determinants; Substitution capacity (IMF, 2021 ; OECD, 2021 ), Exposure necessity(Nisa, C. F. et al., 2021 ; Michaels, D. et al., 2024 ), and Support infrastructure (Durante, R. et al., 2021 ; Fraser, T. et al., 2022 ), to determine the adaptive capacity and explain the aforementioned responsive patterns. Thus, our study demonstrates that effective crisis response should build behavioral capacity, enabling more equitable adaptation and mitigating the disparities caused and intensified through inequality. Methods Data Source and Study Design Using the COVID-19 pandemic as a quasi-natural experiment, this study evaluates the mobility, behavioral adaptation, and underlying socioeconomic evolution with fast-changing external constraints through a cross-sectional nationwide household survey conducted covering all provinces in Sri Lanka, from November 6, 2021, to December 10, 2021(Ilangarathna, G. A. et al, 2024 ). To ensure access to both rural and urban settings, this was structured as a face-to-face Computer-Assisted Personal Interview (CAPI) survey, supported by national and local administrative authorities. The multi-domain questionnaire had 78 core items and 8 demographic questions to capture both socioeconomic and behavioral impacts of COVID-19 through three distinctly identified pandemic waves in Sri Lanka. The first wave was from 27th March 2020 to 3rd October 2020, the second wave was from 4th October 2020 to 14th April 2021, and the third wave was from 15th April 2021 to 31st December 2022 (Epidemiology unit, 2019 ). The main focuses included mobility and behavioral adaptation, income shocks, household economy, psychological changes, food consumption, digital access, and educational disruption, where this particular research focuses on mobility and behavioral adaptation. Participants and Sampling The geographic and demographic representativeness was ensured with multistage cluster sampling, with stratification by population density, COVID-19 risk severity, and urban, rural, and estate-sector communities. The minimum sample size of 2,401 was obtained using Cochran’s formula with a 95% confidence level, 2% margin of error, and p = 0.5. The survey included data on 3,020 households across 9 provinces and 20 districts (Ilangarathna, G. A. et al, 2024 ). The village officers facilitated purposive inclusion of pandemic-affected households. Data Preparation The final dataset used for the analysis contained 1,898 households after performing listwise deletion to remove incomplete or partially filled observations. The retained dataset had a consistent demographic structure with the national distribution as depicted in Fig. 1 . Although this is a cross-sectional survey, the questionnaire contained retrospective information across multiple pandemic phases, including before the pandemic, the first, second, and third COVID-19 waves. The Likert-scale questionnaire items were recorded on a symmetric directional scale (− 1.5, − 1, 1, 1.5) to reflect the magnitude and direction of behavioral change. The mean of the scores from the first, second, and third waves was calculated, and this was subtracted from the corresponding pre-pandemic value to quantify the pandemic-induced behavioral change. This is an established approach for assessing deviations from a baseline under exogenous shocks (Cohen, J et al., 2015 ; Menard, S. W., 2002 ). Variables The questionnaire contained 40 behavior and mobility-related items spanning remote work, digital engagement, commuting patterns, non-work mobility, and service access, which were theorized to represent latent domains such as household adaptation, digital substitution, and rebound in social interactions. Complementing those were the variables of education level, employment sector, household income, settlement (urban/rural/estate), and ethnicity. These socioeconomic variables aided in understanding the structural constraints that shaped behavioral distinctions. Factor Analysis Factor analysis reduces a large set of observed variables to a smaller set of unobserved latent factors. This process captures the essence of the data's structure. The first step involves PCA, which identifies components that represent the shared variance among the variables. Next, varimax rotation is applied. This orthogonal rotation method maximizes the variance of loadings for factors, thereby improving the interpretation of the results (Kaiser, H. F., 1958 ). This step creates uncorrelated factors with a clearer structure. It strengthens the connection between each factor and specific dimensions of the data, allowing for more meaningful interpretations. By simplifying the data, factor analysis boosts its reliability and reinforces its conclusions. Hence, it is a highly effective method for identifying the primary constructs underlying observed behaviors. In the procedure, only the items with factor loadings above 0.4 were retained because they demonstrate a strong connection to the underlying factor (Pett, M. A. et al., 2012 ), thereby improving the reliability of the factor solution. Additionally, items with cross-loadings where the difference was less than 0.1 across factors were removed (Acar Güvendi̇r, M. et al., 2022 ; Jamali, J. et al., 2018 ). Unsupervised Machine Learning: Spectral Clustering and K-Means Clustering Spectral clustering is a powerful unsupervised machine learning technique based on graph theory and linear algebra, which can be used for identifying clusters in data that may not be linearly separable in the high-dimensional feature space (Ng, Andrew et al., 2002 ). Compared to the traditional clustering algorithms, this tool is particularly effective in discovering non-convex and arbitrarily shaped clusters embedded in high-dimensional spaces. First, a similarity graph is obtained, where each node corresponds to a data point and edges are weighted according to pairwise similarities of the data with the use of a Gaussian (RBF) kernel function, as shown in Eq. ( 1 ). $$\:{A}_{ij}=\:{e}^{-\frac{{\Vert\:{x}_{i}-\:{x}_{j}\Vert\:}^{2}}{2{\sigma\:}^{2}}}$$ 1 Here, Aij symbolizes the commonality between the two data points x i​ and x j , and sigma ( \(\:\sigma\:\) ) is a tunable scaling parameter that controls the zooming effect in the standard spectral clustering algorithm. The value of sigma critically determines the structure of the affinity graph, where the small values capture fine-grained clusters, whereas large values connect distant points, highlighting broader global structures. Once the affinity matrix A is constructed, the degree matrix D is computed as a diagonal matrix containing the sum of affinities for each data point in each row, as shown in Eq. ( 2 ). $$\:{D}_{ii}=\:\sum\:_{j}{A}_{ij}$$ 2 Using A and D, the graph Laplacian is formed as in Eq. ( 3 ). $$\:L=\:{D}^{-\frac{1}{2}}\:\left(D-A\right){D}^{-\frac{1}{2}}$$ 3 The eigenvalues of the L ( \(\:{\lambda\:}_{i}\) ) encode the connectivity structure of the graph. Small eigenvalues correspond to well-connected components, while large eigenvalues indicate separations between clusters. Therefore, the difference between successive eigenvalues or the eigengap ( \(\:{\lambda\:}_{i}-\:{\lambda\:}_{j}\) ) provides a quantitative measure of cluster separability, thereby revealing the structure of the data distribution through clustering. A significant increase in the eigengap typically signifies a natural cluster boundary, with the index k of the largest gap suggesting the optimal number of clusters. Then, to capture this optimal number of clusters, the sigma sweep is performed. In this procedure, sigma ( \(\:\sigma\:\) ) is varied logarithmically across a wide range, and the eigengap curve is obtained for each \(\:\sigma\:\) value. This approach enables the identification of modes of clustering. Each mode corresponds to a stable eigengap region, indicating a particular cluster structure. As we zoom in on the cluster space by adjusting the free parameter \(\:\sigma\:\) , sub-clusters start to form within the feature space. It is possible to detect several small clusters submerged within the initial supercluster at various levels of zooming. Hence, the eigengap curve that remains dominant across the widest range of log ( \(\:\sigma\:\) ) is taken to represent the most robust and scale-consistent cluster configuration. Therefore, this process provides a principle, data-driven mechanism to determine the ideal number of clusters. In essence, the required number of clusters (k) for the most persistent cluster formation under zooming is taken as the fundamental organizational structure of the data distribution, which is indicated by the eigengap curve that predominates over a wide range of \(\:\sigma\:\) values. The k-means algorithm is an unsupervised machine learning algorithm that groups similar data points into k clusters by targeting to minimize the distance between data points within each cluster. Mathematically, it aims to minimize the objective function in Eq. ( 4 ): $$\:J=\:\sum\:_{i=1}^{k}\sum\:_{{x}_{j}\in\:{C}_{i}}{‖{x}_{j}-{\mu\:}_{i}‖}^{2}$$ 4 where C i denotes the i th cluster and µ i is its centroid. The algorithm proceeds iteratively through two primary steps: Assignment of each data point to its nearest centroid based on Euclidean distance. The centroids are updated with the recalculation as the mean of all data points assigned to each cluster. This process repeats until convergence, typically when centroid positions stabilize or the reduction in J falls below a threshold. Accordingly, in this study, the k-means clustering is applied to the feature space to obtain the final clusters C 0 ,…, C k−1 , after determining the optimal number of clusters (k) through the sigma-sweep-based spectral clustering procedure. Robustness and Sensitivity Analysis for the Factor Structure and Selection of the Optimal Cluster Number with \(\:\sigma\:\) Variation The robustness of the factor structure was supported by the acceptable internal consistency and construct validity. The Kaiser–Meyer–Olkin (KMO) value of 0.61 exceeded the minimum threshold of 0.60(Kaiser, H. F. et al., 1974 ), supporting the sampling adequacy for factor extraction. Confirmatory factor analysis was conducted using robust maximum likelihood estimation (Askelund, A.D. et al., 2019 ) with Huber–White robust standard errors (White, H., 1980 ) and Yuan–Bentler scaled test statistics(Yuan, K.-H. et al., 1980 ), accounting for deviations from multivariate normality (Mardia’s skewness = 188 935.12, kurtosis = 626.19, p < 0.001; Henze–Zirkler = 9.74, p < 0.001). The model had an acceptable incremental fit (robust CFI = 0.93, robust TLI = 0.90), consistent with the conventional thresholds (≥ 0.90 acceptable(Bentler, P. M., 1990 ; Tuominen, L.S. et al, 2022 ), ≥ 0.95 good(Hu, L. et al., 1999 )). The absolute fit of the model was good (robust RMSEA = 0.03 [90% CI 0.029–0.035], robust SRMR = 0.04), falling well below the conventional thresholds for good fit(Hu, L. et al., 1999 ) (RMSEA ≤ 0.06, SRMR ≤ 0.08). The scaled χ²/df ratio of 2.47 indicated a good absolute model fit, consistent with the literature recommending values below 3.0(Schermelleh-Engel et al., 2003 ; Kline, R.B., 2023 ). Model adequacy is best confirmed when theoretical plausibility exists alongside multiple complementary indicators(Stone, B. M., 2021 ). The 13 latent factors are theoretically justified as they operationalize the conceptual framework of behavioral adaptation. The tunable \(\:\sigma\:\) parameter in Spectral Clustering acts as a multi-scale lens that controls the level of granularity in the similarity graph, making the method particularly robust for identifying the optimal number of behavioral clusters in the high-dimensional survey dataset. By varying sigma, it progressively zooms through different cluster formations in the data, where small sigma values emphasize fine-grained, local similarities, while large sigma values capture broader, global relationships among data points. Observing the eigengap curve’s nature across this sigma sweep allows the detection of stable modes of clustering, i.e., regions where the eigengap remains consistently dominant, which indicates scale-invariant and structurally coherent clusters. This procedure mitigates sensitivity to initialization or noise, ensuring that the selected cluster number reflects the most persistent and intrinsic structure of the data rather than arbitrary local variations revealing the most robust and interpretable latent organization of the behavioural data. Integration of Statistical and Unsupervised Machine Learning with Behavioral Theory A key objective of this study was to expose the latent behaviour of human adaptation during a global disruption. Rather than considering individual mobility alone, we adopted a completely data-driven strategy with Factor Analysis and Unsupervised Machine Learning to identify underlying behavioral constructs consistent with the behavioral science theory that observed actions reflect deeper latent strategies and constraints rather than independent variables. Factor analysis enables the detection of multivariate behavioral patterns emerging from co-occurring routines and decision-making processes, aligning with socioecological perspectives where behavior is embedded in interdependent domains of daily life (e.g., social interaction, work exposure, digital engagement, access to resources). By extracting orthogonal, interpretable factors, we capture distinct adaptive subsystems, each representing a coherent response to environmental constraints, risk, and opportunity. This approach aligns with theories of behavioral ecology, which emphasize that individuals optimize behavior within contextual constraints and bounded rationality (Simon, H. A., 1955 ), whereby people act based on feasible options shaped by structural realities. Building on these latent dimensions, we applied a clustering framework to identify behavioral archetypes where discrete, emergent adaptation patterns represent strategic configurations of behavior, not just socioeconomic groupings. Using spectral clustering is theoretically appropriate because crisis behavior is often non-linear, non-hierarchical, and shaped by complex interactions between structural constraints and adaptive strategies. Spectral methods can detect non-convex cluster boundaries and subtle divisions in behavior space that linear methods may overlook, reflecting the possibility that similar behavioral outcomes (e.g., reduced mobility) may arise from different mechanisms (e.g., remote work vs. unemployment). The subsequent use of k-means for assignment supports interpretability and reinforces each archetype as a distinct mode of behavioral plasticity, inertia, substitution, or compensation. Crucially, this analytical design is grounded in behavioral theory. We assume that individuals adapt using available behavioral capacity defined by their substitution capacity (digital and occupational flexibility), exposure necessity, and support infrastructure. These determinants correspond to well-established constructs in behavioral economics, sociology, and social psychology, including agency vs. structural constraint, risk perception conditioned by necessity, coping strategies, and the socioecological model of behavior. Thus, our methodological choices are not merely statistical; they implement a theoretical framework in which behavior is viewed as situated, structured, and adaptive. The resulting archetypes reveal how structural inequality becomes behavioral inequality, consistent with the emerging recognition that crises produce K-shaped trajectories of adaptation and recovery. Integration of the latent variable modeling with unsupervised clustering, conditioned on the behaviour changes triggered by the pandemic, revealed a hidden structure in the data organization, moving beyond a descriptive mobility analysis and providing a theoretically grounded, data-driven map of human adaptation, qualified by socioeconomics, demographics, and numerous other factors as highlighted in Fig. 2 . This allows us to identify not only what behaviors occurred, but why they occurred, how they co-occur, and how they reflect deeper mechanisms of resilience, vulnerability, substitution, dissent, and protest. Thus, the statistical model becomes an explicit test of behavioral theory and a foundation for conceptual generalization. Results Core Dimensions of Mobility and Behavioral Adaptation We employed PCA together with exploratory factor analysis using Varimax rotation to reduce the information in the high-dimensional survey dataset into 13 principal factors or interpretable behavioral dimensions (Table 1 ) that collectively summarize the structure of human mobility and adaptation under pandemic constraints. We determined which factors to retain based on the Kaiser criterion (Kaiser, H. F., 1960 ) and examined the scree plot, aiming to maintain simplicity while preserving explanatory strength as indicated in Fig. 3 . Collectively, these factors presented accounted for 63.17% of the total variance (Hair, J. F. J. et al., 2018 ) in the behavioral data while remaining conceptually aligned with mechanisms of decision-making under constraint, social capital, infrastructural dependency, and digital capacity, providing a theoretically meaningful basis for studying how people reorganize behavior when risk, rules, and resources change. Factor-loadings in Table 1 link the survey questions to the extracted latent behavioural factors, where each loading represents the strength of association between a question that highlights an observed behavioural indicator and its corresponding factor, revealing distinct domains of human mobility and interaction. They highlight subsystems of everyday behavior, from leisure and social connection to work exposure strategies, infrastructure access, transport portfolios, local resource ecologies, spatial routines, and technology-mediated activity. Emergence of behavioral archetypes through clustering Following factor analysis, we identified 4 robust behavioral archetypes via a two-stage clustering pipeline. The initial step involved using PCA to reduce the original dataset's dimensionality to 13 dimensions to denoise and enhance the data for subsequent clustering. Spectral clustering was then applied to this refined dataset to determine the optimal number of latent clusters (k), as illustrated in Fig. 4 . Finally, K-means clustering was employed to assign each data point to one of the clusters, which captured distinct strategies of adaptation rather than being mere statistical groupings, with privilege-enabled substitution into remote work, tradition-anchored risk management, stability-seeking minimalism under budget constraints, and necessity-driven communal resurgence as shown in Fig. 5 . The first Cluster 0, the “Privileged Adaptors,” comprises individuals with strong structural and demographic privilege, being among the Sinhalese majority, and with the highest educational qualifications. They demonstrated behavioural flexibility and resilience, maintaining middle-to-higher income and formal employment in secure sectors, reflecting greater institutional and economic access. Hence, they empowered a rapid transition to remote work, enabled by digital readiness and occupational flexibility, leading to a marked decline in physical commuting and in-person interactions. Mobility for essential needs reflected resilient maintenance behaviours. Their elevated education and economic stability facilitated informed risk assessment and deliberate behavioural restraint, underscoring how privilege enabled both the capacity and the choice to minimize exposure. Cluster 1, the “Cautious Traditionalists,” represents a lower-middle class with moderate education. Their mixed ethnic composition reflected typical mid-tier socioeconomic diversity in Sri Lanka. They adopted a risk-averse stabilization strategy grounded in traditional, localized adaptation and showed a high income volatility, i.e., many slipped into low-income brackets during early waves and recovered only partially. Compared to Cluster 0, the limited occupational and digital flexibility of Cluster 1 constrained their adaptive options. Their strong pre-pandemic social participation shifted into a highly cautious re-engagement, driven by economic vulnerability and risk aversion, relying moderately on private transport with a sustained drop in public transport usage. Their high home delivery and mobile shops usage reflected proximal sourcing and local dependency as protective strategies. Cluster 2, the “Stable Minimalists,” represents the financially disrupted lower-middle class. With low educational attainment and informal employment, they reflect limited adaptive capacity. Their mixed ethnic composition also mirrors the national socioeconomic middle in Sri Lanka, similar to cluster 1. They demonstrated strong behavioral inertia and necessity-driven resilience. Many fell into low-income brackets from the lower-middle-income tier, with only partial recovery. Although cluster 2 exhibited how the other socioeconomic constraints shaped their rigid maintenance behaviours: private commuting remained minimal, while active mobility such as walking and cycling for work persisted, reflecting economic necessity rather than adaptive choice. Their social participation and digital engagement stayed consistently low, reinforcing a social and digital divide. With moderately stable jobs, they adhered to a budget-friendly financial routine. Collectively, this group embodies constrained behaviour and continuity of low-cost, proximal routines. Cluster 3, the “Resilient Rebounders,” embodies the most socioeconomically vulnerable, displaying behavioral rebound amid severe structural fragility. Persistently in the low-income tier, they rely on daily wages and informal employment, with the lowest educational attainment and minimal digital capacity. Demographically, this is the most ethnically diverse and marginalized cluster, with strong minority representation, reflecting concentrated systemic vulnerability. Despite lacking financial or digital buffers, this group showed a powerful mobility resurgence with social and recreational participation rebounding sharply, suggesting social resilience and collective recovery. Their active commuting and public transport usage rose substantially, driven by economic necessity and communal interdependence. Socioeconomic and demographic mechanisms underlying archetypes A clear socioeconomic division emerges across the identified behavioural clusters, driven by unequal access to education, formal and informal employment (Fig. 6 c, 6 d). The privileged adaptors hold the highest academic attainment, with nearly 30% with advanced or university degrees and secure positions in government or private sectors (71.8%), providing occupational stability and virtual substitution capacity. The cautious traditionalists, with mixed schooling and middle-tier income, exhibit limited credentialing and moderate in-person exposure, managing risk through localized coping. The stable minimalists and resilient rebounders faced constraints in adaptive flexibility and reinforced cost-sensitive, necessity-driven mobility. Ethno-geographic patterns amplify these structural disparities (Fig. 6 a, 6 e); Cluster 0 is predominantly Sinhalese and urban, while Cluster 3 is more ethnically diverse and rural, overrepresented among Tamil and Muslim communities in underdeveloped districts. Thus, behavioral capacity is stratified by structural opportunity, with education, occupation, and geography collectively anchoring the inequality-adaptation nexus in human mobility. Income trajectories and economic resilience across waves Household income trajectories reveal deep asymmetries in economic resilience across archetypes (Fig. 6 b). Pre-pandemic inequalities shaped exposure to shocks and recovery potential. The privileged adaptors, concentrated in Middle-to-Higher income categories (> 85%), absorbed an initial first-wave decline but rapidly recovered, supported by formal employment, digital access, and substitution. The cautious traditionalists showed a sharper contraction, with Low income increasing from 4.9% to 33.4%, and an incomplete rebound by the third wave, marking fragile middle-class stability and constrained adaptive flexibility. The Lower-Middle income of the stable minimalists suffered persistent downgrades, as the Low-income share rose from 4.8% to 35.2% with limited recovery, reflecting dependence on on-site, low-buffer occupations and restricted financial adaptability. The resilient rebounders experienced the most severe and lasting collapse, shifting from 47.2% in the Lower-Middle category to 68.7% in the Low-income category, aligning with informal, daily-wage labor and income, behavior decoupling, and mobility rebound without financial recovery. Their structural downgrade, starkly contrasting with national trends, which recorded a poverty rate of 11.3% in 2019 to 13.1% in 2021 (World Bank Group, 2023 ), isolates Cluster 3 as the most vulnerable group. These trajectories expose how structural precarity and employment informality entrench vulnerability, while privilege sustains economic and behavioral resilience across crises. Behavioral Adaptation Across Pandemic Phases Temporal evolution of the extracted behavioural factors across the pre-pandemic and the three major waves of COVID-19 (Fig. 7 ) captures the heterogeneous responses of different societal clusters. Social and recreational mobility exhibited a universal decline across all clusters during the first pandemic wave (Fig. 7 a), indicating widespread compliance with movement restrictions and collective behavioural suppression, with heterogeneous rebounds thereafter. Notably, Clusters 1 and 3 demonstrated stronger recovery trajectories, reflecting, respectively, social reconnection managed by risk-averse routines and necessity-driven return to communal life, whereas Clusters 0 and 2 remained consistently suppressed, reflecting sustained caution and restricted social participation due to voluntary caution and resource-bounded routine, respectively. Cluster 1’s recovery aligns with its pre-existing social connectivity and active community engagement, while Cluster 3, initially a less recreationally active cluster, displayed a more spirited rebound, potentially reflecting the compensatory pursuit of social interaction following prolonged restriction. In contrast, Cluster 0 maintained a cautious lifestyle with limited exposure, and Cluster 2 preserved habitual routines with minimal change. These divergent rebound paths illustrate that social recovery is not uniform but stratified by socioeconomic conditions, social capital, and risk perception, suggesting further that elevated risk perception among more privileged groups, while promoting safety, may simultaneously foster overcaution and social isolation, potentially heightening experiences of loneliness, emptiness, and be more inclined to use social media as coping mechanisms within these households(Senarath, N. et al. , 2024 ; Thilakasiri, I. et al. , 2025 ). Private commuting and work exposure revealed stratification reflecting class-based occupational flexibility (Fig. 7 b). Clusters 0 and 3 converged at lower mobility levels during the first wave, but for opposing reasons: Cluster 0 benefited from the privilege of remote work, while Cluster 3 likely experienced unemployment or underemployment. Clusters 1 and 2 maintained moderate exposure, consistent with their continued on-site work obligations. This divergence underscores how both privilege and precarity can manifest as reduced mobility, albeit through entirely different mechanisms. The persistence of these differentiated mobility trajectories through subsequent waves highlights the embedded socioeconomic inequalities shaping pandemic-era behavioural adaptation. Home-based work isolation exhibited sharp polarization across clusters (Fig. 7 c). Cluster 0 experienced a pronounced rise during the first wave, reflecting a rapid transition to remote work enabled by occupational and digital privilege. Other clusters remained largely static, evidencing limited remote-work feasibility and consequent in-person exposure by obligation (stable minimalists), by partial necessity with caution (cautious traditionalists), or by precarity (resilient rebounders). This divide underscores the emergence of a digital and occupational inequality, where remote work became a behavioural marker of both safety and privilege. Both financial and essential service access demonstrated marked resilience for Clusters 0 and 1, who maintained steady or slightly increasing mobility across waves (Fig. 7 d,f). In contrast, Clusters 2 and 3 exhibited sustained suppression, reflecting economic vulnerability and constrained access to infrastructure. The stability of access-related mobility among higher-income groups reveals the buffering effect of resource availability, while the contrast in lower-income groups points to a disproportionate impact of logistical and financial disruptions. Unlike the earlier factors, which had a nonlinear, complex relationship to economic status, this factor showcased a linear variation unraveled by the detailed data-driven cluster analysis. Mobility related to public and hired transport declined sharply across all clusters during the first wave (Fig. 7 e,i). However, only Cluster 2, characterized by essential work and limited alternatives, demonstrated partial recovery, reflecting the necessity-driven dependence on shared mobility. Clusters 0 and 1 maintained suppressed usage, substituting private or remote modes of work, illustrating a socioeconomic gradient in transport resilience, where exposure risk was unequally distributed according to occupational necessity and economic means. Bicycle mobility and work exposure (Fig. 7 g,h) reveal how occupational demands shape behavioural adaptation across socioeconomic groups. Cluster 2 demonstrates persistent physical commuting and high interpersonal exposure, reflecting necessity-driven resilience under limited remote work options. In contrast, clusters 0 and 3 maintain low levels of mobility and exposure, cluster 0 benefiting from digital flexibility, while cluster 3 faces unemployment. Access to community resources and group transport displayed complementary behavioural patterns (Figs. 7 j,k). Cluster 1 showed sustained engagement with local networks frequenting delivery services and neighborhood shops, indicating strong communal ties and adaptive social capital. Meanwhile, Clusters 0 and 2 exhibited consistently low community interaction, with Cluster 0 relying on digital systems and Cluster 2 constrained by economic and spatial limitations. Shared transport mobility remained low throughout, highlighting collective risk aversion and reduced interpersonal proximity. Active commuting, such as walking to workplaces, remained remarkably stable for Cluster 2 across all pandemic waves (Fig. 7 l). This implies either a structural necessity or a habitual attachment to routine movement in low-radius livelihoods. Hence, indicating a form of behavioural inertia rooted in occupational rigidity and limited adaptive flexibility. Digital platform engagement of Cluster 0 increased significantly, and it was closely followed by Cluster 1, whose rise still surpassed that of Clusters 2 and 3 during the first wave and plateaued thereafter (Fig. 7 m). Clusters 0 and 1 leveraged digital tools for work, shopping, and communication, reflecting both educational attainment and infrastructural access. In contrast, Clusters 2 and 3 demonstrated minimal digital engagement, underscoring persistent digital divides that parallel economic stratification. The resulting behavioural dichotomy reflects how technology-mediated adaptation became both a buffer and a barrier across socioeconomic tiers. These findings show that effective policy must recognize that divergent behaviours often stem from structural constraints rather than choice, and that risk-mitigation strategies evident for some may be inaccessible or impractical for others (Iio, K. et al., 2021 ; Carranza, A. et al. , 2022 ; Chiou, L. et al., 2020 ). Interdependencies and behavioral trade-offs Behavioral coupling across the system underscores how digital access, work security, and social capital mediated behavioral flexibility, exposing the feedback loops between risk, resources, and routine. Cluster 0 demonstrated the strongest adaptive capacity with a sharp reduction in Private Commuting and Work Exposure and an increment in Home-Based Work Isolation, facilitating a rapid transition to remote work. This was accompanied by decreased Bicycle Mobility and Variable Work Exposure, and a surge in Digital Platform Engagement, a key enabler for maintaining occupational and financial continuity through online services. For Clusters 2 and 3, digital scarcity forced physical presence and transport substitution, increasing exposure and income instability, thereby reinforcing structural precarity. Financial Access Points and Essential Services disruptions of Cluster 3 were compensated by reliance on Community Resource Sourcing. For Cluster 1, community capital acted as a partial hedge, enabling a balance between exposure and livelihood through localized, traditional coping mechanisms. Emergent conceptual model of crisis adaptation We propose a conceptual model of behavioral adaptation under collective disruption with the following determinants which shape whether individuals exhibit behavioral plasticity or inertia, experience agency or compulsion, and achieve resilience or cumulative disadvantage (Fig. 8 ): 1. Substitution capacity (digital access, occupational flexibility) determines the range of practical behavioral substitutes for risky activities. 2. Exposure necessity (formal/informal labor structures, transport portfolios) influence the cost of sustaining livelihoods when mobility is limited. 3. Support infrastructure (local networks, institutional access, urbanity) influences resilience as opposed to path-dependence. Within this framework, inequality is reframed not only as a difference in income or resources, but as a difference in behavioral capacity, which determines the ability to opt for safer alternatives when circumstances change. This reconceptualization aligns with socioecological theories of behavior while adding further extensions in a new crisis-specific context. Discussion The mobility and behavioral distinctions during the pandemic show a complex relationship between common shock responses and deep-rooted socioeconomic divisions (van de Weijer, M. P. et al., 2022 ; Yabe, T. et al, 2024 ; Boz, H. A. et al. , 2024 ; Bavel, J.J.V. et al. , 2020 ; Delaney, L. , 2024 ). The data-driven analysis has revealed that behavioral responses to the pandemic were characterized by significant non-linearities relative to socioeconomic status. While a clear gradient in income capacity is evident across clusters 0 to 3, other critical dimensions, such as resilience in leisure activities, digital adaptation, and health-protective behaviors, deviate from this linear progression. For example, the most economically vulnerable Cluster 3 exhibited a stronger recovery in social and recreational mobility than the other groups. This illustrates that human responses to the crisis are mediated by psychological resilience, cultural norms, access to networks, and structural necessity, not just economic position, challenging simplistic assumptions linking poverty and suppressed behaviors. The existence of these tiered, non-proportional relationships exposes a multifaceted societal structure that was laid bare by the pandemic's disruptive force. The pandemic revealed at least two distinct behavioral logics: proactive substitution and reactive compensation. Privileged Adaptors exemplified the first logic by immediate digital substitution. They engaged in behavioral plasticity facilitated by high substitutability, through remote work, online banking, and virtual social interaction. This aligns with theories of agentic flexibility, where they are often able to restructure routines to maintain stability. The rapid rise in Home-Based Work Isolation and Digital Platform Engagement, paired with reductions in work exposure and physical mobility, redefined daily life through technology. This group showed bounded rationality under low constraints in addition to high perceived control, allowing them to prioritize health without sacrificing income. However, Stable Minimalists and Resilient Rebounders showed behavioral inertia driven by structural constraint. Their lack of access to resources and opportunities translated to their choices being necessity-driven responses as opposed to conscious behavioral choices. They were required to be physically present even during critical pandemic phases to obtain an income. This aligns with theories of bounded rationality, where individuals operate with constrained freedom of decision-making owing to inequalities. Our analysis interprets this phenomenon as risk absorption produced by limited agency, as opposed to risk-seeking, demonstrating that risk perception is not merely psychological, but structurally conditioned. Additionally, the analysis exposed situations where different clusters exhibited similar adaptive behaviours, yet for fundamentally different reasons, where the shared strategies emerged from distinct structural constraints, motivations, and resources unique to each group. As an instance, both Clusters 0 and 3 had a sharp decline in public transport usage during the first wave. While Cluster 0’s reasoning was voluntary avoidance enabled by remote work, Cluster 3 experienced involuntary exclusion caused by job loss and service disruption. Hence, the same mobility reduction diverged into opposite outcomes, where Cluster 0 rapidly recovered economically, whereas Cluster 3 experienced prolonged economic decline. This shows that behavior cannot be understood in isolation from the structural conditions under which it occurs. Another key finding was the crucial role of digital capital as a determinant of behavioral flexibility. Clusters 0 and 1, having higher educational levels and technological literacy, used digital access as a substitution technology, switching to virtual equivalents for work and other needs. Clusters 2 and 3 had a contrasting experience with limited ability to switch to digital resources, compelling them to rely on communal resource sharing or government and NGO support. Social capital also emerged as an alternative form of adaptive capacity in the absence of digital infrastructure. For example, Cluster 1 utilised localized community networks and informal delivery systems to deal with disruption, relying on relational infrastructure rather than shifting to digital platforms. The coping mechanisms to achieve stability for Cluster 3 were reliance on government/NGOs and re-engagement in social gatherings, demonstrating that communal strategies can function as parallel adaptive systems, specifically when digital alternatives are inaccessible. Cluster 2 displayed an adaptation strategy based on preserving routine as opposed to transformation, representing behavioural inertia rooted in occupational immobility, economic necessity. This was prominent through Factors 2, 7, and 12, where they had consistent movements, moderate work exposure, and relatively low digital engagement. Crucially, this points to an interesting observation that the absence of change itself is a behavioural outcome shaped by structural changes. The dynamics of poverty traps and behavioural lock-in are further highlighted through the contrasting responses between Clusters 2 and 3, even though both lacked digital access and higher education. While Cluster 2 maintained low-cost, proximate routines offering small possibilities for upward movement, Cluster 3 showed high physical mobility, in the face of greater economic shocks, as a mechanism to rebuild social and economic stability. However, rather than translating into financial recovery, this led to a structural downgrade in income, even after the superficial recovery of behavioural patterns. This created insights on how behavioural resilience does not guarantee socioeconomic resilience, especially when underlying structures do not support upward movement, hence masking vulnerability. Hence, the clusters demonstrate that adaptive strategies are systematically patterned by the intersection of economic, digital, and social capital. While privileged groups showed individualized, technology-enabled adaptation, moderately resourced groups utilised community-based coping mechanisms, and the more structurally under-privileged individuals had to rely on behavioural inertia or necessity-driven exposure. A socioecological model of behavioural adaptation was supported through these findings, revealing a hierarchy of adaptive freedom, where more resources translated to a wider range of possible strategies. Individual resources, community networks, and institutional access were also factors that contributed in leading to these outcomes. Our findings further highlight that the most powerful determinant of behaviour in crisis is structural opportunity for viable strategies. By integrating these insights, we propose a behavioral framework for crisis adaptation under inequality. At its core, the framework suggests that adaptation capacity is determined by three interacting elements: substitution capacity, exposure necessity, and support infrastructure. Cluster 0 had high substitution capacity, low exposure necessity, and stable support, producing rapid, protective adaptation, enabling the risk aversion to become the Privileged Adaptors. Cluster 1 had moderate substitution and strong community support, producing cautious adaptation. Cluster 2 had low substitution and high exposure necessity, producing behavioral inertia. Cluster 3 had minimal substitution and high exposure necessity, but high communal support, producing a reactive but ultimately economically costly rebound. These differentiated pathways underscore that behavioral flexibility is not a trait but a privilege, which is a function of the resources and structures that surround individuals. Although these identified mechanisms are based on a Sri Lankan context, they encapsulate globally relevant mechanisms of inequality: technological substitutability, livelihood-based exposure, and the buffering role of formal and informal support, which are generalizable beyond Sri Lanka. It offers a transferable lens for understanding heterogeneous crisis adaptations in other countries, where similar constraints shape the behavioral trajectories that emerge under large-scale societal shocks. For example, digital access determined who could maintain economic stability (Francis, D. V. et al., 2021 ; Jahan, N. et al., 2023 ) across the globe. While high-income employees easily switched to remote work, the less fortunate, such as immigrant labourers and minority communities worldwide, had to rely on community or NGO support in the absence of institutional safety nets (Fan, B. E. , 2021 ). Crisis reveals and accelerates pre-existing fault lines in behavioural opportunity spaces as opposed to creating new divisions. The structural inequalities are amplified in a feedback loop (Marmot, M. et al., 2020 ; Kniffin, K. M. et al., 2021 ; Li, H. et al., 2023 ; Bambra, C. et al., 2020 ). Insights for policy and societal resilience are also revealed through our findings. They show that compliance is conditional on structural feasibility, whereas often public health strategies assume voluntary compliance. But the compliance rooted in capacity, flexibility, and necessity can result in resistance at times. In the face of challenges such as a lack of digital access, livelihoods depending on physical presence, simple instructions were neither effective nor sufficient. Therefore, effective policy should focus on expanding behavioural capacity through the improvement of digital infrastructure, enabling remote work opportunities, and strengthening social safety nets. Systems must be designed to facilitate adaptation, as opposed to expecting individuals to adapt to systems. This study advances human behavioural science by demonstrating that adaptation during crisis is shaped primarily by structural opportunity and the knowledge, awareness, and ideology, rather than being driven by individual preference alone. Behavioural flexibility was shown to be a form of privilege, whereas structural constraints were the cause for risk exposure and behavioural inertia rather than choice. Our behavioural framework explains why crises produce K-shaped recoveries and why resilience can coexist with vulnerability, shifting the focus from “what people do” to “what people are able to do”. This reconceptualization opens new directions for behavioural research and policy: rather than attempting to enforce uniform behaviors, effective interventions must build the structural conditions that enable equitable adaptation. One limitation of our study is that it may not fully capture cultural nuances or political and psychological motivations, as the data are from the Sri Lankan context. However, it is important to note that the structural mechanisms identified are consistent with findings across diverse global settings. Future research should test this framework across cultures while exploring how long-term behavioural changes persist or change once the crisis subsides, and there is also a need to integrate behavioural data into the process of policy making, focusing on which structural changes are capable of bringing forth the most effective adaptive capacity enhancements for the vulnerable. In conclusion, the pandemic reconfigured the entire architecture of human behaviour, revealing how adaptation is deeply connected to structural inequalities. Our findings offer a conceptual model of crisis adaptation, highlighting structural opportunity, digital and employment capital, and social networks as core determinants of human behaviour, while also illustrating that behavioural patterns are non-linear, stratified, and interdependent. This study sheds light on the mechanisms that rule adaptation, contributing to a deeper understanding of mobility under constraints, with the potential to aid in designing more resilient societies. Declarations Data availability The dataset used for the research is accessible in Harvard Dataverse, named “A dataset on the socioeconomic and behavioural impacts in Sri Lanka through multiple waves of COVID-19”. (Ilangarathna, G. A. et al, 2023). Competing interests The authors declare no competing interests. Ethical approval With assistance from the Department of Sociology, this study received ethical clearance from the Ethical Review Committee, Faculty of Arts, University of Peradeniya, Sri Lanka (ARTS/ERC/2021/01, September 18, 2021). Additionally, the Administrative clearance was obtained from the relevant Divisional Secretariat (DS) offices, Grama Niladari (GN) offices, and Sri Lanka's Ministry of Home Affairs, for the involvement of human participants. Informed Consent Every participant provided written and verbal informed consent, and their voluntary involvement was guaranteed. Following established procedures and guidelines, data confidentiality and privacy were preserved throughout the whole data collection process (Ilangarathna, G. A. et al, 2024). Author Contributions Statement M.K.– Conceived and designed the experiments, Performed the experiments, Analyzed the data, Contributed materials/analysis tools, Wrote the paper. L.E. - Performed the experiments, Analyzed the data, Contributed materials/analysis tools, Wrote the paper. I.M.- Analyzed the data, Contributed materials/analysis tools, Wrote the paper. R.G. - Conceived and designed the experiments, Analyzed the data, Wrote the paper. P.E. - Wrote the paper. N.K. - Analyzed the data, Wrote the paper. S.D. - Wrote the paper. V.H - Wrote the paper. J.E. - Analyzed the data, Wrote the paper. References Acar Güvendi̇r M, Özer Özkan Y (2022) Item removal strategies conducted in exploratory factor analysis: A comparative study. Int J Assess Tools Educ 9:165–180 Agampodi S (2021) Enhancing the Sri Lankan response to COVID-19 - a communique from the College of Community Physicians of Sri lanka. J Coll Community Physicians Sri Lanka 27:271 Ahmed SA, Majoka Z, Saha J (2023) Revisiting Informality in South Asia: An Agenda for the Future Social Protection and Jobs. Policy Note. Partnership for Economic Inclusion Andronico A et al (2021) Evaluating the impact of curfews and other measures on SARS-COV-2 transmission in French Guiana. Nat Commun 12 Annual Report (2020) : Central Bank of Sri Lanka. Annual Report 2020 | Central Bank of Sri Lanka Available at: https://www.cbsl.gov.lk/en/publications/economic-and-financial-reports/annual-reports/annual-report-2020 (Accessed: 2nd November 2025) Annual report 2022 (2025) Central Bank of Sri Lanka. Annual Report 2022 | Central Bank of Sri Lanka Available at: https://www.cbsl.gov.lk/en/publications/economic-and-financial-reports/annual-reports/annual-report-2022 . (Accessed: 2nd November Askelund AD, Schweizer S, Goodyer IM et al (2019) Positive memory specificity is associated with reduced vulnerability to depression. Nat Hum Behav 3:265–273. https://doi.org/10.1038/s41562-018-0504-3 Bambra C, Riordan R, Ford J, Matthews F (2020) The COVID-19 pandemic and Health Inequalities. J Epidemiol Commun Health 74:964–968 Bavel JJV, Baicker K, Boggio PS et al (2020) Using social and behavioural science to support COVID-19 pandemic response. Nat Hum Behav 4:460–471 Bentler PM (1990) Comparative fit indexes in structural models. Psychol Bull 107:238–246 Bonaccorsi G et al (2020) Economic and social consequences of human mobility restrictions under covid-19. Proceedings of the National Academy of Sciences 117, 15530–15535 Boz HA et al (2024) Investigating neighborhood adaptability using Mobility Networks: A case study of the COVID-19 pandemic. Humanit Social Sci Commun 11 Bronfenbrenner U (1979) The ecology of human development: Experiments by nature and design. Harvard University Press, Cambridge, MA Bronfenbrenner U (1977) Toward an experimental ecology of human development. Am Psychol 32:513–531 Carranza A et al (2022) The Social Divide of social distancing: Shelter-in-place behavior in Santiago during the covid-19 pandemic. Manage Sci 68:2016–2027 Chiou L, Tucker C (2020) Social distancing, internet access and inequality 10.3386/w26982 Cohen J, Cohen P, West SG, Aiken LS (2015) Applied multiple regression. Routledge Covid-19 cases | WHO COVID-19 Dashboard (2025) World Health Organization Available at: https://data.who.int/dashboards/covid19/cases . (Accessed: 1st November Decerf B, Ferreira FHG, Mahler DG, Sterck O (2021) Lives and livelihoods: Estimates of the global mortality and poverty effects of the COVID-19 pandemic. World Dev 146:105561 Delaney L, Watson E (2024) Behavioural science and the Irish covid-19 response. Behav Econ Policy Pandemics 220–231 Durante R, Guiso L, Gulino G (2021) Asocial capital: Civic culture and social distancing during COVID-19. J Public Econ 194:104342 Egger D et al (2021) Falling living standards during the COVID-19 crisis: Quantitative evidence from nine developing countries. Sci Adv 7 Epidemiology unit coronavirus disease 2019 (COVID-19) Available at: https://www.epid.gov.lk/epid/public/storage/post/pdfs/en_63ff1e0b4323b_sitrep-sl-en-06-09_10_21_1.pdf . (Accessed: 3rd November (2025) Fan BE (2021) Migrant workers with covid-19: Recognizing the crucial role non-governmental organizations perform. Lancet Reg Health - Western Pac 9:100145 Ferwana I, Varshney LR (2021) Social capital dimensions are differentially associated with covid-19 vaccinations, masks, and physical distancing. PLoS ONE 16 Francis DV, Weller CE (2021) Economic inequality, the digital divide, and remote learning during COVID-19. Rev Black Political Econ 49:41–60 Fraser T, Page-Tan C, Aldrich DP (2022) Social Capital’s impact on COVID-19 outcomes at local levels. Sci Rep 12 Galgamuwa LS et al (2023) Spatial distribution of COVID-19 patients in Sri Lanka. BMC Public Health 23 GDP growth (annual %) (2025) - sri lanka. World Bank Open Data Available at: https://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?locations=LK . (Accessed: 2nd November Gozzi N et al (2021) Estimating the effect of social inequalities on the mitigation of COVID-19 across communities in Santiago de Chile. Nat Commun 12 Hair JFJ, Black WC, Babin BJ, Anderson RE (2018) Multivariate Data Analysis. Cengage Learning EMEA Household income and expenditure survey – 2019 Available at: https://www.statistics.gov.lk/Resource/en/IncomeAndExpenditure/HouseholdIncomeandExpenditureSurvey2019FinalResults.pdf . (Accessed: 2nd November (2025) Hu L, Bentler PM (1999) Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equation Modeling: Multidisciplinary J 6:1–55 Iio K, Guo X, Kong X, Rees K, Wang B (2021) Covid-19 and social distancing: Disparities in mobility adaptation between income groups. Transp Res Interdisciplinary Perspect 10:100333 Ilangarathna GA et al (2024) A dataset on the socioeconomic and behavioural impacts in Sri Lanka through multiple waves of covid-19. Data Brief 53:110063 Ilangarathna GA et al (2023) Socioeconomic Impact of COVID-19 during three pandemic waves. Harvard Dataverse https://doi.org/10.7910/DVN/CXMJSM International Labour Organization (ILO) (2021) *ILO Monitor: COVID-19 and the world of work. Seventh edition*. https://www.ilo.org/wcmsp5/groups/public/---dgreports/---dcomm/documents/briefingnote/wcms_767028.pdf Jahan N, Zhou Y (2023) Covid-19 and Digital Inclusion: Impact on Employment. J Digit Econ 2:190–203 Jamali J, Ayatollahi S, Jafari P (2018) The effect of cross-loading on measurement equivalence of psychometric multidimensional questionnaires in Mimic Model: A simulation study. Materia Socio Med 30:121–126 Jay J et al (2020) Neighbourhood income and physical distancing during the COVID-19 pandemic in the United States. Nat Hum Behav 4:1294–1302 Kaiser HF, Rice J (1974) Little jiffy, Mark IV. Educ Psychol Meas 34:111–117 Kaiser HF (1960) The application of electronic computers to Factor Analysis. Educ Psychol Meas 20:141–151 Kaiser HF (1958) The Varimax criterion for analytic rotation in factor analysis. Psychometrika 23:187–200 Kline RB (2023) Principles and practice of structural equation modeling. Guilford Kniffin KM et al (2021) Covid-19 and the workplace: Implications, issues, and insights for future research and action. Am Psychol 76:63–77 Li H, Wei YD (2023) Covid-19, cities and inequality. Appl Geogr 160:103059 Marmot M, Allen J (2020) Covid-19: Exposing and amplifying inequalities. J Epidemiol Commun Health 74:681–682 Menard SW (2002) Longitudinal research. SAGE Michaels D, Spieler EA, Wagner G (2024) R. US workers during the COVID-19 pandemic: Uneven risks, inadequate protections, and predictable consequences. BMJ 384 Mid-year population by district and sex (2024) Scribd Available at: https://www.scribd.com/document/784118217/Mid-year-population-by-district-and-sex-2024 . (Accessed: 3rd November 2025) Ng A, Jordan M, Weiss Y (2002) On Spectral Clustering: Analysis and an algorithm. Adv Neural Inf Process Syst 14 Nisa CF et al (2021) Lives versus livelihoods? perceived economic risk has a stronger association with support for covid-19 preventive measures than perceived health risk. Sci Rep 11 Onyeaka H et al (2021) Covid-19 pandemic: A review of the global lockdown and its far-reaching effects. Sci Prog 104 Pan Y et al (2020) Quantifying human mobility behaviour changes during the COVID-19 outbreak in the United States. Sci Rep 10 Pangallo M et al (2023) The unequal effects of the health–economy trade-off during the COVID-19 pandemic. Nat Hum Behav 8:264–275 Petherick A et al (2021) A worldwide assessment of changes in adherence to COVID-19 protective behaviours and hypothesized pandemic fatigue. Nat Hum Behav 5:1145–1160 Pett MA, Lackey, Nancy R, Sullivan JJ (2012) Making sense of factor analysis: The use of factor analysis for instrument development in Health Care Research. SAGE Population density (people per sq (2025) km of land area) - South Asia. World Bank Open Data Available at: https://data.worldbank.org/indicator/EN.POP.DNST?locations=8S . (Accessed: 2nd November Santana C et al (2023) Covid-19 is linked to changes in the time–space dimension of human mobility. Nat Hum Behav 7:1729–1739 Schermelleh-Engel K, Moosbrugger H, Müller H (2003) Evaluating the Fit of Structural Equation Models: Tests of Significance and Descriptive Goodness-of-Fit Measures. Methods Psychol Res Online 8:23–74 Senarath N et al (2024) Family income does not necessarily lead to a holistic e-learning experience during COVID-19: A study in Sri Lanka. Humanit Social Sci Commun 11 Simon HA (1955) A behavioral model of rational choice. Q J Econ 69:99 Sri Lanka Socio Economic Data 2021 Available at: https://www.cbsl.gov.lk/sites/default/files/cbslweb_documents/publications/otherpub/publication_sri_lanka_socio_economic_data_folder_2021_e.pdf . (Accessed: 3rd November (2025) Stone BM (2021) The ethical use of fit indices in structural equation modeling: Recommendations for psychologists. Front Psychol 12 Teleworking in the COVID-19 pandemic: Trends and prospects. OECD Policy Responses to Coronavirus (COVID-19) (2021) 10.1787/72a416b6-en The K-shaped recovery (2025) Examining the diverging fortunes of workers in the recovery from the COVID-19 pandemic using business and household survey microdata. U.S. Bureau of Labor Statistics Available at: https://www.bls.gov/osmr/research-papers/2021/ec210020.htm . (Accessed: 2nd November Thilakasiri I et al (2025) The influence of social interactions in mitigating psychological distress during the COVID – 19 pandemic: A study in Sri Lanka. Front Psychol 15 Tuominen LS, Helle S, Helanterä H et al (2022) Structural equation modeling reveals decoupling of ecological and self-perceived outcomes in a garden box social-ecological system. Sci Rep 12:6425. https://doi.org/10.1038/s41598-022-10178-z United Nations Development Programme (UNDP) (2023) & Oxford Poverty and Human Development Initiative (OPHI). Understanding Multidimensional Vulnerabilities: Impact on People of Sri Lanka. UNDP Sri Lanka, Colombo van Deursen AJ (2020) Digital inequality during a pandemic: Quantitative study of differences in COVID-19–related internet uses and outcomes among the general population. J Med Internet Res 22 van de Weijer MP et al (2022) Genetic and environmental influences on quality of life: The covid-19 pandemic as a natural experiment. Genes Brain Behav 21 Wang J et al (2022) Global evidence of expressed sentiment alterations during the COVID-19 pandemic. Nat Hum Behav 6:349–358 White H (1980) A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48:817 World Bank (2023) Poverty & Equity Brief: Sri Lanka — South Asia (April 2023). World Bank Group, Washington DC World Bank (2020) South Asia Economic Focus, Fall 2020: Beaten or Broken? Informality and COVID-19. World Bank, Washington, DC. https://hdl.handle.net/10986/34517 World Bank. The COVID–19 Impact on Livelihoods and Poverty in Sri Lanka: A Background Note, World Bank Group (2020) Available at: https://documents1.worldbank.org/curated/en/724391618996958268/pdf/Background-Note-to-Sri-Lanka.pdf World Economic Outlook A (2021) : Managing divergent recoveries. IMF (2021).Available at: https://www.imf.org/en/Publications/WEO/Issues/2021/03/23/world-economic-outlook-april-2021 . (Accessed: 1st November 2025) Xi H et al (2023) Quantifying the impact of covid-19 on travel behavior in different socio-economic segments. Transp Policy 136:98–112 Xu F et al (2025) Using human mobility data to quantify experienced urban inequalities. Nat Hum Behav 9:654–664 Yabe T, García Bulle Bueno B, Frank MR, Pentland A, Moro E (2024) Behaviour-based dependency networks between places Shape Urban Economic Resilience. Nat Hum Behav 9:496–506 Yuan K-H, Bentler PM (2000) 5. Three likelihood-based methods for mean and covariance structure analysis with Nonnormal Missing Data. Sociol Methodol 30:165–200 Table 1 Table 1 is available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Table1.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 13 Jan, 2026 Editor assigned by journal 13 Jan, 2026 Submission checks completed at journal 03 Jan, 2026 First submitted to journal 30 Dec, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8482833","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":568699638,"identity":"baefaa1b-318e-4d0f-95a5-5bc9c4c66c2c","order_by":0,"name":"Madhumini Kulathunge","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/0lEQVRIiWNgGAWjYHACMyCWYGBjYGN8ABFIAJMyEkRoYTZA1sJDQAsIsLFJEKVFt/3wtse8ORZ5fAxsaRUfd9QymLPnHmD4UcPAI9mAw4ozaeXGvNskioEOO3Zz5pnjDJY97xIYe44x8EjjctWBHDNpoJbENgb2ttu8bccYDG7kGDDwNjDwyOHScv4NQkvxX6gWxr/4tNyA28J2jJmxrQashRlkC06H3XhWJjkXpIWZLVmyt+0Aj2XPG4PDMsckcHv/fPI2ibfb6hLnt7cZfvjZVidnzp5j+PBNjY2cxAEc1sABM5g8zAOKzwOgyCUW1DEYEK12FIyCUTAKRgoAAEuzUCx+ADTXAAAAAElFTkSuQmCC","orcid":"","institution":"University of Peradeniya","correspondingAuthor":true,"prefix":"","firstName":"Madhumini","middleName":"","lastName":"Kulathunge","suffix":""},{"id":568699640,"identity":"52a911f1-8847-4b9b-911b-aa094f81e6db","order_by":1,"name":"Lelumi Edirisinghe","email":"","orcid":"","institution":"University of Ruhuna","correspondingAuthor":false,"prefix":"","firstName":"Lelumi","middleName":"","lastName":"Edirisinghe","suffix":""},{"id":568699642,"identity":"41684b77-742e-441a-b59a-6ac556b4c944","order_by":2,"name":"Isuri Mapa","email":"","orcid":"","institution":"University of Peradeniya","correspondingAuthor":false,"prefix":"","firstName":"Isuri","middleName":"","lastName":"Mapa","suffix":""},{"id":568699645,"identity":"b55580b6-dd6d-431c-980f-d99fb6734a0b","order_by":3,"name":"Roshan Godaliyadda","email":"","orcid":"","institution":"University of Peradeniya","correspondingAuthor":false,"prefix":"","firstName":"Roshan","middleName":"","lastName":"Godaliyadda","suffix":""},{"id":568699648,"identity":"939b46a8-3506-4b4c-b4e1-57733f7cd6c1","order_by":4,"name":"Parakrama Ekanayake","email":"","orcid":"","institution":"University of Peradeniya","correspondingAuthor":false,"prefix":"","firstName":"Parakrama","middleName":"","lastName":"Ekanayake","suffix":""},{"id":568699651,"identity":"6295ab11-f317-4656-8eb4-174d9e633b84","order_by":5,"name":"Niranjan Karnik","email":"","orcid":"","institution":"University of Illinois at Chicago","correspondingAuthor":false,"prefix":"","firstName":"Niranjan","middleName":"","lastName":"Karnik","suffix":""},{"id":568699654,"identity":"8d3ed5b2-2a91-4177-873c-1a76e0cc7c0f","order_by":6,"name":"Samath Dharmaratne","email":"","orcid":"","institution":"University of Peradeniya","correspondingAuthor":false,"prefix":"","firstName":"Samath","middleName":"","lastName":"Dharmaratne","suffix":""},{"id":568699656,"identity":"79446789-2e56-4d43-99a2-788810461958","order_by":7,"name":"Vijitha Herath","email":"","orcid":"","institution":"University of Peradeniya","correspondingAuthor":false,"prefix":"","firstName":"Vijitha","middleName":"","lastName":"Herath","suffix":""},{"id":568699658,"identity":"517ddadd-bfab-44f6-a2a1-3d2c6886464a","order_by":8,"name":"Janaka Ekanayake","email":"","orcid":"","institution":"University of Peradeniya","correspondingAuthor":false,"prefix":"","firstName":"Janaka","middleName":"","lastName":"Ekanayake","suffix":""}],"badges":[],"createdAt":"2025-12-30 15:53:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8482833/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8482833/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":99792567,"identity":"6cc310d4-8672-4fb5-b108-bf1e085d62f3","added_by":"auto","created_at":"2026-01-08 13:22:32","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6746663,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/b77691a111637f33576e3542.docx"},{"id":99598110,"identity":"9257c330-8897-49b7-9af8-8a0262abf659","added_by":"auto","created_at":"2026-01-06 10:03:36","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":754702,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/79d6243f9db75f981cebfc32.png"},{"id":99598108,"identity":"f23a74c0-7580-498a-8e32-81e3b1860f18","added_by":"auto","created_at":"2026-01-06 10:03:36","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":585557,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/9551a258ea87a456f45ca228.png"},{"id":99792875,"identity":"628fd6c1-994e-42ec-8410-316b4d854c0f","added_by":"auto","created_at":"2026-01-08 13:27:54","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":47656,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/be000c789b9d2cab0d4291ed.png"},{"id":99598112,"identity":"79165c62-6588-4664-b3cb-8b440e08260d","added_by":"auto","created_at":"2026-01-06 10:03:36","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":233446,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/3ecf596b29faeec18a726725.png"},{"id":99793538,"identity":"d711651c-2a1f-49a7-a711-89de985246b5","added_by":"auto","created_at":"2026-01-08 13:31:50","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":463661,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/8204b5f1ca4c48e76f2f6e1e.png"},{"id":99793895,"identity":"27eda879-c393-44df-bf72-b4688887254d","added_by":"auto","created_at":"2026-01-08 13:33:27","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":360960,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/28f24ab5db7ecab6e3581584.png"},{"id":99792998,"identity":"19e047cc-cc34-42fc-8727-95a2b85e0ef8","added_by":"auto","created_at":"2026-01-08 13:30:48","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":611141,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/705f42fc8d9698df6c3775d7.png"},{"id":99792238,"identity":"9f90f41c-7d73-400f-9d79-e9a50a964551","added_by":"auto","created_at":"2026-01-08 13:16:39","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":543228,"visible":true,"origin":"","legend":"","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/1d5218bd696765f8c4f6ab48.png"},{"id":99793397,"identity":"64ebf57a-74f1-4321-b5da-6e30e8859773","added_by":"auto","created_at":"2026-01-08 13:31:32","extension":"json","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9661,"visible":true,"origin":"","legend":"","description":"","filename":"29ae1733fa7a4452b492fee237a5befe.json","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/74285bc0d77c3062200d2bad.json"},{"id":99598127,"identity":"5c688519-ddc0-4201-b577-01cd7c62a877","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"xml","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":177455,"visible":true,"origin":"","legend":"","description":"","filename":"29ae1733fa7a4452b492fee237a5befe1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/bdb2923499849026d406692f.xml"},{"id":99793086,"identity":"e571356c-35d8-4873-92a8-c298fe918ed2","added_by":"auto","created_at":"2026-01-08 13:30:59","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":754702,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/288eda69ff20f46da2e1afb5.png"},{"id":99792401,"identity":"7f5b3a47-9c53-4bea-ba6c-dd20179745b1","added_by":"auto","created_at":"2026-01-08 13:19:18","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":585557,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/225ae72ccc393c8afe60186a.png"},{"id":99793735,"identity":"facbe9e5-78ab-4d86-b350-312aa1c3dcf9","added_by":"auto","created_at":"2026-01-08 13:32:16","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":47656,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/32278e4bf256696ccedd6819.png"},{"id":99598120,"identity":"8bbc7861-02fb-4c3e-b941-771f62ee28ab","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":233446,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/209fb052489d1ecf57f26352.png"},{"id":99793526,"identity":"35452647-a976-4c3f-97e1-add318c62609","added_by":"auto","created_at":"2026-01-08 13:31:46","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":463661,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/2146e61f8cd88a3729b1bc68.png"},{"id":99598129,"identity":"665bc9e6-7c7f-4d97-a363-f6c86d01a69e","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":360960,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/53705b7d65b62f0e4800e5db.png"},{"id":99793528,"identity":"42e2ca0b-7542-45a0-b393-106048182e63","added_by":"auto","created_at":"2026-01-08 13:31:46","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":611141,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/e52b34d16fca7f12bc347cdf.png"},{"id":99598128,"identity":"37212727-f734-4a52-a738-f39038a9911b","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":543228,"visible":true,"origin":"","legend":"","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/e5ecbbaf2da56e4cc8feefc7.png"},{"id":99793900,"identity":"b0e23f30-0bd7-4c0c-88d2-45cee50ffe1d","added_by":"auto","created_at":"2026-01-08 13:33:27","extension":"jpeg","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":797118,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/51f7ca444799b251e9adf125.jpeg"},{"id":99793729,"identity":"cff3f4cd-7864-4d2e-bae8-1d949d7b16eb","added_by":"auto","created_at":"2026-01-08 13:32:15","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":465168,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/faef2f93e8bea6c4f537fb71.png"},{"id":99598123,"identity":"127010cb-4f4f-4996-8e94-61a968204923","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":47656,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/3cdef201a62b9076a236d7e8.png"},{"id":99793149,"identity":"971c8a80-e20e-4b14-a748-4107b55497a1","added_by":"auto","created_at":"2026-01-08 13:31:06","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":233446,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/44cf3fa7905602490f45e775.png"},{"id":99793313,"identity":"03b203b4-c12c-4d16-939a-728dfdb60c52","added_by":"auto","created_at":"2026-01-08 13:31:24","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":439463,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/17693e797a5b79175ce33dc6.png"},{"id":99793439,"identity":"759a77bf-0a3c-429e-a08f-496df4892671","added_by":"auto","created_at":"2026-01-08 13:31:35","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":353000,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/fb0970be5c42a1e173cae08b.png"},{"id":99793088,"identity":"cb37a5de-badf-44ac-a75c-1e4531109067","added_by":"auto","created_at":"2026-01-08 13:30:59","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":604186,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/fd2dafdf02ca7e1f1b054d0a.png"},{"id":99792521,"identity":"523590dd-732a-4486-a5c3-59d76d127421","added_by":"auto","created_at":"2026-01-08 13:21:32","extension":"png","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":435976,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/8edfa7a56732a8d196bb5cf3.png"},{"id":99793196,"identity":"b630a5b2-a352-47fb-b161-122de009e9eb","added_by":"auto","created_at":"2026-01-08 13:31:09","extension":"png","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":157287,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/a8cbbb4c0aef2ac4e4e529a8.png"},{"id":99598149,"identity":"807f37e3-a08b-46b0-8124-e21546b9c9f5","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":150892,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/9c174671e0c2ee193a9eb431.png"},{"id":99598130,"identity":"abb6798d-0e81-4cc7-b154-8df710f2d70d","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13668,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/0f7b0e877b25643ba255a180.png"},{"id":99598136,"identity":"aaa51b0c-d9ec-4aa3-917e-adc0b3256e60","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35727,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/c84d828fcc873cea3933fd36.png"},{"id":99598132,"identity":"20966588-3098-4c12-9c06-dda0f671cb55","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":144683,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/7ee412608e45e55c838fe31f.png"},{"id":99598144,"identity":"39ca325f-9da8-4049-b8b8-af4d3edfcc7c","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":101267,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/c6584156a7ee8e1f572f59e4.png"},{"id":99598145,"identity":"a723a20d-6aea-4084-a1c2-64d5eb6d578b","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":108376,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure7.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/5e720aac05d348c48e429246.png"},{"id":99792893,"identity":"7c281e3b-c249-4c98-b994-714fa43f99b2","added_by":"auto","created_at":"2026-01-08 13:28:16","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":177323,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure8.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/df0aaba0530ab56c316be2f5.png"},{"id":99792887,"identity":"cc5bcf69-3232-4466-9480-c40de1939752","added_by":"auto","created_at":"2026-01-08 13:28:11","extension":"png","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":229696,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/31c1e7b036d585c994356f93.png"},{"id":99792262,"identity":"8a53a455-cf33-4ec6-ae91-df9fed7b9273","added_by":"auto","created_at":"2026-01-08 13:17:16","extension":"png","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":149932,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/592a49b94422400f2b588acb.png"},{"id":99598141,"identity":"e60bb53d-6ad9-48a3-91a8-231fbcb077cc","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":37,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13668,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/64fb9ffa0dfa9c98fb57344e.png"},{"id":99598134,"identity":"267fc97d-1da2-4041-8ee7-a1ee56581c53","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":38,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35727,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/ce001dc9b5a0420afe4db761.png"},{"id":99792645,"identity":"d50d0996-7e22-4522-b7cb-ae6aee0bfea7","added_by":"auto","created_at":"2026-01-08 13:23:24","extension":"png","order_by":39,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":134832,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/25b33b7844afa094940bb3ea.png"},{"id":99598151,"identity":"06fa40f4-b681-4b75-8c63-cb02bd73703f","added_by":"auto","created_at":"2026-01-06 10:03:38","extension":"png","order_by":40,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":100175,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/2934d4fea5436a56bc0aafc2.png"},{"id":99598155,"identity":"b831db41-2960-4a16-8291-a98bae2f646e","added_by":"auto","created_at":"2026-01-06 10:03:38","extension":"png","order_by":41,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":107201,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/22a9cfb57cdc4d93d7a086cb.png"},{"id":99793410,"identity":"268ff84f-1c9c-40c2-812c-8944c902ce3e","added_by":"auto","created_at":"2026-01-08 13:31:32","extension":"png","order_by":42,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":176339,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/16f788a28c1c27e0ff613062.png"},{"id":99793460,"identity":"208c2798-f5f4-4967-8c08-0e4192e152ce","added_by":"auto","created_at":"2026-01-08 13:31:38","extension":"xml","order_by":43,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":171515,"visible":true,"origin":"","legend":"","description":"","filename":"29ae1733fa7a4452b492fee237a5befe1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/c78b5cec640a03af41cdad86.xml"},{"id":99794273,"identity":"dc051662-5714-4f37-90e1-d83266028ff0","added_by":"auto","created_at":"2026-01-08 13:34:25","extension":"html","order_by":44,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":190817,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/99f57589aa9937ff3bcfec31.html"},{"id":99598102,"identity":"5385ee12-ff03-4277-87f4-355964c851b2","added_by":"auto","created_at":"2026-01-06 10:03:36","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":754702,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSampling Framework for the Spatial and Demographic Overview of Sri Lanka vs. Survey Participants. a, \u003c/strong\u003eThe national ethnic distribution from the latest census in Sri Lanka before the pandemic(Sri Lanka Socio Economic Data,\u003ca href=\"#t94xd7lri11j\"\u003e2021\u003c/a\u003e ) is expressed alongside the ethnic distribution of survey participants. \u003cstrong\u003eb\u003c/strong\u003e, The population distribution by district based on 2020 mid-year estimates (Registrar General's Department, \u003ca href=\"#yd6jhc3ivow8\"\u003e2024\u003c/a\u003e) uses a gradient color scale from dark purple for the smallest population shares to bright yellow for the highest shares. \u003cstrong\u003ec, \u003c/strong\u003eThe district-wise distribution of PCR-confirmed COVID-19 cases in Sri Lanka\u003csup\u003e \u003c/sup\u003e(Epidemiology unit, \u003ca href=\"#rp1untacun30\"\u003e2019\u003c/a\u003e) uses the same gradient color scale. \u003cstrong\u003ed\u003c/strong\u003e, District-wise distribution of survey participants, visualized using a colour scale consistent with population shares and PCR-positive case percentages, shows that survey participants were largely concentrated in the same high-population and high-COVID-burden districts such as Colombo and Gampaha, while also showcasing a geographical diversity in the survey data.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/4da90356e2170fac18af06b5.png"},{"id":99598103,"identity":"5e784b42-7f4b-41d7-95e4-bc444db6d731","added_by":"auto","created_at":"2026-01-06 10:03:36","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":585557,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe overall framework of the study. \u003c/strong\u003eA complete data-driven approach integrating nationally representative behavioural survey data with factor analysis and unsupervised clustering revealed four empirically grounded behavioural archetypes: Privileged Adaptors, Cautious Traditionalists, Stable Minimalists, and\u003cem\u003e \u003c/em\u003eResilient Rebounders. These archetypes capture how socioeconomic position, digital access, and occupational security shape differential adaptive pathways. The conceptual synthesis illustrates how structural inequality operates through three behavioural mediators, including substitution capacity, exposure necessity,\u003cem\u003e \u003c/em\u003eand support infrastructure, that generate distinct adaptation modes ranging from behavioural plasticity to forced rebound. The framework culminates in a model for building behavioural capacity for equitable recovery, emphasizing the enhancement of behavioural substitutability, agentic flexibility, social embeddedness, and mitigation of structural exposure constraints. Together, these findings advance understanding of how inequality manifests as behavioural divergence, aligning with the social and structural determinants of human adaptability under global disruption.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/2d9ecc5ff11996495cb76cc6.png"},{"id":99598105,"identity":"df7cf356-c9ca-4ace-8fa7-3428ff98fe92","added_by":"auto","created_at":"2026-01-06 10:03:36","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":47656,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eScree plot. \u003c/strong\u003eThe first 13 components in the scree plot have eigenvalues greater than 1, justifying the selection of 13 principal factors based on the Kaiser Criterion (Kaiser, H.F. et al., 1960).\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/6da9ac5aba8af223a0fdecdc.png"},{"id":99792474,"identity":"93cdfc25-2e77-43ab-8735-c6f894d80076","added_by":"auto","created_at":"2026-01-08 13:20:50","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":233446,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEigen gap curves. \u003c/strong\u003eThese curves illustrate the difference between consecutive eigenvalues of the Laplacian L, which is used to identify the number of clusters (k) in spectral clustering. The choice of k is determined by the curve, which maximizes the variation and demonstrates dominance over an extensive range of log(σ). Hence, the 4th eigengap suggests the most persistent and distinct clustering structure in the data\u003cstrong\u003e, \u003c/strong\u003eindicating 4 inherent clusters in the population, since the other dominant eigengap 1 corresponds to a single cluster, being a trivial solution.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/5ebc65e5c1e40a83572a5e9f.png"},{"id":99793262,"identity":"9bf68f6c-705d-4044-8352-6f569152c85f","added_by":"auto","created_at":"2026-01-08 13:31:17","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":463661,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCluster profiles highlighting demographic, socioeconomic, and mobility-behavioral uniqueness across four groups. \u003c/strong\u003eThese data-driven cluster profiles identified through Unsupervised Machine Learning have a unique behavioral adaptation in mobility during the pandemic, followed by distinct demographic specifications such as ethnicity, education, and employment, highlighting the divergent pandemic experiences of Sri Lanka's underlying socioeconomic stratification.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/7b6c8580d0f151f3c2388fb5.png"},{"id":99792785,"identity":"fe49b589-c2b5-4ac9-b151-6310b3ee85f2","added_by":"auto","created_at":"2026-01-08 13:26:01","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":360960,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSocio-demographic profiles of four clusters\u003c/strong\u003e. This is the summary of the socio-demographic characteristics of the four distinct clusters derived from survey data in Sri Lanka. Collectively, these demographic distinctions justify the intersecting nature of education, occupation, ethnicity, and geographical locations in deriving mobility and behavioral vulnerabilities during the pandemic. \u003cstrong\u003ea,\u003c/strong\u003e Ethnic composition varies significantly across clusters: Cluster 0 is predominantly Sinhalese (81.2%), while Cluster 3 is the most diverse, including high proportions of Tamil (30.2%)\u003cstrong\u003e \u003c/strong\u003eand Muslim (26.1%) individuals. \u003cstrong\u003eb\u003c/strong\u003e, The evolution of household income distributions of the archetypes across COVID-19 phases is visualized using stacked bar plots of the percentages of individuals across five nationally benchmarked (Household income and expenditure survey , \u003ca href=\"#4bo3ipg2b966\"\u003e2019\u003c/a\u003e)\u003csup\u003e \u003c/sup\u003e\u0026nbsp;income categories (very low, lower-middle, middle, upper-middle, and high income) before the pandemic and during three COVID-19 waves. Each panel is assigned to a data-driven cluster derived from data-driven analysis of mobility and behavioural profiles. The visualization justifies both common income contractions across clusters and, as well as the disparities. Cluster 0 has shown marked resilience, clusters 1 and 2 have experienced partial rebounds, and cluster 3 has experienced persistent structural downgrades in household income. \u003cstrong\u003ec, \u003c/strong\u003eTertiary education is predominant in cluster 0 (29.9%), while clusters 2 and 3 are characterized by the lowest levels (3% and 4.8%, respectively). \u003cstrong\u003ed, \u003c/strong\u003eWhen the informal sectors of employment are considered, estate and daily wage workers are prominent in the vulnerable cluster 3 compared to other clusters. \u003cstrong\u003ee, \u003c/strong\u003eThe least privileged cluster 3\u003cstrong\u003e \u003c/strong\u003ehas the highest proportion (17.2%) of individuals residing in districts with high multidimensional vulnerability, as defined in the 2023 UNDP report(UNDP Sri Lanka, Colombo, \u003ca href=\"#v24p8ll6d7m9\"\u003e2023\u003c/a\u003e), compared to 3.4% in the most privileged cluster 0.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/14c3b0e0da211bcbde7e7505.png"},{"id":99598117,"identity":"56fb1689-03b4-4c7d-8ca9-e731bc613207","added_by":"auto","created_at":"2026-01-06 10:03:37","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":611141,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEvolution of the trends of the mobility and behavioral factors disaggregated by cluster, along the different stages of the pandemic. \u003c/strong\u003eThis enabled a more comprehensive view of this disturbance, rather than a mere pre-pandemic and post-pandemic analysis, by tracking its evolution more seamlessly. These plots reveal how distinct patterns of mobility and behavioural adaptation emerged under successive public-health restrictions, providing a dynamic visualization of the shifting equilibrium states within Sri Lankan society. \u003cstrong\u003ea\u003c/strong\u003e, Social and recreational mobility dropped sharply across all clusters during the first wave. Clusters 1 and 3 recovered through risk-managed and necessity-driven reconnection, while clusters 0 and 2 remained consistently low due to caution and constraint. \u003cstrong\u003eb\u003c/strong\u003e, Private commuting and work exposure were stable for clusters 1 and 2 throughout, while clusters 0 and 3 exhibited V-shaped recovery. \u003cstrong\u003ec\u003c/strong\u003e, Home-based work isolation increased significantly in cluster 0 with the increase in their work-from-home frequencies with the pandemic, while other clusters remained widely unchanged. \u003cstrong\u003ed\u003c/strong\u003e, Financial access points correspond to the income status of clusters, as the highest income cluster 0 possesses the highest access frequencies, gradually decreasing to the lowest access frequency of cluster 3. \u003cstrong\u003ee\u003c/strong\u003e, Public transport reliance dropped sharply during the first wave for all clusters and later underwent a partial recovery. \u003cstrong\u003ef\u003c/strong\u003e, Access to Essential Services through government and NGOs dropped sharply during the first wave for the vulnerable cluster 3, yet rebounded sharply in later waves, resulting in a V-shaped trend \u003cstrong\u003eg\u003c/strong\u003e, Bicycle Mobility was significantly higher in cluster 2, reflecting consistent bicycle usage for activities such as work mobility \u003cstrong\u003eh\u003c/strong\u003e, Variable level of work exposure declined for all clusters in the first wave, yet cluster 0 experienced the sharpest decline, consistent with their high remote work adoption. \u003cstrong\u003ei\u003c/strong\u003e, Hired transport dependence, followed a similar pattern as private commuting and work exposure, experiencing a V-shaped pattern for clusters 0 and 3, and clusters 1 and 2 stayed consistently low. \u003cstrong\u003ej\u003c/strong\u003e, Community resource sourcing through mobile shops and home delivery spiked during the first wave. The most significant spikes were observed in the most privileged cluster pair, clusters 0 and 1. \u003cstrong\u003ek\u003c/strong\u003e, Shared transport usage fell sharply for both clusters 0 and 3 during the first wave; by the third, only cluster 3 recovered, reflecting dependence on communal transport, while cluster 0 maintained low usage due to remote work and private mobility. \u003cstrong\u003el\u003c/strong\u003e, Proximal living and active commuting were consistently high in cluster 2, and cluster 3 experienced a V-shaped trend. \u003cstrong\u003em\u003c/strong\u003e, Digital platform engagement spiked for privileged cluster 0, while cluster 1 also maintained high digital engagement compared to clusters 2 and 3.\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/2764e2824d4407e7b879d5c4.png"},{"id":99598114,"identity":"87aec47b-0e4d-410e-a922-792f3b78e980","added_by":"auto","created_at":"2026-01-06 10:03:36","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":543228,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConceptual Framework for behavioral adaptation based on the identified archetypes. \u003c/strong\u003ePrivileged Adaptors exhibit high-substitutability and low-exposure necessity, enabling them to shift into low-risk alternatives with minimal disruption. Cautious Traditionalists balance community buffering with moderate exposure, while cautiously maintaining familiar routines. Stable Minimalists occupy low substitutability yet low-variance routine space. Resilient Rebounders experience low substitutability and high exposure, producing income-behavior decoupling despite mobility rebounds, suggesting that behavioral adaptation can occur even in the absence of economic stabilization.\u003c/p\u003e","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/77832c6935d83c429b315120.png"},{"id":99804329,"identity":"42d0e492-6aee-4b9b-a7aa-3ef865d5c094","added_by":"auto","created_at":"2026-01-08 14:13:10","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4356212,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/03016bc4-b72d-4694-852d-78d6df04d389.pdf"},{"id":99598101,"identity":"81b8fa79-6fc1-440e-94ca-db98da54d1b8","added_by":"auto","created_at":"2026-01-06 10:03:36","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":20504,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8482833/v1/0421e80df4527ebc2e03082b.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Structural Divides Shape the Nonlinear Nature of Human Mobility in COVID-19","fulltext":[{"header":"Introduction","content":"\u003cp\u003eMobility is a core attribute of human behavior that extends beyond physical movements to represent the integrated behavioral outcome of cognitive, emotional, and social processes that define human functioning, reflecting bounded rationality under real-world constraints (Simon, H., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1955\u003c/span\u003e). Social connectivity, economic exchange, and cultural diffusion shape mobility patterns, ultimately determining the very structure of societies.\u003c/p\u003e \u003cp\u003eSimilar to an impulse function in a dynamic system that reveals its intrinsic characteristics, unexpected disruptions such as pandemics may cause changes and reset the equilibrium states of collective behaviour in the societal structural dynamics, exposing latent properties. These distortions to the usual mobility patterns can reveal useful insights into behavioral adaptation, resilience, and vulnerability, reflecting the broader socioecological systems (Bronfenbrenner, U., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e1977\u003c/span\u003e; Bronfenbrenner, U. \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1979\u003c/span\u003e) to understand how people respond in the perception of risk, stress, need, and opportunity.\u003c/p\u003e \u003cp\u003eThe coronavirus disease 2019 (COVID-19) pandemic was such a strong driving factor that revealed this underlying behaviour of human societies all over the globe, by offering a unique opportunity to gauge how different civilisations around the world respond, adapt, and falter under extreme pressure in a sudden disturbance (Pangallo, M. et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Petherick, A. et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Santana, C. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2023\u003c/span\u003e). The effects of the outbreak were widespread, including health, economy, psychology, education, and many more. There were over 772\u0026nbsp;million confirmed cases and approximately 7\u0026nbsp;million deaths in 2023, with accurate figures most likely to be significantly higher due to underreporting and limited testing (WHO COVID-19 Dashboard, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBeyond the immediate health crisis, the pandemic caused the worst global economic downfall since World War II. The International Monetary Fund (IMF) recorded around 3% drop in the world's GDP in 2020 (IMF, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e), and the International Labor Organization (ILO) reported that working hours equivalent to about 255\u0026nbsp;million full-time jobs were lost, destroying livelihoods and worsening poverty (ILO, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn response, governments implemented strict containment measures, including lockdowns, curfews, and social distancing protocols (Onyeaka, H. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; Andronico, A. et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These critical public health constraints redefined human mobility, transforming it from a routine aspect of life into a regulated and often restricted privilege (Jay, J. et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Gozzi, N. et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). This had a cumulative effect on access to essential services, employment, and social interactions, with the burden of these disruptions falling disproportionately along socioeconomic disparities (Xu, F. et al. ,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2025\u003c/span\u003e; Xi, H. et al., \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), exposing the nonlinear dynamics of the civilizations. Existing inequalities in wealth, education, and digital access created great diversity in adaptation and resilience across the society (Jay, J. et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Wang, J. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2022\u003c/span\u003e; van Deursen, A. J., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDue to the South Asian region's dense populations and large informal economic sectors, more than 80% of jobs fall outside formal protections (World Bank Open Data, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2025\u003c/span\u003e; Ahmed, S. A. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2023\u003c/span\u003e). This structural characteristic associated with lockdown measures led to a direct threat to the workforce that relied on daily mobility and physical presence, thereby intensifying the trade-off between public health and economic survival (World Bank, Washington, DC, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn Sri Lanka, the Ministry of Health recorded an official death toll of over 15,000 by February 2022, posing a significant public health burden for a population of approximately 22\u0026nbsp;million (Galgamuwa, L. S. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2023\u003c/span\u003e). This crisis, compounded by stringent and prolonged control measures, including strictly enforced, district-wide lockdowns and curfews, brought social and economic life to a near standstill (World Bank Group, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e; Agampodi, S., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). With its distinctive sociodemographic structure marked by high educational attainment yet limited economic resilience, a diverse ethnic composition, and variable quality of health services, Sri Lanka offers a compelling lens through which to examine how such severe constraints reshaped daily lifestyles, behavioural patterns, and adaptive capacities (World Bank Group, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe pandemic\u0026rsquo;s economic repercussions in Sri Lanka were equally profound, with the output sharply contracting due to prolonged lockdowns (Central Bank of Sri Lanka, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e). While Sri Lanka\u0026rsquo;s economic fragility stemmed from longstanding structural imbalances, the pandemic likely accelerated the trajectory toward collapse by amplifying fiscal pressures and disrupting key income sectors. GDP in Sri Lanka fell by 3.6% in 2020, with sectors such as tourism witnessing income losses exceeding 80% relative to 2019 (Central Bank of Sri Lanka, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e; Central Bank of Sri Lanka, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2022\u003c/span\u003e; World Bank Open Data, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2025\u003c/span\u003e). This downturn pushed an estimated half a million people into poverty, while unemployment rose from 4.7% in the first quarter of 2019 to 5.7% a year later, severely constraining households\u0026rsquo; ability to sustain their mobility and livelihoods under ongoing restrictions (World Bank Group, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe significance of this study is its unique, high-resolution dataset collected through a door-to-door field survey of 3020 households conducted across Sri Lanka during a period of eased restrictions, i.e., from November to December 2021(Ilangarathna, G. A. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2024\u003c/span\u003e). It captured various social effects of the pandemic, including its impact on health, economy, education, psychology, culture, and mobility. Building on earlier research based on this survey (Senarath, N. et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Thilakasiri, I. et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2025\u003c/span\u003e ), this study analyzes human behavior on household mobility during the time of the crisis. This data collection methodology was critical in capturing comprehensive behavioral and socioeconomic data from groups that are often underrepresented in online surveys, such as those from rural areas and lower socioeconomic classes with limited internet access. A longitudinal perspective that is uncommon in the literature was provided by the survey, which specifically recorded retrospective, self-reported changes in mobility patterns across the three separate waves of the pandemic, together with the pre-pandemic period, allowing for a nuanced and more person-centric analysis to uncover the diverse experiences of different societal groups.\u003c/p\u003e \u003cp\u003eWhile numerous studies have investigated the more general changes in human mobility during the pandemic using digital trace data(Andronico, A. et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Jay, J. et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Pan, Y. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e), a scarcity remains in nuanced, data-driven analysis that connects these behavioural changes to the socioeconomic backgrounds of individuals. Most existing literature generally focuses only on what changed and lacks the specificity to reveal why and how different socioeconomic groups exhibited vastly different capabilities in terms of adapting to the unexpected situation. Studies also suggest that risk perception is conditioned by livelihood pressures(Nisa, C. F. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e) and demonstrate that collective resources such as community support can assist in adaptation when formal resources are limited(Durante, R. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; Ferwana, I. et al,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; Fraser, T. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis study addresses this gap with the use of Factor Analysis together with unsupervised machine learning. We identified distinct behavioural clusters within the Sri Lankan population, revealing how pre-existing socioeconomic disparities dictated the behavioural capacity to adapt to a prolonged crisis. This leads to the revelation of latent properties and uncovers a deeper understanding of the nonlinear characteristics of human society in a global epidemic, highlighting its potential to aid in formulating more informed public health strategies in the future.\u003c/p\u003e \u003cp\u003eOur multi-step data-driven approach first quantified the behavioural changes from the pre-pandemic to the pandemic period using factor analysis to identify 13 principal factors related to human mobility. Certain survey questions aligned with them, hence appropriate descriptors were selected for each factor in the subsequent analysis. The dataset was then denoised using Principal Component Analysis (PCA), leaving only the most significant information related to the 13 factors. Then, unsupervised machine learning determined the optimal number of clusters best representing the underlying behavioural distribution of the survey and then revealed 4 distinct archetypes, materializing a complete data-driven approach.\u003c/p\u003e \u003cp\u003eIn the analysis, we observed non-linear asymmetric outcomes, translating into a K-shaped behavioral and economic recovery, where some groups recover or improve while others fall further behind(Bonaccorsi, G. et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; U.S. Bureau of Labor Statistics, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2025\u003c/span\u003e; Egger, D. et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Decerf, B. et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Hence, we propose a conceptual framework of behavioral adaptation under three structural determinants; Substitution capacity (IMF, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; OECD, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e), Exposure necessity(Nisa, C. F. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; Michaels, D. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2024\u003c/span\u003e), and Support infrastructure (Durante, R. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; Fraser, T. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2022\u003c/span\u003e), to determine the adaptive capacity and explain the aforementioned responsive patterns. Thus, our study demonstrates that effective crisis response should build behavioral capacity, enabling more equitable adaptation and mitigating the disparities caused and intensified through inequality.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eData Source and Study Design\u003c/h2\u003e \u003cp\u003eUsing the COVID-19 pandemic as a quasi-natural experiment, this study evaluates the mobility, behavioral adaptation, and underlying socioeconomic evolution with fast-changing external constraints through a cross-sectional nationwide household survey conducted covering all provinces in Sri Lanka, from November 6, 2021, to December 10, 2021(Ilangarathna, G. A. et al, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2024\u003c/span\u003e). To ensure access to both rural and urban settings, this was structured as a face-to-face Computer-Assisted Personal Interview (CAPI) survey, supported by national and local administrative authorities. The multi-domain questionnaire had 78 core items and 8 demographic questions to capture both socioeconomic and behavioral impacts of COVID-19 through three distinctly identified pandemic waves in Sri Lanka. The first wave was from 27th March 2020 to 3rd October 2020, the second wave was from 4th October 2020 to 14th April 2021, and the third wave was from 15th April 2021 to 31st December 2022 (Epidemiology unit, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2019\u003c/span\u003e). The main focuses included mobility and behavioral adaptation, income shocks, household economy, psychological changes, food consumption, digital access, and educational disruption, where this particular research focuses on mobility and behavioral adaptation.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eParticipants and Sampling\u003c/h3\u003e\n\u003cp\u003eThe geographic and demographic representativeness was ensured with multistage cluster sampling, with stratification by population density, COVID-19 risk severity, and urban, rural, and estate-sector communities. The minimum sample size of 2,401 was obtained using Cochran\u0026rsquo;s formula with a 95% confidence level, 2% margin of error, and \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.5. The survey included data on 3,020 households across 9 provinces and 20 districts (Ilangarathna, G. A. et al, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2024\u003c/span\u003e). The village officers facilitated purposive inclusion of pandemic-affected households.\u003c/p\u003e\n\u003ch3\u003eData Preparation\u003c/h3\u003e\n\u003cp\u003eThe final dataset used for the analysis contained 1,898 households after performing listwise deletion to remove incomplete or partially filled observations. The retained dataset had a consistent demographic structure with the national distribution as depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eAlthough this is a cross-sectional survey, the questionnaire contained retrospective information across multiple pandemic phases, including before the pandemic, the first, second, and third COVID-19 waves. The Likert-scale questionnaire items were recorded on a symmetric directional scale (\u0026minus;\u0026thinsp;1.5, \u0026minus;\u0026thinsp;1, 1, 1.5) to reflect the magnitude and direction of behavioral change. The mean of the scores from the first, second, and third waves was calculated, and this was subtracted from the corresponding pre-pandemic value to quantify the pandemic-induced behavioral change. This is an established approach for assessing deviations from a baseline under exogenous shocks (Cohen, J et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2015\u003c/span\u003e ; Menard, S. W., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2002\u003c/span\u003e ).\u003c/p\u003e\n\u003ch3\u003eVariables\u003c/h3\u003e\n\u003cp\u003eThe questionnaire contained 40 behavior and mobility-related items spanning remote work, digital engagement, commuting patterns, non-work mobility, and service access, which were theorized to represent latent domains such as household adaptation, digital substitution, and rebound in social interactions.\u003c/p\u003e \u003cp\u003eComplementing those were the variables of education level, employment sector, household income, settlement (urban/rural/estate), and ethnicity. These socioeconomic variables aided in understanding the structural constraints that shaped behavioral distinctions.\u003c/p\u003e\n\u003ch3\u003eFactor Analysis\u003c/h3\u003e\n\u003cp\u003eFactor analysis reduces a large set of observed variables to a smaller set of unobserved latent factors. This process captures the essence of the data's structure. The first step involves PCA, which identifies components that represent the shared variance among the variables. Next, varimax rotation is applied. This orthogonal rotation method maximizes the variance of loadings for factors, thereby improving the interpretation of the results (Kaiser, H. F., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1958\u003c/span\u003e). This step creates uncorrelated factors with a clearer structure. It strengthens the connection between each factor and specific dimensions of the data, allowing for more meaningful interpretations. By simplifying the data, factor analysis boosts its reliability and reinforces its conclusions. Hence, it is a highly effective method for identifying the primary constructs underlying observed behaviors.\u003c/p\u003e \u003cp\u003eIn the procedure, only the items with factor loadings above 0.4 were retained because they demonstrate a strong connection to the underlying factor (Pett, M. A. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2012\u003c/span\u003e), thereby improving the reliability of the factor solution. Additionally, items with cross-loadings where the difference was less than 0.1 across factors were removed (Acar G\u0026uuml;vendi̇r, M. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2022\u003c/span\u003e ; Jamali, J. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eUnsupervised Machine Learning: Spectral Clustering and K-Means Clustering\u003c/h2\u003e \u003cp\u003eSpectral clustering is a powerful unsupervised machine learning technique based on graph theory and linear algebra, which can be used for identifying clusters in data that may not be linearly separable in the high-dimensional feature space (Ng, Andrew et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2002\u003c/span\u003e). Compared to the traditional clustering algorithms, this tool is particularly effective in discovering non-convex and arbitrarily shaped clusters embedded in high-dimensional spaces. First, a similarity graph is obtained, where each node corresponds to a data point and edges are weighted according to pairwise similarities of the data with the use of a Gaussian (RBF) kernel function, as shown in Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:{A}_{ij}=\\:{e}^{-\\frac{{\\Vert\\:{x}_{i}-\\:{x}_{j}\\Vert\\:}^{2}}{2{\\sigma\\:}^{2}}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eHere, Aij symbolizes the commonality between the two data points x\u003csub\u003ei​\u003c/sub\u003e and x\u003csub\u003ej\u003c/sub\u003e, and sigma (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e) is a tunable scaling parameter that controls the zooming effect in the standard spectral clustering algorithm. The value of sigma critically determines the structure of the affinity graph, where the small values capture fine-grained clusters, whereas large values connect distant points, highlighting broader global structures.\u003c/p\u003e \u003cp\u003eOnce the affinity matrix A is constructed, the degree matrix D is computed as a diagonal matrix containing the sum of affinities for each data point in each row, as shown in Eq.\u0026nbsp;(\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:{D}_{ii}=\\:\\sum\\:_{j}{A}_{ij}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eUsing A and D, the graph Laplacian is formed as in Eq.\u0026nbsp;(\u003cspan refid=\"Equ3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:L=\\:{D}^{-\\frac{1}{2}}\\:\\left(D-A\\right){D}^{-\\frac{1}{2}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe eigenvalues of the L (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{i}\\)\u003c/span\u003e\u003c/span\u003e) encode the connectivity structure of the graph. Small eigenvalues correspond to well-connected components, while large eigenvalues indicate separations between clusters. Therefore, the difference between successive eigenvalues or the eigengap (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{i}-\\:{\\lambda\\:}_{j}\\)\u003c/span\u003e\u003c/span\u003e) provides a quantitative measure of cluster separability, thereby revealing the structure of the data distribution through clustering. A significant increase in the eigengap typically signifies a natural cluster boundary, with the index k of the largest gap suggesting the optimal number of clusters.\u003c/p\u003e \u003cp\u003eThen, to capture this optimal number of clusters, the sigma sweep is performed. In this procedure, sigma (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e) is varied logarithmically across a wide range, and the eigengap curve is obtained for each \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e value. This approach enables the identification of modes of clustering. Each mode corresponds to a stable eigengap region, indicating a particular cluster structure. As we zoom in on the cluster space by adjusting the free parameter \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e, sub-clusters start to form within the feature space. It is possible to detect several small clusters submerged within the initial supercluster at various levels of zooming. Hence, the eigengap curve that remains dominant across the widest range of log (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e) is taken to represent the most robust and scale-consistent cluster configuration.\u003c/p\u003e \u003cp\u003eTherefore, this process provides a principle, data-driven mechanism to determine the ideal number of clusters. In essence, the required number of clusters (k) for the most persistent cluster formation under zooming is taken as the fundamental organizational structure of the data distribution, which is indicated by the eigengap curve that predominates over a wide range of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e values.\u003c/p\u003e \u003cp\u003eThe k-means algorithm is an unsupervised machine learning algorithm that groups similar data points into k clusters by targeting to minimize the distance between data points within each cluster. Mathematically, it aims to minimize the objective function in Eq.\u0026nbsp;(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e):\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e\n$$\\:J=\\:\\sum\\:_{i=1}^{k}\\sum\\:_{{x}_{j}\\in\\:{C}_{i}}{‖{x}_{j}-{\\mu\\:}_{i}‖}^{2}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere C\u003csub\u003ei\u003c/sub\u003e denotes the i\u003csup\u003eth\u003c/sup\u003e cluster and \u0026micro;\u003csub\u003ei\u003c/sub\u003e is its centroid. The algorithm proceeds iteratively through two primary steps:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eAssignment of each data point to its nearest centroid based on Euclidean distance.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eThe centroids are updated with the recalculation as the mean of all data points assigned to each cluster.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eThis process repeats until convergence, typically when centroid positions stabilize or the reduction in J falls below a threshold. Accordingly, in this study, the k-means clustering is applied to the feature space to obtain the final clusters C\u003csub\u003e0\u003c/sub\u003e,\u0026hellip;, C\u003csub\u003ek\u0026minus;1\u003c/sub\u003e, after determining the optimal number of clusters (k) through the sigma-sweep-based spectral clustering procedure.\u003c/p\u003e \u003cp\u003e \u003cb\u003eRobustness and Sensitivity Analysis for the Factor Structure and Selection of the Optimal Cluster Number with\u003c/b\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e \u003cb\u003eVariation\u003c/b\u003e\u003c/p\u003e \u003cp\u003eThe robustness of the factor structure was supported by the acceptable internal consistency and construct validity. The Kaiser\u0026ndash;Meyer\u0026ndash;Olkin (KMO) value of 0.61 exceeded the minimum threshold of 0.60(Kaiser, H. F. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1974\u003c/span\u003e), supporting the sampling adequacy for factor extraction.\u003c/p\u003e \u003cp\u003eConfirmatory factor analysis was conducted using robust maximum likelihood estimation (Askelund, A.D. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2019\u003c/span\u003e) with Huber\u0026ndash;White robust standard errors (White, H., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e1980\u003c/span\u003e) and Yuan\u0026ndash;Bentler scaled test statistics(Yuan, K.-H. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1980\u003c/span\u003e), accounting for deviations from multivariate normality (Mardia\u0026rsquo;s skewness\u0026thinsp;=\u0026thinsp;188 935.12, kurtosis\u0026thinsp;=\u0026thinsp;626.19, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Henze\u0026ndash;Zirkler\u0026thinsp;=\u0026thinsp;9.74, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The model had an acceptable incremental fit (robust CFI\u0026thinsp;=\u0026thinsp;0.93, robust TLI\u0026thinsp;=\u0026thinsp;0.90), consistent with the conventional thresholds (\u0026ge;\u0026thinsp;0.90 acceptable(Bentler, P. M., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1990\u003c/span\u003e; Tuominen, L.S. et al, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2022\u003c/span\u003e ), \u0026ge; 0.95 good(Hu, L. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1999\u003c/span\u003e)). The absolute fit of the model was good (robust RMSEA\u0026thinsp;=\u0026thinsp;0.03 [90% CI 0.029\u0026ndash;0.035], robust SRMR\u0026thinsp;=\u0026thinsp;0.04), falling well below the conventional thresholds for good fit(Hu, L. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1999\u003c/span\u003e) (RMSEA\u0026thinsp;\u0026le;\u0026thinsp;0.06, SRMR\u0026thinsp;\u0026le;\u0026thinsp;0.08). The scaled χ\u0026sup2;/df ratio of 2.47 indicated a good absolute model fit, consistent with the literature recommending values below 3.0(Schermelleh-Engel et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2003\u003c/span\u003e; Kline, R.B., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2023\u003c/span\u003e). Model adequacy is best confirmed when theoretical plausibility exists alongside multiple complementary indicators(Stone, B. M., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e). The 13 latent factors are theoretically justified as they operationalize the conceptual framework of behavioral adaptation.\u003c/p\u003e \u003cp\u003eThe tunable \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e parameter in Spectral Clustering acts as a multi-scale lens that controls the level of granularity in the similarity graph, making the method particularly robust for identifying the optimal number of behavioral clusters in the high-dimensional survey dataset. By varying sigma, it progressively zooms through different cluster formations in the data, where small sigma values emphasize fine-grained, local similarities, while large sigma values capture broader, global relationships among data points. Observing the eigengap curve\u0026rsquo;s nature across this sigma sweep allows the detection of stable modes of clustering, i.e., regions where the eigengap remains consistently dominant, which indicates scale-invariant and structurally coherent clusters. This procedure mitigates sensitivity to initialization or noise, ensuring that the selected cluster number reflects the most persistent and intrinsic structure of the data rather than arbitrary local variations revealing the most robust and interpretable latent organization of the behavioural data.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eIntegration of Statistical and Unsupervised Machine Learning with Behavioral Theory\u003c/h3\u003e\n\u003cp\u003e \u003c/p\u003e \u003cp\u003eA key objective of this study was to expose the latent behaviour of human adaptation during a global disruption. Rather than considering individual mobility alone, we adopted a completely data-driven strategy with Factor Analysis and Unsupervised Machine Learning to identify underlying behavioral constructs consistent with the behavioral science theory that observed actions reflect deeper latent strategies and constraints rather than independent variables. Factor analysis enables the detection of multivariate behavioral patterns emerging from co-occurring routines and decision-making processes, aligning with socioecological perspectives where behavior is embedded in interdependent domains of daily life (e.g., social interaction, work exposure, digital engagement, access to resources). By extracting orthogonal, interpretable factors, we capture distinct adaptive subsystems, each representing a coherent response to environmental constraints, risk, and opportunity. This approach aligns with theories of behavioral ecology, which emphasize that individuals optimize behavior within contextual constraints and bounded rationality (Simon, H. A.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1955\u003c/span\u003e), whereby people act based on feasible options shaped by structural realities.\u003c/p\u003e \u003cp\u003eBuilding on these latent dimensions, we applied a clustering framework to identify behavioral archetypes where discrete, emergent adaptation patterns represent strategic configurations of behavior, not just socioeconomic groupings. Using spectral clustering is theoretically appropriate because crisis behavior is often non-linear, non-hierarchical, and shaped by complex interactions between structural constraints and adaptive strategies. Spectral methods can detect non-convex cluster boundaries and subtle divisions in behavior space that linear methods may overlook, reflecting the possibility that similar behavioral outcomes (e.g., reduced mobility) may arise from different mechanisms (e.g., remote work vs. unemployment). The subsequent use of k-means for assignment supports interpretability and reinforces each archetype as a distinct mode of behavioral plasticity, inertia, substitution, or compensation.\u003c/p\u003e \u003cp\u003eCrucially, this analytical design is grounded in behavioral theory. We assume that individuals adapt using available behavioral capacity defined by their substitution capacity (digital and occupational flexibility), exposure necessity, and support infrastructure. These determinants correspond to well-established constructs in behavioral economics, sociology, and social psychology, including agency vs. structural constraint, risk perception conditioned by necessity, coping strategies, and the socioecological model of behavior. Thus, our methodological choices are not merely statistical; they implement a theoretical framework in which behavior is viewed as situated, structured, and adaptive. The resulting archetypes reveal how structural inequality becomes behavioral inequality, consistent with the emerging recognition that crises produce K-shaped trajectories of adaptation and recovery.\u003c/p\u003e \u003cp\u003eIntegration of the latent variable modeling with unsupervised clustering, conditioned on the behaviour changes triggered by the pandemic, revealed a hidden structure in the data organization, moving beyond a descriptive mobility analysis and providing a theoretically grounded, data-driven map of human adaptation, qualified by socioeconomics, demographics, and numerous other factors as highlighted in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. This allows us to identify not only what behaviors occurred, but why they occurred, how they co-occur, and how they reflect deeper mechanisms of resilience, vulnerability, substitution, dissent, and protest. Thus, the statistical model becomes an explicit test of behavioral theory and a foundation for conceptual generalization.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eCore Dimensions of Mobility and Behavioral Adaptation\u003c/h2\u003e\n \u003cp\u003eWe employed PCA together with exploratory factor analysis using Varimax rotation to reduce the information in the high-dimensional survey dataset into 13 principal factors or interpretable behavioral dimensions (Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e) that collectively summarize the structure of human mobility and adaptation under pandemic constraints. We determined which factors to retain based on the Kaiser criterion (Kaiser, H. F.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e1960\u003c/span\u003e ) and examined the scree plot, aiming to maintain simplicity while preserving explanatory strength as indicated in Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eCollectively, these factors presented accounted for 63.17% of the total variance (Hair, J. F. J. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2018\u003c/span\u003e) in the behavioral data while remaining conceptually aligned with mechanisms of decision-making under constraint, social capital, infrastructural dependency, and digital capacity, providing a theoretically meaningful basis for studying how people reorganize behavior when risk, rules, and resources change. Factor-loadings in Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e link the survey questions to the extracted latent behavioural factors, where each loading represents the strength of association between a question that highlights an observed behavioural indicator and its corresponding factor, revealing distinct domains of human mobility and interaction. They highlight subsystems of everyday behavior, from leisure and social connection to work exposure strategies, infrastructure access, transport portfolios, local resource ecologies, spatial routines, and technology-mediated activity.\u0026nbsp;\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eEmergence of behavioral archetypes through clustering\u003c/h2\u003e\n \u003cp\u003eFollowing factor analysis, we identified 4 robust behavioral archetypes via a two-stage clustering pipeline. The initial step involved using PCA to reduce the original dataset\u0026apos;s dimensionality to 13 dimensions to denoise and enhance the data for subsequent clustering. Spectral clustering was then applied to this refined dataset to determine the optimal number of latent clusters (k), as illustrated in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. Finally, K-means clustering was employed to assign each data point to one of the clusters, which captured distinct strategies of adaptation rather than being mere statistical groupings, with privilege-enabled substitution into remote work, tradition-anchored risk management, stability-seeking minimalism under budget constraints, and necessity-driven communal resurgence as shown in Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eThe first Cluster 0, the \u0026ldquo;Privileged Adaptors,\u0026rdquo; comprises individuals with strong structural and demographic privilege, being among the Sinhalese majority, and with the highest educational qualifications.\u003c/p\u003e\n \u003cp\u003eThey demonstrated behavioural flexibility and resilience, maintaining middle-to-higher income and formal employment in secure sectors, reflecting greater institutional and economic access. Hence, they empowered a rapid transition to remote work, enabled by digital readiness and occupational flexibility, leading to a marked decline in physical commuting and in-person interactions. Mobility for essential needs reflected resilient maintenance behaviours. Their elevated education and economic stability facilitated informed risk assessment and deliberate behavioural restraint, underscoring how privilege enabled both the capacity and the choice to minimize exposure.\u003c/p\u003e\n \u003cp\u003eCluster 1, the \u0026ldquo;Cautious Traditionalists,\u0026rdquo; represents a lower-middle class with moderate education. Their mixed ethnic composition reflected typical mid-tier socioeconomic diversity in Sri Lanka.\u003c/p\u003e\n \u003cp\u003eThey adopted a risk-averse stabilization strategy grounded in traditional, localized adaptation and showed a high income volatility, i.e., many slipped into low-income brackets during early waves and recovered only partially. Compared to Cluster 0, the limited occupational and digital flexibility of Cluster 1 constrained their adaptive options. Their strong pre-pandemic social participation shifted into a highly cautious re-engagement, driven by economic vulnerability and risk aversion, relying moderately on private transport with a sustained drop in public transport usage. Their high home delivery and mobile shops usage reflected proximal sourcing and local dependency as protective strategies.\u003c/p\u003e\n \u003cp\u003eCluster 2, the \u0026ldquo;Stable Minimalists,\u0026rdquo; represents the financially disrupted lower-middle class. With low educational attainment and informal employment, they reflect limited adaptive capacity. Their mixed ethnic composition also mirrors the national socioeconomic middle in Sri Lanka, similar to cluster 1.\u003c/p\u003e\n \u003cp\u003eThey demonstrated strong behavioral inertia and necessity-driven resilience. Many fell into low-income brackets from the lower-middle-income tier, with only partial recovery. Although cluster 2 exhibited how the other socioeconomic constraints shaped their rigid maintenance behaviours: private commuting remained minimal, while active mobility such as walking and cycling for work persisted, reflecting economic necessity rather than adaptive choice. Their social participation and digital engagement stayed consistently low, reinforcing a social and digital divide. With moderately stable jobs, they adhered to a budget-friendly financial routine. Collectively, this group embodies constrained behaviour and continuity of low-cost, proximal routines.\u003c/p\u003e\n \u003cp\u003eCluster 3, the \u0026ldquo;Resilient Rebounders,\u0026rdquo; embodies the most socioeconomically vulnerable, displaying behavioral rebound amid severe structural fragility. Persistently in the low-income tier, they rely on daily wages and informal employment, with the lowest educational attainment and minimal digital capacity. Demographically, this is the most ethnically diverse and marginalized cluster, with strong minority representation, reflecting concentrated systemic vulnerability.\u003c/p\u003e\n \u003cp\u003eDespite lacking financial or digital buffers, this group showed a powerful mobility resurgence with social and recreational participation rebounding sharply, suggesting social resilience and collective recovery. Their active commuting and public transport usage rose substantially, driven by economic necessity and communal interdependence.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eSocioeconomic and demographic mechanisms underlying archetypes\u003c/h2\u003e\n \u003cp\u003eA clear socioeconomic division emerges across the identified behavioural clusters, driven by unequal access to education, formal and informal employment (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ec, \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ed). The privileged adaptors hold the highest academic attainment, with nearly 30% with advanced or university degrees and secure positions in government or private sectors (71.8%), providing occupational stability and virtual substitution capacity. The cautious traditionalists, with mixed schooling and middle-tier income, exhibit limited credentialing and moderate in-person exposure, managing risk through localized coping. The stable minimalists and resilient rebounders faced constraints in adaptive flexibility and reinforced cost-sensitive, necessity-driven mobility.\u003c/p\u003e\n \u003cp\u003eEthno-geographic patterns amplify these structural disparities (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ea, \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003ee); Cluster 0 is predominantly Sinhalese and urban, while Cluster 3 is more ethnically diverse and rural, overrepresented among Tamil and Muslim communities in underdeveloped districts. Thus, behavioral capacity is stratified by structural opportunity, with education, occupation, and geography collectively anchoring the inequality-adaptation nexus in human mobility.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003eIncome trajectories and economic resilience across waves\u003c/h2\u003e\n \u003cp\u003eHousehold income trajectories reveal deep asymmetries in economic resilience across archetypes (Fig. \u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003eb). Pre-pandemic inequalities shaped exposure to shocks and recovery potential. The privileged adaptors, concentrated in Middle-to-Higher income categories (\u0026gt;\u0026thinsp;85%), absorbed an initial first-wave decline but rapidly recovered, supported by formal employment, digital access, and substitution. The cautious traditionalists showed a sharper contraction, with Low income increasing from 4.9% to 33.4%, and an incomplete rebound by the third wave, marking fragile middle-class stability and constrained adaptive flexibility.\u003c/p\u003e\n \u003cp\u003eThe Lower-Middle income of the stable minimalists suffered persistent downgrades, as the Low-income share rose from 4.8% to 35.2% with limited recovery, reflecting dependence on on-site, low-buffer occupations and restricted financial adaptability. The resilient rebounders experienced the most severe and lasting collapse, shifting from 47.2% in the Lower-Middle category to 68.7% in the Low-income category, aligning with informal, daily-wage labor and income, behavior decoupling, and mobility rebound without financial recovery. Their structural downgrade, starkly contrasting with national trends, which recorded a poverty rate of 11.3% in 2019 to 13.1% in 2021 (World Bank Group, \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2023\u003c/span\u003e), isolates Cluster 3 as the most vulnerable group. These trajectories expose how structural precarity and employment informality entrench vulnerability, while privilege sustains economic and behavioral resilience across crises.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003eBehavioral Adaptation Across Pandemic Phases\u003c/h2\u003e\n \u003cp\u003eTemporal evolution of the extracted behavioural factors across the pre-pandemic and the three major waves of COVID-19 (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003e) captures the heterogeneous responses of different societal clusters.\u003c/p\u003e\n \u003cp\u003eSocial and recreational mobility exhibited a universal decline across all clusters during the first pandemic wave (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003ea), indicating widespread compliance with movement restrictions and collective behavioural suppression, with heterogeneous rebounds thereafter. Notably, Clusters 1 and 3 demonstrated stronger recovery trajectories, reflecting, respectively, social reconnection managed by risk-averse routines and necessity-driven return to communal life, whereas Clusters 0 and 2 remained consistently suppressed, reflecting sustained caution and restricted social participation due to voluntary caution and resource-bounded routine, respectively. Cluster 1\u0026rsquo;s recovery aligns with its pre-existing social connectivity and active community engagement, while Cluster 3, initially a less recreationally active cluster, displayed a more spirited rebound, potentially reflecting the compensatory pursuit of social interaction following prolonged restriction. In contrast, Cluster 0 maintained a cautious lifestyle with limited exposure, and Cluster 2 preserved habitual routines with minimal change. These divergent rebound paths illustrate that social recovery is not uniform but stratified by socioeconomic conditions, social capital, and risk perception, suggesting further that elevated risk perception among more privileged groups, while promoting safety, may simultaneously foster overcaution and social isolation, potentially heightening experiences of loneliness, emptiness, and be more inclined to use social media as coping mechanisms within these households(Senarath, N. et al. ,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2024\u003c/span\u003e; Thilakasiri, I. et al. ,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003ePrivate commuting and work exposure revealed stratification reflecting class-based occupational flexibility (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eb). Clusters 0 and 3 converged at lower mobility levels during the first wave, but for opposing reasons: Cluster 0 benefited from the privilege of remote work, while Cluster 3 likely experienced unemployment or underemployment. Clusters 1 and 2 maintained moderate exposure, consistent with their continued on-site work obligations. This divergence underscores how both privilege and precarity can manifest as reduced mobility, albeit through entirely different mechanisms. The persistence of these differentiated mobility trajectories through subsequent waves highlights the embedded socioeconomic inequalities shaping pandemic-era behavioural adaptation.\u003c/p\u003e\n \u003cp\u003eHome-based work isolation exhibited sharp polarization across clusters (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003ec). Cluster 0 experienced a pronounced rise during the first wave, reflecting a rapid transition to remote work enabled by occupational and digital privilege. Other clusters remained largely static, evidencing limited remote-work feasibility and consequent in-person exposure by obligation (stable minimalists), by partial necessity with caution (cautious traditionalists), or by precarity (resilient rebounders). This divide underscores the emergence of a digital and occupational inequality, where remote work became a behavioural marker of both safety and privilege.\u003c/p\u003e\n \u003cp\u003eBoth financial and essential service access demonstrated marked resilience for Clusters 0 and 1, who maintained steady or slightly increasing mobility across waves (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003ed,f). In contrast, Clusters 2 and 3 exhibited sustained suppression, reflecting economic vulnerability and constrained access to infrastructure. The stability of access-related mobility among higher-income groups reveals the buffering effect of resource availability, while the contrast in lower-income groups points to a disproportionate impact of logistical and financial disruptions. Unlike the earlier factors, which had a nonlinear, complex relationship to economic status, this factor showcased a linear variation unraveled by the detailed data-driven cluster analysis.\u003c/p\u003e\n \u003cp\u003eMobility related to public and hired transport declined sharply across all clusters during the first wave (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003ee,i). However, only Cluster 2, characterized by essential work and limited alternatives, demonstrated partial recovery, reflecting the necessity-driven dependence on shared mobility. Clusters 0 and 1 maintained suppressed usage, substituting private or remote modes of work, illustrating a socioeconomic gradient in transport resilience, where exposure risk was unequally distributed according to occupational necessity and economic means.\u003c/p\u003e\n \u003cp\u003eBicycle mobility and work exposure (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003eg,h) reveal how occupational demands shape behavioural adaptation across socioeconomic groups. Cluster 2 demonstrates persistent physical commuting and high interpersonal exposure, reflecting necessity-driven resilience under limited remote work options. In contrast, clusters 0 and 3 maintain low levels of mobility and exposure, cluster 0 benefiting from digital flexibility, while cluster 3 faces unemployment.\u003c/p\u003e\n \u003cp\u003eAccess to community resources and group transport displayed complementary behavioural patterns (Figs. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003ej,k). Cluster 1 showed sustained engagement with local networks frequenting delivery services and neighborhood shops, indicating strong communal ties and adaptive social capital. Meanwhile, Clusters 0 and 2 exhibited consistently low community interaction, with Cluster 0 relying on digital systems and Cluster 2 constrained by economic and spatial limitations. Shared transport mobility remained low throughout, highlighting collective risk aversion and reduced interpersonal proximity.\u003c/p\u003e\n \u003cp\u003eActive commuting, such as walking to workplaces, remained remarkably stable for Cluster 2 across all pandemic waves (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003el). This implies either a structural necessity or a habitual attachment to routine movement in low-radius livelihoods. Hence, indicating a form of behavioural inertia rooted in occupational rigidity and limited adaptive flexibility.\u003c/p\u003e\n \u003cp\u003eDigital platform engagement of Cluster 0 increased significantly, and it was closely followed by Cluster 1, whose rise still surpassed that of Clusters 2 and 3 during the first wave and plateaued thereafter (Fig. \u003cspan class=\"InternalRef\"\u003e7\u003c/span\u003em). Clusters 0 and 1 leveraged digital tools for work, shopping, and communication, reflecting both educational attainment and infrastructural access. In contrast, Clusters 2 and 3 demonstrated minimal digital engagement, underscoring persistent digital divides that parallel economic stratification. The resulting behavioural dichotomy reflects how technology-mediated adaptation became both a buffer and a barrier across socioeconomic tiers. These findings show that effective policy must recognize that divergent behaviours often stem from structural constraints rather than choice, and that risk-mitigation strategies evident for some may be inaccessible or impractical for others (Iio, K. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; Carranza, A. et al. ,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2022\u003c/span\u003e; Chiou, L. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e ).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003eInterdependencies and behavioral trade-offs\u003c/h2\u003e\n \u003cp\u003eBehavioral coupling across the system underscores how digital access, work security, and social capital mediated behavioral flexibility, exposing the feedback loops between risk, resources, and routine. Cluster 0 demonstrated the strongest adaptive capacity with a sharp reduction in Private Commuting and Work Exposure and an increment in Home-Based Work Isolation, facilitating a rapid transition to remote work. This was accompanied by decreased Bicycle Mobility and Variable Work Exposure, and a surge in Digital Platform Engagement, a key enabler for maintaining occupational and financial continuity through online services.\u003c/p\u003e\n \u003cp\u003eFor Clusters 2 and 3, digital scarcity forced physical presence and transport substitution, increasing exposure and income instability, thereby reinforcing structural precarity. Financial Access Points and Essential Services disruptions of Cluster 3 were compensated by reliance on Community Resource Sourcing. For Cluster 1, community capital acted as a partial hedge, enabling a balance between exposure and livelihood through localized, traditional coping mechanisms.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eEmergent conceptual model of crisis adaptation\u003c/h2\u003e\n \u003cp\u003eWe propose a conceptual model of behavioral adaptation under collective disruption with the following determinants which shape whether individuals exhibit behavioral plasticity or inertia, experience agency or compulsion, and achieve resilience or cumulative disadvantage (Fig. \u003cspan class=\"InternalRef\"\u003e8\u003c/span\u003e):\u003c/p\u003e\n \u003cp\u003e1. Substitution capacity (digital access, occupational flexibility) determines the range of practical behavioral substitutes for risky activities.\u003c/p\u003e\n \u003cp\u003e2. Exposure necessity (formal/informal labor structures, transport portfolios) influence the cost of sustaining livelihoods when mobility is limited.\u003c/p\u003e\n \u003cp\u003e3. Support infrastructure (local networks, institutional access, urbanity) influences resilience as opposed to path-dependence.\u003c/p\u003e\n \u003cp\u003eWithin this framework, inequality is reframed not only as a difference in income or resources, but as a difference in behavioral capacity, which determines the ability to opt for safer alternatives when circumstances change. This reconceptualization aligns with socioecological theories of behavior while adding further extensions in a new crisis-specific context.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe mobility and behavioral distinctions during the pandemic show a complex relationship between common shock responses and deep-rooted socioeconomic divisions (van de Weijer, M. P. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2022\u003c/span\u003e; Yabe, T. et al,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2024\u003c/span\u003e; Boz, H. A. et al. ,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2024\u003c/span\u003e; Bavel, J.J.V. et al. ,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e; Delaney, L. ,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2024\u003c/span\u003e ).\u003c/p\u003e \u003cp\u003eThe data-driven analysis has revealed that behavioral responses to the pandemic were characterized by significant non-linearities relative to socioeconomic status. While a clear gradient in income capacity is evident across clusters 0 to 3, other critical dimensions, such as resilience in leisure activities, digital adaptation, and health-protective behaviors, deviate from this linear progression. For example, the most economically vulnerable Cluster 3 exhibited a stronger recovery in social and recreational mobility than the other groups. This illustrates that human responses to the crisis are mediated by psychological resilience, cultural norms, access to networks, and structural necessity, not just economic position, challenging simplistic assumptions linking poverty and suppressed behaviors. The existence of these tiered, non-proportional relationships exposes a multifaceted societal structure that was laid bare by the pandemic's disruptive force.\u003c/p\u003e \u003cp\u003eThe pandemic revealed at least two distinct behavioral logics: proactive substitution and reactive compensation. Privileged Adaptors exemplified the first logic by immediate digital substitution. They engaged in behavioral plasticity facilitated by high substitutability, through remote work, online banking, and virtual social interaction. This aligns with theories of agentic flexibility, where they are often able to restructure routines to maintain stability. The rapid rise in Home-Based Work Isolation and Digital Platform Engagement, paired with reductions in work exposure and physical mobility, redefined daily life through technology. This group showed bounded rationality under low constraints in addition to high perceived control, allowing them to prioritize health without sacrificing income.\u003c/p\u003e \u003cp\u003eHowever, Stable Minimalists and Resilient Rebounders showed behavioral inertia driven by structural constraint. Their lack of access to resources and opportunities translated to their choices being necessity-driven responses as opposed to conscious behavioral choices. They were required to be physically present even during critical pandemic phases to obtain an income. This aligns with theories of bounded rationality, where individuals operate with constrained freedom of decision-making owing to inequalities. Our analysis interprets this phenomenon as risk absorption produced by limited agency, as opposed to risk-seeking, demonstrating that risk perception is not merely psychological, but structurally conditioned.\u003c/p\u003e \u003cp\u003eAdditionally, the analysis exposed situations where different clusters exhibited similar adaptive behaviours, yet for fundamentally different reasons, where the shared strategies emerged from distinct structural constraints, motivations, and resources unique to each group. As an instance, both Clusters 0 and 3 had a sharp decline in public transport usage during the first wave. While Cluster 0\u0026rsquo;s reasoning was voluntary avoidance enabled by remote work, Cluster 3 experienced involuntary exclusion caused by job loss and service disruption. Hence, the same mobility reduction diverged into opposite outcomes, where Cluster 0 rapidly recovered economically, whereas Cluster 3 experienced prolonged economic decline. This shows that behavior cannot be understood in isolation from the structural conditions under which it occurs.\u003c/p\u003e \u003cp\u003eAnother key finding was the crucial role of digital capital as a determinant of behavioral flexibility. Clusters 0 and 1, having higher educational levels and technological literacy, used digital access as a substitution technology, switching to virtual equivalents for work and other needs. Clusters 2 and 3 had a contrasting experience with limited ability to switch to digital resources, compelling them to rely on communal resource sharing or government and NGO support.\u003c/p\u003e \u003cp\u003eSocial capital also emerged as an alternative form of adaptive capacity in the absence of digital infrastructure. For example, Cluster 1 utilised localized community networks and informal delivery systems to deal with disruption, relying on relational infrastructure rather than shifting to digital platforms. The coping mechanisms to achieve stability for Cluster 3 were reliance on government/NGOs and re-engagement in social gatherings, demonstrating that communal strategies can function as parallel adaptive systems, specifically when digital alternatives are inaccessible.\u003c/p\u003e \u003cp\u003eCluster 2 displayed an adaptation strategy based on preserving routine as opposed to transformation, representing behavioural inertia rooted in occupational immobility, economic necessity. This was prominent through Factors 2, 7, and 12, where they had consistent movements, moderate work exposure, and relatively low digital engagement. Crucially, this points to an interesting observation that the absence of change itself is a behavioural outcome shaped by structural changes.\u003c/p\u003e \u003cp\u003eThe dynamics of poverty traps and behavioural lock-in are further highlighted through the contrasting responses between Clusters 2 and 3, even though both lacked digital access and higher education. While Cluster 2 maintained low-cost, proximate routines offering small possibilities for upward movement, Cluster 3 showed high physical mobility, in the face of greater economic shocks, as a mechanism to rebuild social and economic stability. However, rather than translating into financial recovery, this led to a structural downgrade in income, even after the superficial recovery of behavioural patterns. This created insights on how behavioural resilience does not guarantee socioeconomic resilience, especially when underlying structures do not support upward movement, hence masking vulnerability.\u003c/p\u003e \u003cp\u003eHence, the clusters demonstrate that adaptive strategies are systematically patterned by the intersection of economic, digital, and social capital. While privileged groups showed individualized, technology-enabled adaptation, moderately resourced groups utilised community-based coping mechanisms, and the more structurally under-privileged individuals had to rely on behavioural inertia or necessity-driven exposure. A socioecological model of behavioural adaptation was supported through these findings, revealing a hierarchy of adaptive freedom, where more resources translated to a wider range of possible strategies. Individual resources, community networks, and institutional access were also factors that contributed in leading to these outcomes. Our findings further highlight that the most powerful determinant of behaviour in crisis is structural opportunity for viable strategies.\u003c/p\u003e \u003cp\u003eBy integrating these insights, we propose a behavioral framework for crisis adaptation under inequality. At its core, the framework suggests that adaptation capacity is determined by three interacting elements: substitution capacity, exposure necessity, and support infrastructure. Cluster 0 had high substitution capacity, low exposure necessity, and stable support, producing rapid, protective adaptation, enabling the risk aversion to become the Privileged Adaptors. Cluster 1 had moderate substitution and strong community support, producing cautious adaptation. Cluster 2 had low substitution and high exposure necessity, producing behavioral inertia. Cluster 3 had minimal substitution and high exposure necessity, but high communal support, producing a reactive but ultimately economically costly rebound. These differentiated pathways underscore that behavioral flexibility is not a trait but a privilege, which is a function of the resources and structures that surround individuals.\u003c/p\u003e \u003cp\u003eAlthough these identified mechanisms are based on a Sri Lankan context, they encapsulate globally relevant mechanisms of inequality: technological substitutability, livelihood-based exposure, and the buffering role of formal and informal support, which are generalizable beyond Sri Lanka. It offers a transferable lens for understanding heterogeneous crisis adaptations in other countries, where similar constraints shape the behavioral trajectories that emerge under large-scale societal shocks. For example, digital access determined who could maintain economic stability (Francis, D. V. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; Jahan, N. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2023\u003c/span\u003e) across the globe. While high-income employees easily switched to remote work, the less fortunate, such as immigrant labourers and minority communities worldwide, had to rely on community or NGO support in the absence of institutional safety nets (Fan, B. E. ,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e). Crisis reveals and accelerates pre-existing fault lines in behavioural opportunity spaces as opposed to creating new divisions. The structural inequalities are amplified in a feedback loop (Marmot, M. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e; Kniffin, K. M. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2021\u003c/span\u003e; Li, H. et al., \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2023\u003c/span\u003e; Bambra, C. et al.,\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e2020\u003c/span\u003e ).\u003c/p\u003e \u003cp\u003eInsights for policy and societal resilience are also revealed through our findings. They show that compliance is conditional on structural feasibility, whereas often public health strategies assume voluntary compliance. But the compliance rooted in capacity, flexibility, and necessity can result in resistance at times. In the face of challenges such as a lack of digital access, livelihoods depending on physical presence, simple instructions were neither effective nor sufficient. Therefore, effective policy should focus on expanding behavioural capacity through the improvement of digital infrastructure, enabling remote work opportunities, and strengthening social safety nets. Systems must be designed to facilitate adaptation, as opposed to expecting individuals to adapt to systems.\u003c/p\u003e \u003cp\u003eThis study advances human behavioural science by demonstrating that adaptation during crisis is shaped primarily by structural opportunity and the knowledge, awareness, and ideology, rather than being driven by individual preference alone. Behavioural flexibility was shown to be a form of privilege, whereas structural constraints were the cause for risk exposure and behavioural inertia rather than choice. Our behavioural framework explains why crises produce K-shaped recoveries and why resilience can coexist with vulnerability, shifting the focus from \u0026ldquo;what people do\u0026rdquo; to \u0026ldquo;what people are able to do\u0026rdquo;. This reconceptualization opens new directions for behavioural research and policy: rather than attempting to enforce uniform behaviors, effective interventions must build the structural conditions that enable equitable adaptation.\u003c/p\u003e \u003cp\u003eOne limitation of our study is that it may not fully capture cultural nuances or political and psychological motivations, as the data are from the Sri Lankan context. However, it is important to note that the structural mechanisms identified are consistent with findings across diverse global settings. Future research should test this framework across cultures while exploring how long-term behavioural changes persist or change once the crisis subsides, and there is also a need to integrate behavioural data into the process of policy making, focusing on which structural changes are capable of bringing forth the most effective adaptive capacity enhancements for the vulnerable.\u003c/p\u003e \u003cp\u003eIn conclusion, the pandemic reconfigured the entire architecture of human behaviour, revealing how adaptation is deeply connected to structural inequalities. Our findings offer a conceptual model of crisis adaptation, highlighting structural opportunity, digital and employment capital, and social networks as core determinants of human behaviour, while also illustrating that behavioural patterns are non-linear, stratified, and interdependent. This study sheds light on the mechanisms that rule adaptation, contributing to a deeper understanding of mobility under constraints, with the potential to aid in designing more resilient societies.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe dataset used for the research is accessible in Harvard Dataverse, named \u0026ldquo;A dataset on the socioeconomic and behavioural impacts in Sri Lanka through multiple waves of COVID-19\u0026rdquo;. (Ilangarathna, G. A. et al, 2023).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWith assistance from the Department of Sociology, this study received ethical clearance from the Ethical Review Committee, Faculty of Arts, University of Peradeniya, Sri Lanka (ARTS/ERC/2021/01, September 18, 2021). Additionally, the Administrative clearance was obtained from the relevant Divisional Secretariat (DS) offices, Grama Niladari (GN) offices, and Sri Lanka\u0026apos;s Ministry of Home Affairs, for the involvement of human participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInformed Consent\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEvery participant provided written and verbal informed consent, and their voluntary involvement was guaranteed. Following established procedures and guidelines, data confidentiality and privacy were preserved throughout the whole data collection process (Ilangarathna, G. A. et al, 2024).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions Statement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eM.K.\u0026ndash; Conceived and designed the experiments, Performed the experiments, Analyzed the data, Contributed materials/analysis tools, Wrote the paper. \u0026nbsp;L.E. - Performed the experiments, Analyzed the data, Contributed materials/analysis tools, Wrote the paper. \u0026nbsp;I.M.- Analyzed the data, Contributed materials/analysis tools, Wrote the paper. \u0026nbsp; R.G. - Conceived and designed the experiments, Analyzed the data, Wrote the paper. \u0026nbsp;P.E. - Wrote the paper. \u0026nbsp;N.K. - Analyzed the data, Wrote the paper. \u0026nbsp;S.D. - Wrote the paper. \u0026nbsp;V.H - Wrote the paper. \u0026nbsp;J.E. - Analyzed the data, Wrote the paper.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAcar G\u0026uuml;vendi̇r M, \u0026Ouml;zer \u0026Ouml;zkan Y (2022) Item removal strategies conducted in exploratory factor analysis: A comparative study. Int J Assess Tools Educ 9:165\u0026ndash;180\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAgampodi S (2021) Enhancing the Sri Lankan response to COVID-19 - a communique from the College of Community Physicians of Sri lanka. J Coll Community Physicians Sri Lanka 27:271\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhmed SA, Majoka Z, Saha J (2023) Revisiting Informality in South Asia: An Agenda for the Future Social Protection and Jobs. Policy Note. Partnership for Economic Inclusion\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndronico A et al (2021) Evaluating the impact of curfews and other measures on SARS-COV-2 transmission in French Guiana. Nat Commun 12\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnnual Report (2020) : Central Bank of Sri Lanka. \u003cem\u003eAnnual Report 2020 | Central Bank of Sri Lanka\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cbsl.gov.lk/en/publications/economic-and-financial-reports/annual-reports/annual-report-2020\u003c/span\u003e\u003cspan address=\"https://www.cbsl.gov.lk/en/publications/economic-and-financial-reports/annual-reports/annual-report-2020\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (Accessed: 2nd November 2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnnual report 2022 (2025) Central Bank of Sri Lanka. Annual Report 2022 | Central Bank of Sri Lanka Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cbsl.gov.lk/en/publications/economic-and-financial-reports/annual-reports/annual-report-2022\u003c/span\u003e\u003cspan address=\"https://www.cbsl.gov.lk/en/publications/economic-and-financial-reports/annual-reports/annual-report-2022\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 2nd November\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAskelund AD, Schweizer S, Goodyer IM et al (2019) Positive memory specificity is associated with reduced vulnerability to depression. Nat Hum Behav 3:265\u0026ndash;273. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41562-018-0504-3\u003c/span\u003e\u003cspan address=\"10.1038/s41562-018-0504-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBambra C, Riordan R, Ford J, Matthews F (2020) The COVID-19 pandemic and Health Inequalities. J Epidemiol Commun Health 74:964\u0026ndash;968\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBavel JJV, Baicker K, Boggio PS et al (2020) Using social and behavioural science to support COVID-19 pandemic response. Nat Hum Behav 4:460\u0026ndash;471\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBentler PM (1990) Comparative fit indexes in structural models. Psychol Bull 107:238\u0026ndash;246\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBonaccorsi G et al (2020) Economic and social consequences of human mobility restrictions under covid-19. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 117, 15530\u0026ndash;15535\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoz HA et al (2024) Investigating neighborhood adaptability using Mobility Networks: A case study of the COVID-19 pandemic. Humanit Social Sci Commun 11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBronfenbrenner U (1979) The ecology of human development: Experiments by nature and design. Harvard University Press, Cambridge, MA\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBronfenbrenner U (1977) Toward an experimental ecology of human development. Am Psychol 32:513\u0026ndash;531\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarranza A et al (2022) The Social Divide of social distancing: Shelter-in-place behavior in Santiago during the covid-19 pandemic. Manage Sci 68:2016\u0026ndash;2027\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChiou L, Tucker C (2020) \u003cem\u003eSocial distancing, internet access and inequality\u003c/em\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3386/w26982\u003c/span\u003e\u003cspan address=\"10.3386/w26982\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCohen J, Cohen P, West SG, Aiken LS (2015) Applied multiple regression. Routledge\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCovid-19 cases | WHO COVID-19 Dashboard (2025) \u003cem\u003eWorld Health Organization\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://data.who.int/dashboards/covid19/cases\u003c/span\u003e\u003cspan address=\"https://data.who.int/dashboards/covid19/cases\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 1st November\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDecerf B, Ferreira FHG, Mahler DG, Sterck O (2021) Lives and livelihoods: Estimates of the global mortality and poverty effects of the COVID-19 pandemic. World Dev 146:105561\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDelaney L, Watson E (2024) Behavioural science and the Irish covid-19 response. Behav Econ Policy Pandemics 220\u0026ndash;231\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDurante R, Guiso L, Gulino G (2021) Asocial capital: Civic culture and social distancing during COVID-19. J Public Econ 194:104342\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEgger D et al (2021) Falling living standards during the COVID-19 crisis: Quantitative evidence from nine developing countries. Sci Adv 7\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u003cem\u003eEpidemiology unit coronavirus disease 2019 (COVID-19)\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.epid.gov.lk/epid/public/storage/post/pdfs/en_63ff1e0b4323b_sitrep-sl-en-06-09_10_21_1.pdf\u003c/span\u003e\u003cspan address=\"https://www.epid.gov.lk/epid/public/storage/post/pdfs/en_63ff1e0b4323b_sitrep-sl-en-06-09_10_21_1.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 3rd November (2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFan BE (2021) Migrant workers with covid-19: Recognizing the crucial role non-governmental organizations perform. Lancet Reg Health - Western Pac 9:100145\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerwana I, Varshney LR (2021) Social capital dimensions are differentially associated with covid-19 vaccinations, masks, and physical distancing. PLoS ONE 16\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrancis DV, Weller CE (2021) Economic inequality, the digital divide, and remote learning during COVID-19. Rev Black Political Econ 49:41\u0026ndash;60\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFraser T, Page-Tan C, Aldrich DP (2022) Social Capital\u0026rsquo;s impact on COVID-19 outcomes at local levels. Sci Rep 12\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGalgamuwa LS et al (2023) Spatial distribution of COVID-19 patients in Sri Lanka. BMC Public Health 23\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGDP growth (annual %) (2025) - sri lanka. \u003cem\u003eWorld Bank Open Data\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?locations=LK\u003c/span\u003e\u003cspan address=\"https://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?locations=LK\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 2nd November\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGozzi N et al (2021) Estimating the effect of social inequalities on the mitigation of COVID-19 across communities in Santiago de Chile. Nat Commun 12\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHair JFJ, Black WC, Babin BJ, Anderson RE (2018) Multivariate Data Analysis. Cengage Learning EMEA\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u003cem\u003eHousehold income and expenditure survey \u0026ndash; 2019\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.statistics.gov.lk/Resource/en/IncomeAndExpenditure/HouseholdIncomeandExpenditureSurvey2019FinalResults.pdf\u003c/span\u003e\u003cspan address=\"https://www.statistics.gov.lk/Resource/en/IncomeAndExpenditure/HouseholdIncomeandExpenditureSurvey2019FinalResults.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 2nd November (2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu L, Bentler PM (1999) Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equation Modeling: Multidisciplinary J 6:1\u0026ndash;55\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIio K, Guo X, Kong X, Rees K, Wang B (2021) Covid-19 and social distancing: Disparities in mobility adaptation between income groups. Transp Res Interdisciplinary Perspect 10:100333\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIlangarathna GA et al (2024) A dataset on the socioeconomic and behavioural impacts in Sri Lanka through multiple waves of covid-19. Data Brief 53:110063\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIlangarathna GA et al (2023) Socioeconomic Impact of COVID-19 during three pandemic waves. Harvard Dataverse https://doi.org/10.7910/DVN/CXMJSM\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eInternational Labour Organization (ILO) (2021) *ILO Monitor: COVID-19 and the world of work. Seventh edition*. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ilo.org/wcmsp5/groups/public/---dgreports/---dcomm/documents/briefingnote/wcms_767028.pdf\u003c/span\u003e\u003cspan address=\"https://www.ilo.org/wcmsp5/groups/public/---dgreports/---dcomm/documents/briefingnote/wcms_767028.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJahan N, Zhou Y (2023) Covid-19 and Digital Inclusion: Impact on Employment. J Digit Econ 2:190\u0026ndash;203\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJamali J, Ayatollahi S, Jafari P (2018) The effect of cross-loading on measurement equivalence of psychometric multidimensional questionnaires in Mimic Model: A simulation study. Materia Socio Med 30:121\u0026ndash;126\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJay J et al (2020) Neighbourhood income and physical distancing during the COVID-19 pandemic in the United States. Nat Hum Behav 4:1294\u0026ndash;1302\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaiser HF, Rice J (1974) Little jiffy, Mark IV. Educ Psychol Meas 34:111\u0026ndash;117\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaiser HF (1960) The application of electronic computers to Factor Analysis. Educ Psychol Meas 20:141\u0026ndash;151\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaiser HF (1958) The Varimax criterion for analytic rotation in factor analysis. Psychometrika 23:187\u0026ndash;200\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKline RB (2023) Principles and practice of structural equation modeling. Guilford\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKniffin KM et al (2021) Covid-19 and the workplace: Implications, issues, and insights for future research and action. Am Psychol 76:63\u0026ndash;77\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi H, Wei YD (2023) Covid-19, cities and inequality. Appl Geogr 160:103059\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarmot M, Allen J (2020) Covid-19: Exposing and amplifying inequalities. J Epidemiol Commun Health 74:681\u0026ndash;682\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMenard SW (2002) Longitudinal research. SAGE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMichaels D, Spieler EA, Wagner G (2024) R. US workers during the COVID-19 pandemic: Uneven risks, inadequate protections, and predictable consequences. BMJ 384\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMid-year population by district and sex (2024) \u003cem\u003eScribd\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.scribd.com/document/784118217/Mid-year-population-by-district-and-sex-2024\u003c/span\u003e\u003cspan address=\"https://www.scribd.com/document/784118217/Mid-year-population-by-district-and-sex-2024\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 3rd November 2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNg A, Jordan M, Weiss Y (2002) On Spectral Clustering: Analysis and an algorithm. Adv Neural Inf Process Syst 14\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNisa CF et al (2021) Lives versus livelihoods? perceived economic risk has a stronger association with support for covid-19 preventive measures than perceived health risk. Sci Rep 11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOnyeaka H et al (2021) Covid-19 pandemic: A review of the global lockdown and its far-reaching effects. Sci Prog 104\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePan Y et al (2020) Quantifying human mobility behaviour changes during the COVID-19 outbreak in the United States. Sci Rep 10\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePangallo M et al (2023) The unequal effects of the health\u0026ndash;economy trade-off during the COVID-19 pandemic. Nat Hum Behav 8:264\u0026ndash;275\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetherick A et al (2021) A worldwide assessment of changes in adherence to COVID-19 protective behaviours and hypothesized pandemic fatigue. Nat Hum Behav 5:1145\u0026ndash;1160\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePett MA, Lackey, Nancy R, Sullivan JJ (2012) Making sense of factor analysis: The use of factor analysis for instrument development in Health Care Research. SAGE\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePopulation density (people per sq (2025) km of land area) - South Asia. \u003cem\u003eWorld Bank Open Data\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://data.worldbank.org/indicator/EN.POP.DNST?locations=8S\u003c/span\u003e\u003cspan address=\"https://data.worldbank.org/indicator/EN.POP.DNST?locations=8S\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 2nd November\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantana C et al (2023) Covid-19 is linked to changes in the time\u0026ndash;space dimension of human mobility. Nat Hum Behav 7:1729\u0026ndash;1739\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchermelleh-Engel K, Moosbrugger H, M\u0026uuml;ller H (2003) Evaluating the Fit of Structural Equation Models: Tests of Significance and Descriptive Goodness-of-Fit Measures. Methods Psychol Res Online 8:23\u0026ndash;74\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSenarath N et al (2024) Family income does not necessarily lead to a holistic e-learning experience during COVID-19: A study in Sri Lanka. Humanit Social Sci Commun 11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimon HA (1955) A behavioral model of rational choice. Q J Econ 69:99\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u003cem\u003eSri Lanka Socio Economic Data 2021\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cbsl.gov.lk/sites/default/files/cbslweb_documents/publications/otherpub/publication_sri_lanka_socio_economic_data_folder_2021_e.pdf\u003c/span\u003e\u003cspan address=\"https://www.cbsl.gov.lk/sites/default/files/cbslweb_documents/publications/otherpub/publication_sri_lanka_socio_economic_data_folder_2021_e.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 3rd November (2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStone BM (2021) The ethical use of fit indices in structural equation modeling: Recommendations for psychologists. Front Psychol 12\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTeleworking in the COVID-19 pandemic: Trends and prospects. OECD Policy Responses to Coronavirus (COVID-19) (2021) \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1787/72a416b6-en\u003c/span\u003e\u003cspan address=\"10.1787/72a416b6-en\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe K-shaped recovery (2025) Examining the diverging fortunes of workers in the recovery from the COVID-19 pandemic using business and household survey microdata. \u003cem\u003eU.S. Bureau of Labor Statistics\u003c/em\u003e Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.bls.gov/osmr/research-papers/2021/ec210020.htm\u003c/span\u003e\u003cspan address=\"https://www.bls.gov/osmr/research-papers/2021/ec210020.htm\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 2nd November\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThilakasiri I et al (2025) The influence of social interactions in mitigating psychological distress during the COVID\u0026thinsp;\u0026ndash;\u0026thinsp;19 pandemic: A study in Sri Lanka. Front Psychol 15\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuominen LS, Helle S, Helanter\u0026auml; H et al (2022) Structural equation modeling reveals decoupling of ecological and self-perceived outcomes in a garden box social-ecological system. Sci Rep 12:6425. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-022-10178-z\u003c/span\u003e\u003cspan address=\"10.1038/s41598-022-10178-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUnited Nations Development Programme (UNDP) (2023) \u0026amp; Oxford Poverty and Human Development Initiative (OPHI). Understanding Multidimensional Vulnerabilities: Impact on People of Sri Lanka. UNDP Sri Lanka, Colombo\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Deursen AJ (2020) Digital inequality during a pandemic: Quantitative study of differences in COVID-19\u0026ndash;related internet uses and outcomes among the general population. J Med Internet Res 22\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan de Weijer MP et al (2022) Genetic and environmental influences on quality of life: The covid-19 pandemic as a natural experiment. Genes Brain Behav 21\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang J et al (2022) Global evidence of expressed sentiment alterations during the COVID-19 pandemic. Nat Hum Behav 6:349\u0026ndash;358\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhite H (1980) A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48:817\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Bank (2023) Poverty \u0026amp; Equity Brief: Sri Lanka \u0026mdash; South Asia (April 2023). World Bank Group, Washington DC\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Bank (2020) South Asia Economic Focus, Fall 2020: Beaten or Broken? Informality and COVID-19. World Bank, Washington, DC. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://hdl.handle.net/10986/34517\u003c/span\u003e\u003cspan address=\"https://hdl.handle.net/10986/34517\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Bank. The COVID\u0026ndash;19 Impact on Livelihoods and Poverty in Sri Lanka: A Background Note, World Bank Group (2020) Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://documents1.worldbank.org/curated/en/724391618996958268/pdf/Background-Note-to-Sri-Lanka.pdf\u003c/span\u003e\u003cspan address=\"https://documents1.worldbank.org/curated/en/724391618996958268/pdf/Background-Note-to-Sri-Lanka.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Economic Outlook A (2021) : Managing divergent recoveries. \u003cem\u003eIMF\u003c/em\u003e (2021).Available at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.imf.org/en/Publications/WEO/Issues/2021/03/23/world-economic-outlook-april-2021\u003c/span\u003e\u003cspan address=\"https://www.imf.org/en/Publications/WEO/Issues/2021/03/23/world-economic-outlook-april-2021\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (Accessed: 1st November 2025)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXi H et al (2023) Quantifying the impact of covid-19 on travel behavior in different socio-economic segments. Transp Policy 136:98\u0026ndash;112\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu F et al (2025) Using human mobility data to quantify experienced urban inequalities. Nat Hum Behav 9:654\u0026ndash;664\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYabe T, Garc\u0026iacute;a Bulle Bueno B, Frank MR, Pentland A, Moro E (2024) Behaviour-based dependency networks between places Shape Urban Economic Resilience. Nat Hum Behav 9:496\u0026ndash;506\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan K-H, Bentler PM (2000) 5. Three likelihood-based methods for mean and covariance structure analysis with Nonnormal Missing Data. Sociol Methodol 30:165\u0026ndash;200\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Table 1","content":"\u003cp\u003eTable 1 is available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"humanities-and-social-sciences-communications","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"palcomms","sideBox":"Learn more about [Humanities \u0026 Social Sciences Communications](http://www.nature.com/palcomms/)","snPcode":"41599","submissionUrl":"https://submission.springernature.com/new-submission/41599/3","title":"Humanities and Social Sciences Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8482833/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8482833/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMobility reflects collective human behaviour, revealing how people interact with their physical and social environments. The COVID-19 pandemic provided a unique opportunity to identify the hidden behavioural and structural properties of society through its mobility responses, with its unprecedented impulse. Data-driven modelling was applied to a nationwide survey dataset collected from Sri Lanka in 2021, to identify 4 distinct archetypes, namely, Privileged Adaptors, Cautious Traditionalists, Stable Minimalists, and Resilient Rebounders, showing diverse adaptations across 13 key factors. These inductively derived clusters from their crisis-driven mobility adaptations subsequently revealed distinct ethnic, educational, and socioeconomic patterns, showing that human behaviour and socioeconomic status are linked not linearly but through structurally conditioned, nonlinear trajectories of risk absorption. These trajectories are governed by the interplay of substitution capacity, exposure necessity, and support infrastructures of the archetypes, offering a generalizable framework for understanding human behavioural inequality in a global disruption.\u003c/p\u003e","manuscriptTitle":"Structural Divides Shape the Nonlinear Nature of Human Mobility in COVID-19","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-06 10:03:27","doi":"10.21203/rs.3.rs-8482833/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-01-13T14:29:19+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-13T12:57:59+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-03T08:05:45+00:00","index":"","fulltext":""},{"type":"submitted","content":"Humanities and Social Sciences Communications","date":"2025-12-30T15:36:47+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"humanities-and-social-sciences-communications","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"palcomms","sideBox":"Learn more about [Humanities \u0026 Social Sciences Communications](http://www.nature.com/palcomms/)","snPcode":"41599","submissionUrl":"https://submission.springernature.com/new-submission/41599/3","title":"Humanities and Social Sciences Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"164a606e-2ac1-4483-ad88-efd20881c596","owner":[],"postedDate":"January 6th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":60513616,"name":"Humanities/Complex networks"},{"id":60513617,"name":"Social science/Complex networks"},{"id":60513618,"name":"Physical sciences/Mathematics and computing"},{"id":60513619,"name":"Physical sciences/Physics"}],"tags":[],"updatedAt":"2026-01-21T23:08:32+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-06 10:03:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8482833","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8482833","identity":"rs-8482833","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}