Landscape Complexity Shapes the Role of Network Density and Diversity in Collective Adaptation Under Disruption

Landscape Complexity Shapes the Role of Network Density and Diversity in Collective Adaptation Under Disruption | 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 Landscape Complexity Shapes the Role of Network Density and Diversity in Collective Adaptation Under Disruption Ross Tieman, Robert Ackland, Katherine Daniell, Steven J. Lade This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8115470/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 11 You are reading this latest preprint version Abstract The polycrisis presents a future characterized by multiple co-occurring crises, demanding human collectives capable of navigating complex challenges simultaneously. This requires understanding how human collectives should be structured to improve their capacity to find and maintain effective solutions despite ongoing and frequent disruptions. We model collective problem-solving using an NK landscape impacted by intermittent disruptions to examine how network connectivity and agent diversity influence performance under ongoing disruptions. We measure the capacity for different collective structures to find and maintain high-performing solutions through various disruption regimes (frequency of disruption, impact size, impact distribution). Results reveal that disruptions degrade collective cumulative performance in simple landscapes (K = 0) but enhance it in complex ones (K = 7). In complex landscapes, moderate connectivity (d ≈ 0.2–0.32) and high diversity maximize cumulative performance, with disruptions amplifying these benefits. Additionally, we observe that impact distribution across agent groups affects cumulative performance, with skewed distributions. Our analysis demonstrates that in complex landscapes, disruptions can actually improve collective problem-solving capacity, but only with appropriate collective structures. This suggests that effective collective design requires matching structure to both problem complexity and disruption environment rather than applying universal principles. Physical sciences/Mathematics and computing Physical sciences/Physics collective intelligence collective adaptation NK landscape resilience social learning agent-based model networks Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1 Introduction Climate change, the polycrisis (Lawrence et al., 2024), and other sources of major disruptions present human collectives with unpredictable challenges requiring simultaneous adaptation across multiple domains. For example, agricultural collectives facing increased climate volatility must adapt crop selection, supply chain management, and resource allocation while maintaining productivity (Folke et al., 2010 ). The potential for critical transitions in such systems, which can cause abrupt shifts in dynamics requiring rapid adaptation (Scheffer et al., 2009 ), raises questions about how collectives should be structured to maintain performance in increasingly turbulent environments. Existing research primarily examines collective problem-solving in stable environments. Network structure shapes information diffusion and collective intelligence, with dense connectivity accelerating social learning but potentially promoting premature convergence on suboptimal solutions (Burt, 2004 ; Centola, 2022 ; Derex & Boyd, 2016; Lazer & Friedman, 2007 ; Mason & Watts, 2012). Organisational learning theory addresses the exploration-exploitation trade-off, where collectives balance refining current solutions against searching for alternatives (March, 1991 ). However, these frameworks assume static problem landscapes and require extension to account for environmental disruptions that can render existing solutions obsolete. A critical gap remains in understanding how network structure and agent diversity jointly influence collective performance under ongoing disruptions, and whether these relationships depend on problem complexity. While early work examined how specific structures respond to single perturbations (Siggelkow & Rivkin, 2005 ), and recent studies show network structure moderates diversity effects in static environments (Baumann et al., 2024 ), we lack systematic analysis of ongoing disruption regimes. The Collective Adaptation framework (Galesic et al., 2023 ) provides conceptual foundations for studying dynamic problem landscapes, but computational implementations testing specific mechanisms remain limited. This question has broad relevance across complexity science, from biological regulatory systems to human governance (Gao et al., 2016 ; Levin, 1998; Ostrom, 2010 ). To answer the question of how to structure collectives to maintain performance under ongoing disruptions, we adopt a socio-ecological resilience perspective because of its focus on structure-environment interactions. Unlike approaches that seek universal structural principles, this perspective recognizes that effective social structures must match their environmental context (Bodin, 2017 ; Folke, 2006 ), with established resilience principles specifically highlighting the importance of maintaining diversity and managing connectivity (Biggs et al., 2012 ). By foregrounding the interaction between disruption regimes and collective structures in shaping performance, the socio-ecological resilience perspective aligns with core insights from the collective adaptation framework concerning the tuning of collective properties to given problem structures (Galesic et al., 2023 ). While motivated by human collectives responding to increasingly turbulent environments, the approach employed is sufficiently general to capture collective learning processes wherever networked agents search among interdependent options. We extend Baumann et al.'s ( 2024 ) NK landscape model of collective learning by incorporating stochastic disruptions to problem landscapes. This extension enables systematic examination of how network connectivity and agent diversity jointly determine collective performance under varying disruption regimes. Disruption regimes are defined by three characteristics in our analysis: frequency, severity, and distribution of impact across agent groups. By examining these dynamics across both simple and complex problem landscapes, where problem complexity determines the degree of interdependence between decisions, we can test whether optimal collective structures depend on interactions between problem complexity, disruption regime characteristics, and collective composition (network structure, agent diversity). We systematically examine how network connectivity and agent diversity influence collective performance across simple and complex problem landscapes under varying disruption regimes. We characterise disruption regimes by frequency, severity, and distribution of impacts across agent groups, enabling analysis of how these factors interact with collective structure and problem complexity. Our approach extends collective intelligence research on network effects in learning (Centola, 2022 ; Baumann et al., 2024 ) by examining how these effects persist or change under environmental disruption, while providing computational evidence for socio-ecological resilience principles regarding connectivity and diversity (Biggs et al., 2012 ) and polycentric governance structures (Ostrom, 2010 ). 2 Methods 2.1 Model Framework We extend the collective learning model of Baumann et al. ( 2024 ) by introducing stochastic disruptions that reconfigure problem landscapes. Our framework models collective adaptation as a collective search problem where N interacting agents search a space of solutions over some number of timesteps T (Fig. 1 .). This solution space could represent combinations of decisions organisations must make or ways to arrange system components. Agents search the space through individual exploration and social learning from connected neighbours, with high-performing solutions spreading through the collective via network ties. The quality of a solution (S) is quantified by a payoff function P(S) . Disruptions modify these payoff functions, changing the value of solutions and the overall performance of the group. Exposing collectives to disruption regimes creates ongoing adaptive challenges that allows systematic investigation of collective properties (network structure, agent diversity) that can facilitate performance under disruption. 2.2 NK Landscape Structure The fitness landscape is represented using the NK model (Kauffman & Levin, 1987 ), which provides a tunably rugged landscape suitable for investigating adaptive processes. The NK model is commonly deployed to investigate social phenomena such as innovation processes and decision making and social learning in collectives such as human organisations (Levinthal, 1997 ). The NK model is defined by two parameters: N NK , the number of binary decisions, and K NK , the interdependence (epistatic interactions) of decisions. Each solution S is a vector of N NK = 15 binary elements b i ∈ {0,1}, and the payoff function is computed as: Where I i represents the set of K NK interacting elements and f are uniformly distributed contribution functions. For simple landscapes K NK = 0, meaning I i is empty and each decision contributes interpedently. For complex landscapes K NK = 7, meaning each decision's contribution depends on 7 randomly assigned interacting decisions, creating rugged fitness landscapes with multiple local optima. Following Baumann et al. ( 2024 ), payoffs are normalised by P max and raised to power 8 to create sparse high-quality solutions. K NK = 7 represents approximately half of all decisions being interdependent, which provides a threshold for meaningful complexity while also balancing computational requirements which grow as K NK approaches N NK (Baumann et al., 2024 ). 2.3 Network Structure Collective problem-solving processes are known to be strongly influenced by communication networks (Lazer, D. & Friedman, 2007 ; Barkoczi, D. & Galesic, M. 2016; Centola, 2022 ). To capture this effect agents are embedded in undirected Erdős–Rényi (ER) random networks (Erdős, P. & Rényi, A., 1959) with varying densities following the approach of Baumann et al. ( 2024 ) (see Table 1 ). ER networks provide a baseline topology with homogeneous structural properties (Newman M. E. J., 2003 ; Baumann et al., 2024 ), allowing density effects to be examined independently of confounding topological features such as clustering or degree heterogeneity. The model implementation allows straightforward selection of alternative network topologies (ER, small-world, Barabási-Albert, etc.) via configuration parameters, enabling future work to systematically examine alternative structural assumptions. Networks are generated at initialisation before agents are randomly assigned to network positions. 2.4 Agent Diversity The n = 100 agents are partitioned into G groups of equal size (see Table 1 for group configurations). Each group g has a distinct payoff function P g ( S ) generated by unique contribution functions f g (b i | I i ) ~ Uniform(0,1). For the same solution S , agents in different groups extract different payoffs: P α ( S ) ≠ P β ( S ) for α ≠ β. This captures how different skills, perspectives, or functional roles influence performance evaluation in collective systems. Agents are evenly distributed across groups, and network attachment probability is independent of group membership.2.5 Agent Behaviour At each timestep t ∈ {1,...,T}, agents synchronously update through a two-stage process. First, agents attempt social learning by adopting the best neighbour solution if it improves their group-specific payoff (referred to below as social learning events). If no improvement is found through social learning, agents attempt innovation by randomly flipping one bit in their binary solution vector and adopting the new solution if it yields improvement (innovation events). If neither social learning nor innovation yields improvement, agents retain their current solution. This implements satisficing behaviour balancing exploitation and exploration. 2.5 Disruption Mechanism Disruption regimes are modelled through stochastic perturbations to fitness landscapes, characterized by three parameters with specific values provided in Table 1 . Disruption frequency ω determines the probability at each timestep that a disruption occurs, following a Bernoulli process. Moderate impact scenarios reflect typical historical disruptions, while severe scenarios represent the amplified disruptions projected under polycrisis conditions (Lawrence M., et al. 2024)(see Table 1 ). The shape parameters α and β for the Beta distribution are determined by minimising the squared error between target percentiles and those produced by candidate parameter values, using Nelder-Mead optimisation. Beta distributions provide flexible bounded distributions on (0,1) suitable for modeling proportional impacts and are widely used in risk analysis such as earthquake impact modelling (Vose, 2008 ; Lallemant & Kiremidjian, 2015 ). The heavy-tailed parameterisations capture patterns where most disruptions have moderate impacts while severe impacts occur with non-negligible probability (Clauset et al., 2009 ; Helbing 2013 ) (See Supplementary Figure S1 -S3 for disruption distributions and characteristics). The distribution of impact magnitude across groups follows an exponential decay τ g = τ·e −λ(g−1) , where λ determines the degree of inequality in impact distribution across groups. When λ = 0, all groups experience identical impact. When λ > 0, impacts follow an exponential gradient creating substantial differences between most and least affected groups. This captures how distinct functional units of the collective (e.g. organisational departments) are likely to have differential exposure to any given disruption. For each disruption, vulnerability rankings are randomly permuted across groups. Total disruption magnitude is conserved for each disruption event. When disruption occurs, each group's contribution functions are modified through linear interpolation following Siggelkow and Rivkin ( 2005 ): where u g,i ~ Uniform (0,1) is independently sampled for each group and contribution function. Each of the G groups has its contribution functions updated according to its group-specific impact magnitude τ g . For K NK = 0, each group has 30 contribution values corresponding to N NK positions and 2 states per position. For K NK = 7, each group has 3,840 contribution values corresponding to N NK positions and 2^(K NK +1) sub combinations per position. The interaction matrix M defines the epistatic relationships between positions, the specific pattern of interdependencies where each position's contribution to the payoff depends not only on its own state but also on the states of K other positions. M is preserved across disruptions, reflecting that perturbations typically alter performance within existing structural constraints rather than fundamentally rewiring interdependencies (Siggelkow & Rivkin, 2005 ; Scheffer et al., 2012). Frequency and magnitude are modelled as independent factors to enable systematic factorial exploration of how collective configurations perform under varying disruption characteristics. This represents a simplification as many natural hazards and disruption events such as blackouts exhibit frequency-magnitude relationships (Newman, M.E.J., 2005 ; Carreras, B.A., 2004; Becerra O., et al., 2006), though we believe is suitable to enable streamlined investigation of pure frequency impact size effects. 2.6 Performance Measurement and Experimental Design Collective performance at time t is measured as median payoff across all N = 100 agents, P ~ (t) = median {P g (S i (t)): i = 1,...,N}. Cumulative performance Π=∑ T t=1 P ~ (t) aggregates sustained functionality, capturing resilience as persistence of function under disruption (Folke, 2006 ). We employ a full factorial design crossing network density, agent diversity, landscape complexity, and disruption regimes (Table 1 ). Network density and agent group values use denser sampling at lower values where theoretical predictions and prior empirical findings (Lazer & Friedman, 2007 ; Baumann et al., 2024 ) indicate that marginal effects are expected to be larger. This parameter spacing focuses simulations on areas where qualitatively important transitions occur i.e. from homogeneous to diverse collectives as the first few groups are added (G = 1–3 groups) and sparse to intermediate connectivity (d = 0.04–0.12) while limiting the factorial to a 7×7 configuration for computational tractability. We run simulations for all diversity and density parameter combinations with no disruptions which functions as a control to compare impacts of disruptions against. Each parameter combination is replicated 83 times, a total of 54,544 simulations. This sample size was the result of successive sampling rounds (3,30,50 replicates) to ensure sufficient statistical power for main effects while balancing computational requirements. Median cumulative performance across replicates for a given parameter set is reported with bootstrap confidence intervals (5,000 samples) used to quantify uncertainty in median estimates. Wilcoxon rank-sum tests are used to assess cross-landscape differences. Table 1 Simulation parameters and factor choices for disruption regimes. Parameter Values Timesteps (t) 400 Agents (n) 100 Network Density (d) 0.04, 0.08, 0.12, 0.20, 0.32, 0.52, 0.76 Agent Groups (G) 1, 2, 3, 5, 8, 13, 21 Landscape Complexity (K) 0, 7 Disruption Regimes Low Factor High Factor Frequency (ω) (infrequent, frequent) 1/100 1/25 Impact Size (moderate, severe) 5th % = 0.001 95th % = 0.5 5th% = 0.001 95th % = 0.65 Impact Inequality (λ) (uniform, skewed) 0 0.5 3 Results We examine how network connectivity and agent diversity influence collective performance under disruption across simple (K NK = 0) and complex (K NK = 7) landscapes. Our core finding is that landscape complexity determines whether disruptions degrade or enhance collective performance, with simple and complex landscapes exhibiting markedly different responses to identical disruption regimes. We first present this cross-landscape comparison, then detail the mechanisms and configuration effects for each landscape type. All cumulative performance values are reported as medians with 95% bootstrap confidence intervals unless otherwise noted. Disruptions trigger immediate performance changes that vary systematically by landscape complexity (Fig. 2 ). In simple landscapes, disruptions consistently reduce performance by interfering with convergence toward the global optimum. Notably, in complex landscapes, disruptions occasionally produce immediate performance increases when landscape changes allow agents to escape local optima where they were previously trapped. This "creative destruction" phenomenon is well-documented in search processes and complex adaptive systems (Kauffman & Weinberger, 1989; Allen et al., 2014 ). 3.1 Cross-Landscape Analysis: Complexity Moderates Creative Destruction Comparative analysis reveals disruptions produce opposite performance effects depending on landscape complexity (Fig. 3 ). Simple landscapes (K NK = 0) show small negative reductions in median cumulative performance across all disruption regimes, with median performance declining between − 0.06% and − 0.75% relative to the control (no disruptions). In contrast, complex landscapes (K NK = 7) show substantial positive effects across all disruption regimes, with performance increasing between 5.62% and 16.79% relative to control. This occurs under identical disruption mechanisms and is highly significant (Wilcoxon rank-sum test, W = 0.0, p = 0.0078, comparing disruption effects between landscapes across eight regime combinations). Impact inequality (skew) effects only matter in complex landscapes for severe disruptions (Fig. 3 ). In simple landscapes, uniform versus skewed distribution produces nearly identical outcomes within each disruption regime. In complex landscapes, skew appears to only become relevant for severe disruptions irrelevant of frequency. The impact is most pronounced under frequent-severe conditions, uniformly distributed impacts yields a median performance effect of 16.79% (95% CI: [15.67%, 17.89%]) compared to 7.21% (95% CI: [5.71%, 8.74%]) under skewed impacts, a median difference of 9.58 percentage points. The median cumulative performance ranges conditions differ dramatically between simple and complex landscapes. Median cumulative performance in simple landscapes span 0.036 units across all disruption conditions ranging from 0.958 (G = 1, d = 0.08, [frequent, severe, skewed]) to 0.994 (various disruption regimes) (Supplementary Figure S4). In complex landscapes the median cumulative performance spans 0.435 units, ranging from 0.325 (G = 13, d = 0.04, [infrequent, severe, skewed]) to 0.760 (G = 21, d = 0.052, [frequent, severe, skewed]), a ten-fold difference (Supplementary Figure S5). This reflects that simple landscapes operate near ceiling performance with limited room for variation, while complex landscapes pose greater collective learning challenges where collective configuration choices meaningfully impact outcomes. Performance difference heatmaps (disruption minus control, Supplementary Figures S5-S6) reveal distinct mechanisms that increase resilience to disruptions. In simple landscapes, disruption effects concentrate on low-diversity configurations regardless of density: at G = 1–2, severe disruptions reduce performance by 3–4 percentage points, but by G = 5, disruption effects become negligible across all density levels (Figure S5). This demonstrates that diversity, not density, provides most of the resilience. In complex landscapes, the patterns are much more complex. Positive effects appear in diverse configurations including, both high-density-low-diversity (G = 1, d = 0.76) and low-density-high-diversity (G = 21, d = 0.04) combinations. Notably, high diversity (G = 21) shows consistently positive or neutral effects across all density levels and disruption regimes, indicating robust benefits of maintaining diverse perspectives. The heterogeneity in difference patterns across regimes suggests that disruption-configuration interactions warrant further investigation to identify specific mechanisms underlying performance gains. These patterns reflect fundamentally different adaptation challenges. In simple landscapes, disruptions interfere with convergence toward a single global optimum, though high baseline performance limits absolute impacts. In complex landscapes, disruptions enable escape from local maxima, facilitating exploration of superior solution regions. Uniform distribution amplifies this creative destruction effect by preserving diverse perspectives needed to navigate rugged landscapes, while concentrated impacts eliminate specific viewpoints from the collective search process. 3.2 Simple Landscapes (K = 0) Network density and agent diversity serve distinct functions in simple landscapes (Fig. 4 ). Under control conditions, performance peaks 0.994 at high density and low diversity (Fig. 4 A). Under disruptions, collectives with greater diversity (G > 3) and density (d > 0.2) performing better. Density appears to drive baseline performance through rapid information diffusion, exhibiting a logarithmic-like increase with steep gains below d = 0.2 and a plateau beyond (Fig. 3 C). Diversity provides minimal baseline benefits beyond G = 2–3. This pattern replicates Baumann et al. ( 2024 ) and reflects that dense networks enable rapid diffusion of the single global optimum in simple landscapes (Centola, 2022 ), while diversity provides minimal additional benefit navigating the smooth landscape. The diversity-performance relationship shows threshold behaviour: in response to disruptions, performance jumps from G = 1 up to G = 3, after which additional diversity provides less value. This is most prominent under frequent-severe disruptions where performance jumps from a median ≈ 0.964 (G = 1) to ≈ 0.991 (G = 3) then saturates through G = 21 (Fig. 3 D). This threshold reflects the transition from complete homogeneity, where all agents seek identical solutions and disruptions affect everyone similarly, to diversity where additional payoff functions provide redundancy. Critically, diversity ≥ 5 largely eliminates disruption effects entirely. Performance difference heatmaps (Supplementary Figure S6) show that severe disruptions reduce performance by ~ 1.0–3.3 percentage points at G = 1–2 but produce near-zero effects at G ≥ 5, regardless of network density. Additional diversity beyond G = 5 provides neither baseline gains nor resilience benefits, as sufficient redundancy has been achieved. 3.3 Complex Landscapes (K = 7) The complex landscape requires fundamentally different collective properties to achieve good cumulative performance compared to simple landscapes (Fig. 5 .). Under control conditions, performance peaks 0.577 at moderate density and high diversity (Fig. 5 A). Under disruption, performance increases substantially across most configurations, with roughly the same combinations favoured as the control, with a cumulative performance peak of 0.677 (Fig. 5 B). The density-performance relationship exhibits an inverted-U with peaks around d = 0.2–0.32, declining at both low (d 0.5) density (Fig. 5 C). This non-monotonic pattern persists across all disruption regimes and suggests that moderate connectivity balances information sharing benefits against costs of excessive coupling. Sparse networks limit collective learning, while dense networks may promote premature convergence or rapid spread of solutions reducing effective diversity limiting future innovation (Levinthal 1997 ; Lazer & Friedman, 2007 ; Centola, 2022 ). This type of convergence is also likely to reduce resilience and adaptability to disruptions due to limiting diversity of unique solutions leading to greater effective heterogeneity of agents known to increase potential for critical transitions (Scheffer 2012). This inverted-U persists across all disruption regimes and is also present in the control, with uniform distribution outperforming skewed at all density levels, except infrequent-mild which is roughly equivalent. Diversity increase performance gradually through G = 8–13 (Fig. 5 D), contrasting with simple landscapes' sharp threshold and plateau (Fig. 4 D). Across the regimes shown, performance continues increasing through the highest diversity level examined (G = 21), with different disruption regimes achieving different absolute performance levels. Performance difference heatmaps (Supplementary Figure S7) show that disruptions are broadly beneficial across complex landscapes, though patterns are heterogeneous and complex. High diversity (G = 21) shows consistently positive or neutral effects across all density levels and disruption regimes. The varied spatial patterns of positive effects across different regime heatmaps suggest that multiple configurations may achieve performance gains through different mechanisms, warranting further investigation of potential mechanism. 4 Discussion Our analysis reveals that landscape complexity fundamentally determines whether disruption enhances or degrades collective performance. Simple landscapes (K = 0) demonstrate universal performance degradation under disruption (performance effects ranging from − 0.06% to -0.78%), while complex landscapes (K = 7) show universal performance enhancement under disruption (median effects ranging from 5.78% to 16.79%) (Fig. 3 ). The difference in effects between landscapes is significant (Wilcoxon W = 0.0, p < 0.001) and challenges assumptions about disruption effects suggesting that the relationship between disruption and collective performance is highly dependent on the environment collectives must navigate. 4.1 Extending Collective Intelligence Frameworks The reversal between simple and complex landscapes extends existing collective intelligence frameworks in important ways. While previous work has shown that network structure moderates diversity effects in static environments (Baumann et al., 2024 ), our results demonstrate that disruptions fundamentally alter these relationships. In simple landscapes, disruptions interfere with convergence toward global optima, confirming that volatility impedes performance when solutions are straightforward. However, in complex landscapes, disruptions facilitate escape from local optima, functioning similarly to simulated annealing by facilitating exploration of solution spaces. This finding aligns with theoretical predictions for rugged fitness landscapes (Kauffman & Weinberger, 1989; Weinberger, 1991 ) and provides computational support for the collective adaptation framework's proposition that collectives must match their structure to problem characteristics (Galesic et al., 2023 ). The creative destruction mechanism we observe, where disruptions enable performance gains in complex environments, has been theorised in socio-ecological systems (Allen et al., 2014 ) but with limited observations in computational models and in collective learning contexts. 4.2 Network Structure and Resilience Mechanism For complex landscapes, network connectivity exhibits an inverted-U relationship with performance (Fig. 5 C), aligning with patterns observed across diverse complex systems from supply chains (Chen & Wen, 2023 ) to infrastructure networks (Brummitt et al., 2012 ). Sparse connectivity limits information diffusion necessary for collective learning, while excessive connectivity promotes premature convergence or rapid spread of suboptimal solutions. Critically, this inverted-U persists under disruption, demonstrating robustness of the connectivity optimum across environmental regimes and providing empirical grounding for theoretical predictions about connectivity thresholds in complex adaptive systems (Gao et al., 2016 ). The difference between disrupted and control conditions (Supplementary Figures S6-S7) illuminates distinct resilience mechanisms across landscape types. In simple landscapes, network density optimizes baseline performance by enabling rapid diffusion of the single optimal solution, but diversity provides resilience by maintaining redundant solution approaches that buffer against disruption. Once sufficient diversity is achieved (G ≥ 5), additional diversity provides neither baseline gains nor resilience benefits, as adequate coverage of the solution space has been attained. In complex landscapes, both connectivity and diversity contribute to resilience, with moderate connectivity (avoiding both under-connection and over-coupling) and high diversity (maintaining multiple perspectives for navigating local optima) required for maximal disruption-enabled performance gains. The role of distributional equity emerges clearly in complex landscape difference maps, where uniform distribution amplifies positive effects throughout the parameter space. 4.3 Distribution of Impacts and Collective Performance The 4–7 percentage point performance advantage of severe-uniform over severe-skewed impacts indicates that equality of impacts matters for maintaining collective performance (Fig. 3 .). When certain groups face disproportionately severe impacts it is possible for skewed distributions to become τg = 1 for certain groups (see Supplementary Information figures S 2–3). This results in a completely new contribution function being redrawn from the uniform distribution which effectively randomizes the position on the landscape of agents in those groups. In the severe skewed disruptions this likely results in agents being unable to contribute meaningfully to collective learning, functionally reducing the diversity that is critical for performance in complex landscapes (Baumann et al., 2024 ; Centola, 2022 ; Hong & Page, 2004). This implies that minimising potential (or at least managing) potential for highly unequal impacts in turbulent complex environments can have benefits for collective performance and adaptability. Future work should investigate threshold effects and how network homophily might amplify inequality effects by limiting information flow between differentially affected groups. Practically, this implies that minimising inequality of impacts in turbulent complex environments, or at least managing them below thresholds that completely eliminate groups' knowledge, is beneficial for collective performance. This points toward future work investigating group-level impacts to determine threshold effects and more mechanistic analysis of how agent locations in social learning networks contribute. The current model randomly distributes agents across network positions regardless of group membership, but real-world collectives likely exhibit homophily where similar agents preferentially connect. For instance, complexity science researchers are more likely to learn from other complexity scientists than from individuals in different fields, which could amplify inequality effects by limiting information flow from less-affected to severely-affected groups. 4.4 Connections to Socio-Ecological Resilience Our results provide quantitative support for established resilience principles while revealing how they apply differently across problem complexities. This aligns with socio-ecological resilience findings which show connectivity and diversity play a central role in system resilience that is context dependent (Biggs et al., 2015). Practically this implies the need for more context aware management approaches. The pattern connects to polycentric governance principles, which emphasize distributed authority and diverse decision-making centres for enhancing system resilience (Ostrom, 2010 ). Polycentric governance increases diversity and autonomy across levels, fostering innovation and flexibility beneficial under changing circumstances (Singh & Kumar, 2024). Our findings demonstrate that diversity consistently enhances performance under disruption across landscape types, while connectivity requirements vary with environmental context. Collectives navigating complex environments require fundamentally different management approaches than those in simpler environments. Maintained diversity and moderate connectivity yield highest performance under disruption, aligning with polycentric governance principles. 4.5 Multiple Modes of Adaptation The optimal configuration for collectives to adapt fundamentally depends on both landscape complexity and the characteristics of environmental disruption. Looking at differences in cumulative performance compared to controls for complex landscapes across different regimes, shows that both low diversity, high network density; high network density, high diversity; and high diversity, low network density configurations can all see relatively large (> 15%) and significant increases (Supplementary Fig. 7.). This may be indicative of different dominant modes of adaptation to new landscapes which have different resilience and response characteristics. For instance, rapid high connectivity low diversity collectives might rapidly share good solutions that result from fortunate disruptions that place agents on high performing locations of the landscape. While low network density high diversity groups may be more slowly and independently accruing a greater distribution of high performing solutions, the slower convergence and greater diversity of landscape solutions would then mean that the likelihood of all solutions massively dropping in payoff value is lower. This aligns with existing evidence in resilience and network science of collective intelligence literature (Centola, 2022 ). Further mechanistic analysis of social learning and innovation along with time series analysis of performance drops and recoveries is required to confirm different modes contributing to resilience and recovery which we defer to future work. 4.6 Limitations and Future Directions While our analysis advances understanding of collective adaptation under disruption, several limitations warrant acknowledgment. Our focus on random network structures restricts insights into how specific topologies influence adaptation, and the binary comparison between K = 0 and K = 7 leaves intermediate complexity levels unexplored. Future work examining broader ranges of landscape complexity could reveal phase transitions in the disruption-performance relationship and identify when collectives should shift strategies. Additionally, temporal analysis of performance drops and recoveries could distinguish when rapid information sharing versus sustained exploration provides advantages. The basic NK landscape implementation doesn't capture agents' capacity to influence their environment, missing key characteristics like niche construction observed in real-world systems. Incorporating endogenous landscape dynamics through NKES (Suzuki & Arita, 2005 ) or NKEZ landscapes (Gavetti et al., 2017 ) would allow agents to deliberately modify their problem spaces, similar to policy interventions or ecosystem engineering. Such extensions could reveal how collectives might strategically complexify or simplify their decision landscapes to leverage or avoid disruption effects, potentially generating endogenous disruption dynamics. Given that multiple network-diversity configurations yield performance gains under disruption, investigating adaptive networks represents a crucial next step. Simulations where network connections evolve based on social learning or innovation patterns could reveal how collectives discover appropriate structures for their problem landscapes. This aligns with the collective adaptation framework's emphasis on meta-learning (Galesic et al., 2023 ; Baumann et al., 2024 ) and earlier conceptual work on Open Systems Theory exploring how organisations need to adapt to changing environments of different complexity (Emery & Trist, 1965 ; Emery, 1999 ). Extension of the proposed model through adaptive network approaches (Berner et al., 2023 ; Sayama et al., 2013 ) could allow exploration of these meta-learning and organisational adaptation themes. Understanding when and how collectives should switch between rapid-convergence versus distributed-exploration configurations would provide practical insights for designing resilient decision-making structures. Our results demonstrate that collectives navigating complex environments require fundamentally different structures than those in simple environments. The reversal between landscapes, where disruptions hinder performance in simple problems but enhance it in complex ones, suggests that with appropriate structures, collectives can navigate and even benefit from environmental volatility rather than merely enduring it. The convergence of our findings around moderate connectivity, maintained diversity, and distributional equity aligns with polycentric governance principles while revealing how these principles apply differently across problem complexities. By maintaining or enhancing these key properties in collectives it may be possible to adapt to the disruption laden future entailed by the polycrisis. Declarations Acknowledgements AI was used in the development of manuscript. AI coding tools were used in development of model code: Cursor-small, Claude Sonnet 4.5, Claude Opus 4.1. Generative AI was used improve readability of manuscript text: Claude Sonnet 4.5, Claude Opus 4.1. Funding Declaration S.J.L. This research has been supported by funding from the Australian Government (Australian Research Council Future Fellowship FT200100381 to S.J.L.). K.D. receives funding from the OneBasin CRC. Author Contributions R.J.T. Developed research concept, developed model and analysis approach, coded the model, ran simulations, performed data analysis, interpreted results, created all figure and wrote manuscript. S.J.L. Contributed to research concept developed, assisted in model design and analysis approach, assisted in research framing, contributed to interpretation of results and reviewed the manuscript. R.A. and K.D. Provided input on research concept & modelling approach, contributed to interpretation of results, reviewed manuscript. Code and Data Availability Code for the simulation model is available at: https://github.com/rosstie/caud Competing Interests None of the authors have competing interests. References Allen, C. R., Angeler, D. G., Garmestani, A. S., Gunderson, L. H., & Holling, C. S. (2014). Panarchy: Theory and Application. Ecosystems , 17 (4), 578–589. https://doi.org/10.1007/s10021-013-9744-2 Baumann, F., Czaplicka, A., & Rahwan, I. (2024). Network structure shapes the impact of diversity in collective learning. Scientific Reports , 14 (1), 2491. https://doi.org/10.1038/s41598-024-52837-3 Berner, R., Gross, T., Kuehn, C., Kurths, J., & Yanchuk, S. (2023). Adaptive Dynamical Networks (No. arXiv:2304.05652). arXiv. http://arxiv.org/abs/2304.05652 Bodin, Ö. (2017). Collaborative environmental governance: Achieving collective action in social-ecological systems. Science (New York, N.Y.) , 357 (6352). https://doi.org/10.1126/science.aan1114 Biggs, Reinette, Maja Schlüter, Duan Biggs, et al. 2012. “Toward Principles for Enhancing the Resilience of Ecosystem Services.” Annual Review of Environment and Resources 37 (Volume 37, 2012): 421–48. https://doi.org/10.1146/annurev-environ-051211-123836. Brummitt, C. D., D’Souza, R. M., & Leicht, E. A. (2012). Suppressing cascades of load in interdependent networks. Proceedings of the National Academy of Sciences , 109 (12), E680–E689. https://doi.org/10.1073/pnas.1110586109 Burt, R. S. (2004). Structural Holes and Good Ideas. American Journal of Sociology , 110 (2), 349–399. https://doi.org/10.1086/421787 Carreras, B.A., D.E. Newman, I. Dobson, and A.B. Poole. 2004. “Evidence for Self-Organized Criticality in a Time Series of Electric Power System Blackouts.” IEEE Transactions on Circuits and Systems I: Regular Papers 51 (9): 1733–40. https://doi.org/10.1109/TCSI.2004.834513. Centola, D. (2022). The network science of collective intelligence. Trends in Cognitive Sciences , 26 (11), 923–941. https://doi.org/10.1016/j.tics.2022.08.009 Chen, J., & Wen, H. (2023). The application of complex network theory for resilience improvement of knowledge-intensive supply chains. Operations Management Research , 16 (3), 1140–1161. https://doi.org/10.1007/s12063-023-00365-0 Clauset, Aaron, Cosma Rohilla Shalizi, and M. E. J. Newman. 2009. “Power-Law Distributions in Empirical Data.” SIAM Review 51 (4): 661–703. https://doi.org/10.1137/070710111. Emery, F. E., and E. L. Trist. 1965. “The Causal Texture of Organizational Environments.” Human Relations 18 (1): 21–32. https://doi.org/10.1177/001872676501800103. Emery, Merrelyn. 1999. Searching: The Theory and Practice of Making Cultural Change. John Benjamins. https://doi.org/10.1075/dowi.4. Folke, C. (2006). Resilience: The emergence of a perspective for social–ecological systems analyses. Global Environmental Change , 16 (3), 253–267. https://doi.org/10.1016/j.gloenvcha.2006.04.002 Folke, C., Carpenter, S., Walker, B., Scheffer, M., Chapin, T., & Rockström, J. (2010). Resilience Thinking: Integrating Resilience, Adaptability and Transformability. Ecology and Society , 15 (4). https://doi.org/10.5751/ES-03610-150420 Galesic, M., Barkoczi, D., Berdahl, A. M., Biro, D., Carbone, G., Giannoccaro, I., Goldstone, R. L., Gonzalez, C., Kandler, A., Kao, A. B., Kendal, R., Kline, M., Lee, E., Massari, G. F., Mesoudi, A., Olsson, H., Pescetelli, N., Sloman, S. J., Smaldino, P. E., & Stein, D. L. (2023). Beyond collective intelligence: Collective adaptation. Journal of The Royal Society Interface , 20 (200), 20220736. https://doi.org/10.1098/rsif.2022.0736 Gao, J., Barzel, B., & Barabási, A.-L. (2016). Universal resilience patterns in complex networks. Nature , 530 (7590), 307–312. https://doi.org/10.1038/nature16948 Gavetti, G., Helfat, C. E., & Marengo, L. (2017). Searching, Shaping, and the Quest for Superior Performance. Strategy Science , 2 (3), 194–209. https://doi.org/10.1287/stsc.2017.0036 Helbing, Dirk. 2013. “Globally Networked Risks and How to Respond.” Nature 497 (7447): 51–59. https://doi.org/10.1038/nature12047. Kauffman, S., & Levin, S. (1987). Towards a general theory of adaptive walks on rugged landscapes. Journal of Theoretical Biology , 128 (1), 11–45. https://doi.org/10.1016/S0022-5193(87)80029-2 Lallemant, David, and Anne Kiremidjian. 2015. “A Beta Distribution Model for Characterizing Earthquake Damage State Distribution.” Earthquake Spectra 31 (3): 1337–52. https://doi.org/10.1193/012413EQS013M. Lazer, D., & Friedman, A. (2007). The Network Structure of Exploration and Exploitation. Administrative Science Quarterly , 52 (4), 667–694. https://doi.org/10.2189/asqu.52.4.667 Levinthal, D. A. (1997). Adaptation on Rugged Landscapes. Management Science , 43 (7), 934–950. March, J. G. (1991). Exploration and Exploitation in Organizational Learning. Organization Science , 2 (1), 71–87. Newman, M. E. J. 2003. “The Structure and Function of Complex Networks.” SIAM Review 45 (2): 167–256. https://doi.org/10.1137/S003614450342480. Newman, M. E. J. 2005. “Power Laws, Pareto Distributions and Zipf’s Law.” Contemporary Physics 46 (5): 323–51. https://doi.org/10.1080/00107510500052444. Ostrom, E. (2010). Beyond Markets and States: Polycentric Governance of Complex Economic Systems. American Economic Review , 100 (3), 641–672. https://doi.org/10.1257/aer.100.3.641 Sayama, H., Pestov, I., Schmidt, J., Bush, B. J., Wong, C., Yamanoi, J., & Gross, T. (2013). Modeling complex systems with adaptive networks. Computers & Mathematics with Applications , 65 (10), 1645–1664. https://doi.org/10.1016/j.camwa.2012.12.005 Scheffer, M., Bascompte, J., Brock, W. A., Brovkin, V., Carpenter, S. R., Dakos, V., Held, H., van Nes, E. H., Rietkerk, M., & Sugihara, G. (2009). Early-warning signals for critical transitions. Nature , 461 (7260), 53–59. https://doi.org/10.1038/nature08227 Siggelkow, N., & Rivkin, J. W. (2005). Speed and Search: Designing Organizations for Turbulence and Complexity. Organization Science , 16 (2), 101–122. https://doi.org/10.1287/orsc.1050.0116 Suzuki, R., & Arita, T. (2005). How Niche Construction Can Guide Coevolution. In M. S. Capcarrère, A. A. Freitas, P. J. Bentley, C. G. Johnson, & J. Timmis (Eds.), Advances in Artificial Life (pp. 373–382). Springer. https://doi.org/10.1007/11553090_38 Vose, David. 2008. Risk Analysis : A Quantitative Guide. With Internet Archive. Chichester, England ; Hoboken, NJ : Wiley. http://archive.org/details/riskanalysisquan0000vose. Weinberger, E. D. (1991). Local properties of Kauffman’s N-k model: A tunably rugged energy landscape. Physical Review A , 44 (10), 6399–6413. https://doi.org/10.1103/PhysRevA.44.6399 Additional Declarations No competing interests reported. Supplementary Files LandscapeComplexityShapestheRoleofNetworkDensityandDiversityinCollectiveAdaptationUnderDisruptionSupplementaryInformation.docx Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 05 Mar, 2026 Reviews received at journal 23 Feb, 2026 Reviews received at journal 10 Feb, 2026 Reviewers agreed at journal 24 Jan, 2026 Reviewers agreed at journal 20 Jan, 2026 Reviews received at journal 19 Jan, 2026 Reviewers agreed at journal 10 Jan, 2026 Reviewers invited by journal 08 Jan, 2026 Editor assigned by journal 26 Nov, 2025 Submission checks completed at journal 24 Nov, 2025 First submitted to journal 14 Nov, 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-8115470","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":551501733,"identity":"a9dc0cf1-ddf2-445e-8a64-bed9c880227b","order_by":0,"name":"Ross Tieman","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6ElEQVRIiWNgGAWjYPACCRDB+ADCSSBGRwJYC7MBKVrAJJsEUVrM29sff2D8YSGn2372WMXPHYcZ+NlzDBh+tuHWInPmjJkE0GHGZmfy0m72njnMINnzxoCxF48WCYkcNpBfErcdyDG7wdt2mMHgBtAWXrxa0h9/AGqp33b+jVnhX6AWe6AWxr94tSQYgByWYHYjx4wZbItEjgEzXlt4gH5JSJMw3HbjjbG0bFs6j8SZZwWHZc7h0cIODLEPNnXyZudzDD++bbOW429P3vjwTRluLWCQgMTmAREHCGgYBaNgFIyCUUAAAAAf8ktI2HfohwAAAABJRU5ErkJggg==","orcid":"","institution":"Australian National University","correspondingAuthor":true,"prefix":"","firstName":"Ross","middleName":"","lastName":"Tieman","suffix":""},{"id":551501735,"identity":"5b697876-fdc0-4b5a-a5d3-b4ddf26a51fa","order_by":1,"name":"Robert Ackland","email":"","orcid":"","institution":"Australian National University","correspondingAuthor":false,"prefix":"","firstName":"Robert","middleName":"","lastName":"Ackland","suffix":""},{"id":551501736,"identity":"4a892589-5274-405a-a274-2eea7fe2301d","order_by":2,"name":"Katherine Daniell","email":"","orcid":"","institution":"Australian National University","correspondingAuthor":false,"prefix":"","firstName":"Katherine","middleName":"","lastName":"Daniell","suffix":""},{"id":551501738,"identity":"f4d826a6-7b62-4862-ae8e-d94556f29206","order_by":3,"name":"Steven J. Lade","email":"","orcid":"","institution":"Australian National University","correspondingAuthor":false,"prefix":"","firstName":"Steven","middleName":"J.","lastName":"Lade","suffix":""}],"badges":[],"createdAt":"2025-11-14 13:23:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8115470/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8115470/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97666209,"identity":"d4279402-d209-4d53-b703-75d40eee6f39","added_by":"auto","created_at":"2025-12-08 09:20:38","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4712004,"visible":true,"origin":"","legend":"","description":"","filename":"LandscapeComplexityShapestheRoleofNetworkDensityandDiversityinCollectiveAdaptationUnderDisruptiontr.docx","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/90db3208dcc8a4be5b4123aa.docx"},{"id":97411010,"identity":"f603f012-9e37-4044-a742-212fd31a862e","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6380,"visible":true,"origin":"","legend":"","description":"","filename":"d98940d32c444259868df01935892bd2.json","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/3bd3da0f9e684a9c18b14f86.json"},{"id":97411026,"identity":"87c7101c-d06f-4b0c-a3a4-7ee66373481c","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2264196,"visible":true,"origin":"","legend":"","description":"","filename":"LandscapeComplexityShapestheRoleofNetworkDensityandDiversityinCollectiveAdaptationUnderDisruptionSupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/020b4bde354d0373ee3415bd.docx"},{"id":97667850,"identity":"a7b7a1bf-318d-4775-a54a-c3b1385fd2a0","added_by":"auto","created_at":"2025-12-08 09:24:22","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":112553,"visible":true,"origin":"","legend":"","description":"","filename":"d98940d32c444259868df01935892bd21enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/027fe89ed0fe4ad1a475d0ea.xml"},{"id":97411016,"identity":"953ed986-92b4-4064-ba5b-fe39da58972c","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1591998,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/b65467f91384b0e5d6409a58.png"},{"id":97411011,"identity":"37f77c5e-5798-435e-9aeb-245daba64f96","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":57571,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/b80e392f51aa8b9741e380fc.png"},{"id":97666228,"identity":"1ce149fb-dd7c-4781-8ff3-4b435efae5a8","added_by":"auto","created_at":"2025-12-08 09:20:41","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94715,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/ab20c5d75f218882c8657075.png"},{"id":97666243,"identity":"f07e9c64-ab4e-45a9-8d33-1e94165bd783","added_by":"auto","created_at":"2025-12-08 09:20:42","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94623,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/3f29e150dcee816ffb3badb7.png"},{"id":97411017,"identity":"d6ef99b2-0ae5-4eb7-ac4f-d1634ecc3ce0","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94988,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/d64f3ebfbccde801a956d3b0.png"},{"id":97666368,"identity":"094d04d1-c3d2-4435-9633-b169ac5522ff","added_by":"auto","created_at":"2025-12-08 09:21:04","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":96228,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/06aaae60d158d6636de341f9.png"},{"id":97668014,"identity":"9af62d97-2313-4198-aad2-a6bd7ffa9c31","added_by":"auto","created_at":"2025-12-08 09:24:39","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7982,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/869b6150a760fbb247c8267d.png"},{"id":97666288,"identity":"1e54e512-d22c-420b-9132-f713dec5efd5","added_by":"auto","created_at":"2025-12-08 09:20:51","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7047,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/0585a88a7db6d26b0a80e318.png"},{"id":97411034,"identity":"661c8740-e439-46ae-a7c8-3f7e2330a7a4","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":105282,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/cb064e9c4e7008008366982c.png"},{"id":97411024,"identity":"cacc8b81-4d1c-4326-b76c-b39068eb15f8","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":116045,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/53d7074e5a9c312f9c52b36d.png"},{"id":97666819,"identity":"0aac927e-880f-4f6d-9c8b-8172b3a59fb9","added_by":"auto","created_at":"2025-12-08 09:22:11","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":291086,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/58aff25ddbec0374328e61a1.png"},{"id":97411043,"identity":"29d14bab-c48f-4cc2-bfc6-63f8c548ff88","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":355846,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/136a2e33f83130929e91deed.png"},{"id":97411023,"identity":"6c62be28-7ed3-40d9-912c-77edda1a9dc8","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":59500,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/018c5b70a6ebc976f9e3e042.png"},{"id":97666261,"identity":"1b64fe37-2a91-415e-96c1-3da802e29090","added_by":"auto","created_at":"2025-12-08 09:20:46","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":109083,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/0a2a0e8788761fc1abaabc10.png"},{"id":97411045,"identity":"bac824d9-a6dc-4001-90c0-246933360190","added_by":"auto","created_at":"2025-12-04 05:25:48","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":942294,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/4e0c13521fc6a70733b88fe4.png"},{"id":97666000,"identity":"073c4071-559b-47be-9fab-dd094f4cccad","added_by":"auto","created_at":"2025-12-08 09:20:14","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19167,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/bef7fe6c54b702ab43ef34cd.png"},{"id":97411031,"identity":"c1d313db-ba1a-4557-9b65-b8352172dd61","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30183,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/c10572ccaa0b2625fe95d2cf.png"},{"id":97411028,"identity":"ab726020-cbf2-40ce-9ee3-a636caabc779","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29431,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/01b0478c6a0879008ce139ea.png"},{"id":97666125,"identity":"71729f8d-aabd-4788-80ab-916e5af23147","added_by":"auto","created_at":"2025-12-08 09:20:29","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30377,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/a6aab7360b89038a173b501b.png"},{"id":97411037,"identity":"a7bc4a4f-aac1-4154-8d3d-d98f84e7b073","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29194,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/78fda4d97ef8a497627e7771.png"},{"id":97666441,"identity":"208e3ba7-96e0-4154-9265-85927ce4dcb8","added_by":"auto","created_at":"2025-12-08 09:21:14","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4085,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/de368b5ee0c02a38a4a37b40.png"},{"id":97411041,"identity":"7b7d2180-f07e-451d-b179-427b3ccf2389","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4036,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/82ad4262b73e7a9709406d51.png"},{"id":97666137,"identity":"a3fcbabd-695e-41d7-bbaf-4b54baefa13a","added_by":"auto","created_at":"2025-12-08 09:20:29","extension":"png","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":26358,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/222f55337e8ad81e11658cbc.png"},{"id":97667314,"identity":"146a1830-a408-47db-a412-609aa4d26e96","added_by":"auto","created_at":"2025-12-08 09:23:14","extension":"png","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28488,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/64ab5b6de426dad784348aae.png"},{"id":97411040,"identity":"83081549-8233-4181-afce-17c4dd96ab94","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":68452,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/37748c17edf28f057ff4cde5.png"},{"id":97411042,"identity":"9982ed07-996e-41b0-83a0-495eeca8089c","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":79919,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/d9a49019c39a59513f25a40b.png"},{"id":97411039,"identity":"2e596406-360a-4e81-90dc-034bd0b8d30e","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17936,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/c9770165c16f385b252afd30.png"},{"id":97666296,"identity":"f0d935ad-f2c6-4e56-91b9-a568d38e01e3","added_by":"auto","created_at":"2025-12-08 09:20:55","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":34016,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/e6a1bb9467a99b2124331fb1.png"},{"id":97411044,"identity":"6af388a0-6ec1-4f7d-9901-a2b46fa69ef3","added_by":"auto","created_at":"2025-12-04 05:25:48","extension":"xml","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":110548,"visible":true,"origin":"","legend":"","description":"","filename":"d98940d32c444259868df01935892bd21structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/e444adf86d9544b6fc415c8d.xml"},{"id":97667036,"identity":"98606df9-2b7f-4221-9009-850a24bbe1bc","added_by":"auto","created_at":"2025-12-08 09:22:40","extension":"html","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":119251,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/e96d421835e615dd9c765688.html"},{"id":97411007,"identity":"f97870fc-66cb-4d91-ab6a-b2c1c7a8be8a","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1591998,"visible":true,"origin":"","legend":"\u003cp\u003eA graphical representation of the multi-agent model. (A) model is defined by complexity of Task (NK landscape), density of social learning network and diversity of agents. (B) Disruptions are regime is defined by frequency (number of disruptions) and disruption characteristics size of disruption impact and inequality of impacts – skew of impacts across agent groups. (C) A simulation with diverse agents, connected by a sparse social network explore a complex NK landscape to find good solutions. The agents socially learn from other agents in network or innovate to independently find new solutions directly from landscape, different agent groups value solutions based on unique payoff functions, until disruptions occur. (D) Disruptions result in changes to agent group payoff functions, proportional to impacts and skew, the social network is not impacted, agents then continue collective learning process for the defined number of timesteps, with ongoing disruptions occurring. (E) Throughout simulation collective performance is recorded, simulations are run under various disruption regimes with median cumulative performance of collective to measure capacity of collective to maintain performance under disruption regimes. A Factorial approach is used to compare across of simulation parameters (Landscape Complexity, Diversity, Network Density, Disruption Regime [Frequency, Impact Size, Impact Inequality]) with 83 run per condition totalling 55,000 + simulations.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/d1c8e0aaeed73bfc6a04dbf3.png"},{"id":97666933,"identity":"cd1869b0-4297-4d0a-b665-76df5cd13ef9","added_by":"auto","created_at":"2025-12-08 09:22:29","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":105282,"visible":true,"origin":"","legend":"\u003cp\u003eTemporal dynamics in a representative simulation run (K\u003csub\u003eNK \u003c/sub\u003e= 7, G = 5, d = 0.76, N = 100, T = 200, disruption regime: frequent, severe and skewed). The vertical dotted lines show when disruptions occurred. (Top) The red line shows the performance of the collective over time, the green line shows the number of unique items (solutions) held by all agents in the collective. The light black lines are individual agents. (Bottom) social learning and innovation events are also logged over time and spike following disruptions as agents socially learn and innovate new solutions based on altered payoff functions.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/20c731572f3264e37ebc7d1e.png"},{"id":97665898,"identity":"cfb6c451-d116-4d7a-8c84-b90f9598c0b4","added_by":"auto","created_at":"2025-12-08 09:19:55","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":116045,"visible":true,"origin":"","legend":"\u003cp\u003ePerformance effects relative to control (no disruptions) across all disruption regime combinations. The x axis shows the different disruptions regimes (Impact frequency [frequent, infrequent], impact size [mild, severe], impact inequality [uniform, skewed]) see Table 1 for parameters. Simple landscapes (K\u003csub\u003eNK \u003c/sub\u003e= 0) show small negative effects while complex landscapes (K\u003csub\u003eNK \u003c/sub\u003e= 7) show large positive effects. Impact inequality (uniform vs skewed) (skewed show diagonal cross hatching) that distributional characteristics matter primarily in complex landscapes and only for Severe impacts. Error bars represent 95% bootstrap confidence intervals.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/71ebab94ff46952e5491491c.png"},{"id":97411009,"identity":"c1ca0bdf-c17a-4ebf-8951-38871e0d239d","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":291086,"visible":true,"origin":"","legend":"\u003cp\u003eNetwork configuration and diversity effects on cumulative median performance in simple landscapes (K=0). (A) Control: Under control conditions, performance favours high density, low diversity. (B) Disruption: Under disruption conditions, there is a shift toward higher diversity requirements (G = 3-5). (C) Density effects: logarithmic-like performance increases observed as density increases across disruption regimes (frequent or infrequent, mild or severe, uniform or skewed). (D) Diversity effects: sharp increase in performance for frequent-severe disruptions at G = 2-3 (though G≥5 eliminates disruption effects; see Supplementary Figure S6).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/ed4322313234febfdd22190b.png"},{"id":97411013,"identity":"e385ba68-7084-40b8-81e3-15db8f378ade","added_by":"auto","created_at":"2025-12-04 05:25:47","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":355846,"visible":true,"origin":"","legend":"\u003cp\u003eNetwork configuration and diversity effects on cumulative performance in complex landscapes (K=7). (A) control: Performance heatmap under control conditions (no disruptions) showing preference for moderate diversity levels. (B) Performance heatmap under disruption conditions showing enhanced performance for preferred collective structure under control conditions (moderate density high diversity) (C) Performance versus network density across all disruption regime combinations (9 lines – disruption regimes [frequency × severity × impact distribution] and control). (D) Performance versus agent diversity across all disruption regime combinations, illustrating the groups × impact distribution interaction effects.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/a9dfa9b008d891062194c185.png"},{"id":97677540,"identity":"4ab9819b-dcc2-4969-883c-c6f8389a0af3","added_by":"auto","created_at":"2025-12-08 09:53:25","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1699761,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/bb60f0d3-5922-400d-a673-bcaf25bdec6d.pdf"},{"id":97666972,"identity":"8f6e3260-b22e-43aa-9a4b-4e81edd9123d","added_by":"auto","created_at":"2025-12-08 09:22:33","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":2264196,"visible":true,"origin":"","legend":"","description":"","filename":"LandscapeComplexityShapestheRoleofNetworkDensityandDiversityinCollectiveAdaptationUnderDisruptionSupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-8115470/v1/20e48db1a1ced42e258b2b1e.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Landscape Complexity Shapes the Role of Network Density and Diversity in Collective Adaptation Under Disruption","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eClimate change, the polycrisis (Lawrence et al., 2024), and other sources of major disruptions present human collectives with unpredictable challenges requiring simultaneous adaptation across multiple domains. For example, agricultural collectives facing increased climate volatility must adapt crop selection, supply chain management, and resource allocation while maintaining productivity (Folke et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). The potential for critical transitions in such systems, which can cause abrupt shifts in dynamics requiring rapid adaptation (Scheffer et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), raises questions about how collectives should be structured to maintain performance in increasingly turbulent environments.\u003c/p\u003e\u003cp\u003eExisting research primarily examines collective problem-solving in stable environments. Network structure shapes information diffusion and collective intelligence, with dense connectivity accelerating social learning but potentially promoting premature convergence on suboptimal solutions (Burt, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Centola, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Derex \u0026amp; Boyd, 2016; Lazer \u0026amp; Friedman, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Mason \u0026amp; Watts, 2012). Organisational learning theory addresses the exploration-exploitation trade-off, where collectives balance refining current solutions against searching for alternatives (March, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e1991\u003c/span\u003e). However, these frameworks assume static problem landscapes and require extension to account for environmental disruptions that can render existing solutions obsolete.\u003c/p\u003e\u003cp\u003eA critical gap remains in understanding how network structure and agent diversity jointly influence collective performance under ongoing disruptions, and whether these relationships depend on problem complexity. While early work examined how specific structures respond to single perturbations (Siggelkow \u0026amp; Rivkin, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), and recent studies show network structure moderates diversity effects in static environments (Baumann et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), we lack systematic analysis of ongoing disruption regimes. The Collective Adaptation framework (Galesic et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) provides conceptual foundations for studying dynamic problem landscapes, but computational implementations testing specific mechanisms remain limited. This question has broad relevance across complexity science, from biological regulatory systems to human governance (Gao et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Levin, 1998; Ostrom, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2010\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTo answer the question of how to structure collectives to maintain performance under ongoing disruptions, we adopt a socio-ecological resilience perspective because of its focus on structure-environment interactions. Unlike approaches that seek universal structural principles, this perspective recognizes that effective social structures must match their environmental context (Bodin, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Folke, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2006\u003c/span\u003e), with established resilience principles specifically highlighting the importance of maintaining diversity and managing connectivity (Biggs et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). By foregrounding the interaction between disruption regimes and collective structures in shaping performance, the socio-ecological resilience perspective aligns with core insights from the collective adaptation framework concerning the tuning of collective properties to given problem structures (Galesic et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). While motivated by human collectives responding to increasingly turbulent environments, the approach employed is sufficiently general to capture collective learning processes wherever networked agents search among interdependent options.\u003c/p\u003e\u003cp\u003eWe extend Baumann et al.'s (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) NK landscape model of collective learning by incorporating stochastic disruptions to problem landscapes. This extension enables systematic examination of how network connectivity and agent diversity jointly determine collective performance under varying disruption regimes. Disruption regimes are defined by three characteristics in our analysis: frequency, severity, and distribution of impact across agent groups. By examining these dynamics across both simple and complex problem landscapes, where problem complexity determines the degree of interdependence between decisions, we can test whether optimal collective structures depend on interactions between problem complexity, disruption regime characteristics, and collective composition (network structure, agent diversity).\u003c/p\u003e\u003cp\u003eWe systematically examine how network connectivity and agent diversity influence collective performance across simple and complex problem landscapes under varying disruption regimes. We characterise disruption regimes by frequency, severity, and distribution of impacts across agent groups, enabling analysis of how these factors interact with collective structure and problem complexity. Our approach extends collective intelligence research on network effects in learning (Centola, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Baumann et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) by examining how these effects persist or change under environmental disruption, while providing computational evidence for socio-ecological resilience principles regarding connectivity and diversity (Biggs et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) and polycentric governance structures (Ostrom, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2010\u003c/span\u003e).\u003c/p\u003e"},{"header":"2 Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Model Framework\u003c/h2\u003e\u003cp\u003eWe extend the collective learning model of Baumann et al. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) by introducing stochastic disruptions that reconfigure problem landscapes. Our framework models collective adaptation as a collective search problem where N interacting agents search a space of solutions over some number of timesteps T (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.). This solution space could represent combinations of decisions organisations must make or ways to arrange system components. Agents search the space through individual exploration and social learning from connected neighbours, with high-performing solutions spreading through the collective via network ties. The quality of a solution (S) is quantified by a payoff function \u003cem\u003eP(S)\u003c/em\u003e. Disruptions modify these payoff functions, changing the value of solutions and the overall performance of the group. Exposing collectives to disruption regimes creates ongoing adaptive challenges that allows systematic investigation of collective properties (network structure, agent diversity) that can facilitate performance under disruption.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 NK Landscape Structure\u003c/h2\u003e\u003cp\u003eThe fitness landscape is represented using the NK model (Kauffman \u0026amp; Levin, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1987\u003c/span\u003e), which provides a tunably rugged landscape suitable for investigating adaptive processes. The NK model is commonly deployed to investigate social phenomena such as innovation processes and decision making and social learning in collectives such as human organisations (Levinthal, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). The NK model is defined by two parameters: \u003cem\u003eN\u003c/em\u003e\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e,\u003c/sub\u003e the number of binary decisions, and \u003cem\u003eK\u003c/em\u003e\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e\u003c/sub\u003e, the interdependence (epistatic interactions) of decisions. Each solution \u003cem\u003eS\u003c/em\u003e is a vector of \u003cem\u003eN\u003c/em\u003e\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e\u003c/sub\u003e = 15 binary elements b\u003csub\u003ei\u003c/sub\u003e \u0026isin; {0,1}, and the payoff function is computed as:\u003c/p\u003e\u003cp\u003e\u003cimg src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAcIAAACHCAYAAACF+lkuAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAAFiUAABYlAUlSJPAAAEFISURBVHhe7d15XFT1/vjx1yww7LssAgICirvmrpl6MzNzLTXLvkXmbd9Lq1vd7N7bbmrd9izLNrXScsu1SBEUS1EWkUVA9lVgWGY/vz8unJ+MKKCAIJ/n43Ee5Xw+cxiGmfM+n+39UUiSJCEIgiAI3ZTS+gFBEARB6E5EIBQEQRC6NREIBUEQhG5NBEJBEAShWxOBUBAEQejWRCAUBEEQujURCAVBEIRuTSHWEQqCIAidhcViwWAwYDKZAFCr1dja2qJUtl+7TQRCQRAEodPIz88nLi6OjIwMFAoF4eHhDBs2jICAAOuqbUYEQkEQBKFT+PDDD4mPj6dHjx6MHj2ampoa9u3bh1arZebMmcyfPx+NRmP9tMvWfm1NQRAEQWgBg8FAfHw8X3zxBQUFBYwdO5ZRo0Yxfvx4goODSU1N5eOPP6aoqMj6qW1CBEKhWTqdjoqKCkpLSyktLaWkpISSkhKKi4spKSmhrKwMvV5v/TRBELq5mpqaRteKhv+WlpZSVlZGVVUVANXV1ezbt4/U1FTGjRvHjBkz8PX1pVevXixcuJCRI0dy8OBB8vPzaY9OTNE1KjTr4MGDxMXFUVBQgFKpRJIk+cOoUChwc3Nj+vTpDBkyxPqpgiB0Yzt37uTAgQMYjUb5MYVCgUKhACAsLIwpU6bg5ubGqlWr+PDDD1m2bBlLly6V6+fk5LB69Wr++9//sm/fPsaNG4dKpZLL24JoEQrNSkxMZOPGjaxbt47Dhw8D4OHhgYuLCy4uLjg5OWFjY2P9NEEQujk7OztcXFxwdnbGxcUFLy8vtFote/bs4YsvvmDHjh1kZWXh6urK6NGjMRgMJCYmyi1FgNTUVFJTU4mIiCA4OLjNgyCAavny5cutHxSEc0VFRbFnzx4KCgro378/c+fOZcyYMQQHBxMeHk5oaCj+/v7Y29tbP1UQhG7MxsYGHx8fwsLCCA8PZ+DAgRgMBg4fPkxmZiYeHh6MHj2avn374ubmhtlsJjs7m5iYGAoKCjh06BA///wzKpWKRx55hGuuuaZdAqFoEQrNUqlU2NjYYGdnh7+/PyNGjGDAgAGMGDGC4cOHM2jQINzd3a2fJghCN+fv788111wjXyv69evH4MGD8fLygvpA2dBN2qNHD66//nqo71KNiopi69atHDlyBBsbG8aOHdtuPU8iEAotJkkSZrMZg8FgXSQIgtAiBoMBi8Ui/7shEJaUlPDqq6+iUCj47LPP+Pbbb/nll19YtGgR8fHxPP7441RUVJxzprYjAqEgCIJwxdjY2KDX61m/fj3Z2dmMGjWKadOmoVQqsbOzY+bMmQwfPpx9+/aRnJwsZ5xpSyIQCoLQLiRJorS0lJMnT/LXX3+RkpLSaPagyWQiNzeXpKQk4uLiSEhIoLa2tl2mxwudl0qlora2lhMnTmA2m/Hx8WlUHhISQlBQEGfPnqWoqKhdlmqJQCgIQruQJAmTyURsbCyfffYZn3/+OampqXLXuiRJ1NbWsm3bNjZu3MixY8cwm81yV5nQPUiShEqlwtHREb1e32jGKEBdXR01NTVoNBqcnZ1Rq9WNytuCCISCILQLpVKJr68vOTk5fPfdd3z44Yds376dsrIyqO8SCw0NJSoqisOHD2NjY4Ozs7P1aYSrnNlsxsnJifHjx2OxWDh16hTFxcVy78GxY8dIT0+nT58+DBgwQKRYEwSha+rZsyd9+/Zl//79FBcXy4+rVCpMJhP9+/eXZwwK3YvRaESpVHLzzTezdOlS0tLSuPXWW3n++edZsmQJb731Fk5OTnz44YfybNO2JgKhIAjtqrCwEE9PT2666SaOHDlCbm6uXGYymaitrcXT0xNvb+9GzxO6h4YxYQcHBxYuXMizzz7LkiVLGDt2LDfeeCOPPfYYjz32GKNHj8bW1tb66W1CBEJBENqNwWCgqqqKoKAgpkyZQklJCSdPnqSurg6AM2fOoNFoxDrUbuzcMeGAgACmTZvG3Xffzc0338z8+fNZsGBBu64hRARCQRDai9lsJiUlBTs7O4YMGcKAAQPw8fHh6NGjJCUlYTKZOH78OAEBAe2615zQNdnZ2Vk/1G5EIBQEoV2YzWYSExNxdHSkV69eODg4MGHCBDIzM0lKSsJsNpOUlISfnx/+/v7WTxeEDiMCoSAI7aIhEDo7O9OrVy80Gg1TpkyhqKiIpKQkqqqqSEhIwN/fXwRC4YoSgVAQhHZhsVg4ceIELi4uBAcHY2Njw8SJE9FoNKSnp5OYmEhSUhL+/v6ia1S4okQgFAShXRgMBtLT03F1dZVbfH379qVv374UFxeza9cuCgsL8fX1bZe1YYLQUiIQCoLQLsrLyzEajbi4uDR6fOzYsVgsFrZu3UrPnj3FInrhihOBUBCENldTU0NaWhpBQUHnLYIeM2YMPXr0ICMjg4iICJycnBqVC0JHE4FQEIQ2l5CQwLp168jJyWmUSQZg+PDh9OrVC71eT//+/XFwcGhUTn0gLSgoID8/X061ZTQaKS8vJzc3l6KiIurq6rBYLJSWlpKXl0d+fj5nz54FoKioiIyMDHJyckQib6FZIhAKgtCmGgJWSUkJPXr0wGKxoNVq5XInJydGjx7NpEmTGDlyJI6Ojo2eD1BbW8vx48f5/vvvOXXqFAaDAYVCQUlJCT///DOpqalYLBYkSaKoqIgtW7Zw4MABeaH+/v372bFjBwkJCej1ehEIhYsSgVAQhDZlY2PDTTfdxK5du4iOjub//u//zhsHXLRoEfv27WP69OnnlQG4uLiQnZ3NCy+8wPr168nLy0OtVuPi4sKaNWuoqKjA0dERhUKBi4sLn376KdHR0fTs2RPqdziPi4vDbDbj7u6OUikudcKFiU+H0CylUolSqUShUKBWq8U2OUKzFApFo8+NtYbypsoAKioqcHBwYNKkScTExJCdnQ2AXq8nIiICT09PqF+rePbsWfz8/Ojbty8AP/zwAyEhIdxxxx0ikXcnpFQqUalUUJ90vTPcpFz5VyB0eqWlpVRWVmI2m8nNzUWn01lXEYQ2lZOTg9ls5tZbb5XH/0wmEzk5OQwePFgOhAaDgdTUVMLCwlCr1Rw6dIjS0lKmTJnCdddd1+T4o3Bl1dTUUFRUBEBxcbE8rnsliUAoNKvhw2o2m8nOzm6XHaIF4VxZWVlUVFQwceJEfHx8KC0tJT4+nrS0NPr27YubmxvUtxBPnDiBo6MjSUlJ/PDDD4SFhTF8+HAcHR3F2GAnpNVqyc/PB6CgoEDen/JKEoFQaJbRaMRsNiNJEgaDAYvFYl1FENpUUVERpaWlhISEMGPGDM6ePcuOHTvIzMw8LxAmJSWh1WqJj4/nwIED9OnTR+56u1DXq3DlmM1mDAYD1G/DZTabrat0OBEIBUHodCwWC0ajERsbG2644QbOnj3L3r17KS0tJSAgQM5EYzAYOH78OFOnTuWpp55iyJAh/Pzzz52iu03oOkQgFAShU5EkCVtbW1xdXQEYOnQoZrOZrKwsDAZDo81ZtVotkiTh5+fHTTfdxOjRo/nxxx+prKw854yCcHEiEAqC0GlIksThw4c5c+aMvBGrQqHA29sbT09PnJyc5HG/4uJiDh06BPVdpHZ2doSHh3Ps2DH279/faO2iIFyMCISCIHQakiQRHR1NRkZGo2n1Q4cOZfr06URERMjjf9nZ2Rw8eBAXFxc5+4ybmxuBgYH8+uuvxMfHy88XhItRSGJaldCMxYsXs3btWhQKBWFhYWzcuJGhQ4daVxOENqHVajEajdja2sp5SHU6HQaDAZVKhYODAwqFAoPBQE1NDUajEVdXVzQajZzVxsbGBkdHR7GrRSe0YcMGlixZQnV1Na6urixfvpwnnnjCulqHEi1CQRA6FWdnZzw8PBol47azs8PFxUXOJtMwjuju7o63tzcajQZJkrCxscHHxwcPDw8RBIUWE4FQ6JK0Wi2VlZViTWMr1dXVUVhY2CmmrF+OppZFNPWY0Hau5s5DEQiFLsVkMpGfn8/OnTuJj4+noqLCuorQBJPJRHV1Nb///jtbt26Vx9SE9mUwGCgvL+8Ui8Yvl0KhoKamhsLCwja7AZUkqVMEWBEIhWad+0FtyPh/JVgsFn777Tfuu+8+7rrrLn7//Xd5Ya5wcadOneKJJ55g4cKFfPLJJ1fsb9jdREdH89xzz/HPf/7TuqhL+vnnn5k3bx67d++mtrbWurhFJEmSk3KIQCh0GSEhIfj4+KBWqxk0aFCHb6RqMpmIi4tj5cqV7N69Gx8fH3Q6XZvdlV7NUlJSePfdd9m4cSNGoxGNRkNVVZXoRmwnDRd1o9HIxo0bef7559m3b1+TO2x0RSqViry8PJ5++mm+/vprSktLras0y9PTU06QHhYWRkBAgHWVDicCodAsNzc3nJ2dUalU+Pj4NFrQ3FFUKhVeXl4MHz6cyZMnX3RnA+H/U6lUuLq60rdvX8aOHUvv3r3F+9aOFAoFJpOJ2NhYVq1aRWFhIX/729+YO3euddUuafjw4dxzzz3U1NTw2WefsWXLFkwmk3W1i7K3t5eTpru7u3eKmwQRCIVmmc1mOdeo0Wjs8K4MtVrN4MGDiYyM5Pbbb2fAgAGoVCqR87QFwsLCiIyM5I477mD06NG4urp2+YkynV1iYiJfffUVcXFxLFy4kBdffJHRo0dbV+tyJEkiPDyc5557jieeeIJTp06xbt06YmNjratelMVikYOn2WzuFN9jEQiFLqEhywj13U5Cy5zb8hMJ09ufJEns3r2bDRs2cOONNxIZGUlQUJB1tS6p4bNka2vLk08+ya233kpCQgKrVq3q8t9JEQgFQRDaSFRUFLt378bW1pbHH3+c3r17W1e5KqjVaiIjIxk0aBCHDh0iKiqKuro662pdhgiEgiAIbWTnzp2cPHmSa665hnHjxl3Vi/qvu+46xo4dS3V1NRs2bOjSic6vukCo1+upq6tr9QCuNb1eT21tbYePhwnC1c5sNlNXV0dlZSVVVVVUVVVRWVnZ7NFQt+HQ6XTWp243ZrMZrVZLRUUFlZWV510bJEni7NmzxMbGYrFYmDJlykUngZjNZmpqaqipqaG6upqampo2WQpksVjQarXU1NRc9jWwOUqlkmHDhhEQEMCuXbsoLCy0rtJlXJFA2LB2pLmjtcxmM8ePHyc5OfmiC63P/RkWi0VeG3fuz83JyeHPP/8kLy+vy/d/C0JnUl1dzfHjx/nll1/Ytm0b27ZtY/v27ezYseOCx/bt2+W627ZtY+vWrZw8edL61O1Cp9ORmZlJTEwMW7duZdu2bfz111+Nrgt6vZ4DBw6Qk5NDaGgokydPvug1rLq6mvj4eKKjo9m1axcxMTEUFhZe9hiuVqtl586dHDx4sNnAZH29vZTrbt++fRk/fjy5ubkkJSVd8trCK63Dk25nZmYSFRVFVlYWSqUSqX7Xc7PZjEqlwtHRkZ49ezJw4EBGjhxp/fQmGY1G9u/fz6pVqygvL+fWW2/ljjvuwM/Pz7oqAPHx8URFRZGYmEhFRQU2NjZ4eXkRFBREv379GDduHBkZGXz//ffs2bOHiRMncu+993bbRNOrVq3iv//9L/n5+SxatIiXXnqJ4OBg62od5siRI4wfP55nnnmGhx56qFOsQ+oK4uLieOGFF8jKyiIpKemKLIMBqKys5L333uOtt97C1dUVGxsbPDw8cHV1lVtF1ss7KisrKS8vR6fTodFosFgsPPzww7z44ouN6rW1qKgovvvuOwoKChg2bBhxcXGkpqYyfvx43nvvPdzd3QGoqqri1Vdf5b333mPu3LmsW7cOtVptfTpZUVERP/30E+vWrSM1NZXrrruOF198kaFDh170ec1JSUlh7ty59OnTh4ceeogbb7zRugrUX4f3799PcnIyTk5O8kbIgwcPZsGCBdbVL+js2bN89913PPLIIzz66KM89NBDREREWFdrZP/+/bz88stERUUxceJEli5dys0332xdrUN1eIvQy8sLX19fDh8+zNdff822bdvIycnB1tYWSZLIzMzkp59+4oUXXmD58uWcOnXK+hSNlJaW8umnn/LWW29hb2/PrbfeynXXXYeLi4t1VcrLy3nnnXf497//TXR0NF5eXgwePJiQkBBOnz7Nhx9+yBtvvEFCQgJ+fn5MmzaNOXPmcOrUKd5++202bNhgfcpu4dw1e2q1+ryLVEc7d3ueK3Ux74oalpw0/B2vFFdXVwIDA/Hw8CAvL4+srCxcXFyYP38+S5YsYcmSJSxevFg+7r33Xh577DEWLlyIr68v+fn5FBYWXtJi7tbYu3cvH3/8MadOnWLu3LlMmzaNiRMnyt2BDdtBUX8zHhsbi0KhIDg4uNn3183NjeHDh0N9MCkvLycwMLDROS+FXq8nIyODnJwcqqurrYtlXl5eaDQaUlJSWLt2LV999RV1dXX06dPHuupFubu7ExERga2tLUePHuXMmTPWVc6jVCrl31OlUjX6Pl8pHf4KnJ2dueaaawDIyMjAYDAwYsQIZs6cyfTp05k8eTKenp5ERUXxxhtvsH379gtusKnT6di4cSNfffUVJSUl3HfffTz99NOMHDkSR0fHRnXLy8v5+eefeeONN0hPT2fYsGHceeedPPTQQ9x///2MHj2asrIyoqOjKS8vx9/fnxtuuIFXXnmF6dOnc+LECd5//33++OOPdu9772xqamrQ6XRYLBYqKiquyO+v1+s5c+YMMTEx7Nu3D6PRyIkTJ/jtt99IS0ujvLxcrI9rgl6vp7CwkISEBH7//Xeys7MpLy9n69atHD16lOLi4svujrsUU6ZM4f7778fOzg4ABwcHhg4dSmRkJJGRkdxzzz3yERkZyZIlS/jXv/7Fo48+KrfCmgs2l6O6uppPPvmEmJgYhgwZwuLFixk3bhyRkZE8/fTTzJs3D3t7e7m+VqslJSUFLy8vevXq1ehc1iRJQqPRMHr0aHx8fLCxsaF37974+Phc9k2mUqnE3t4eOzu7iwZVZ2dn+vTpg4uLC5mZmdTW1jJy5MhL6vXy8fHB29ubU6dOUVBQYF18Hr1eL1/TtVptp+hO7fBASP3dU8Od/IABA5g1axajRo1iwoQJLFq0iBdffJE5c+ag1+vZtWtXkws29Xo9hw8f5u2330apVPLGG29www03WFeTHTt2jH//+9/odDpeeOEFXnjhBQYOHCh3ic6ePZvbbrsNX1/fRt1+CoWCJ554gnvuuYf8/Hyee+458vLyGp37apebm0tJSQkmk4nU1NQO/+BaLBYqKyv55Zdf+Ne//sVHH32Eg4MDMTExrFy5kk8//ZTk5GSRcq0JFRUVREdHs3r1aj799FMKCgowmUz885//5M033yQ2NvaKjIEHBARw2223yckRduzYweuvv37RVoyDgwMLFiyQ03O1VwCvq6sjNjaW6OhoAgMDmTlzplzm5+fHuHHjGD58uLy21WKxkJubi1arpWfPnvj6+p5ztvM1BLvq6mpSU1Nxd3dvdUusLUiSJAfLadOmNduleSEajYawsDBKS0spLi62Lj5PZWUlmZmZAJw5c4aioiLrKh3uigXC9PR0AHr27ElISEijcmdnZ4YPH469vT15eXnk5+c3KgcoKytj+fLl1NXVMWvWLKZOnWpdRWYwGEhOTiYrK4tRo0ad9/MAfH19mT59Ovfddx89evSwLua2225jwYIFHDlyhPXr1zf5mq5Wer0eg8GAJEnU1tZ2eMtLqVTi6enJXXfdxRdffEFUVBRJSUkcPXqUTZs28dxzzzFixIhGd+jC/3h5eTFt2jRef/119u7dS2JiIvHx8Wzbto13332XKVOmXLHu5Z49e/LSSy/JY7zx8fF8/PHHVFVVWVeV2draMmXKFNzd3dvthkyr1bJp0yaKioqIiIiQuzAvpKKigrS0NMxmM35+fnL6sIuRJIkzZ85QVlZGr169CA8Pt67S7kpLS8nOzkapVNKvX78mr3stodFo5KQBBQUFzf5dTCaTPOO34dpypV2RQHj27FlycnJQKBT07NnTuhilUomdnd0F80nW1NRw8OBBDh8+LLckL6asrEyeQWVnZ9fk2h53d3euvfZa7r77bry8vKyL8ff359prr8Xb25tvv/2WlJQU6ypXrXPnU3Xw3CpZQ87Mnj17EhQURHBwMMHBwQQEBODp6Ymdnd15nxPhf++bk5MT3t7e9OrVi5CQEIKDgwkMDMTX11fe6PZKsLe35/rrr2fq1Kk4OTmRk5PDF198QXJysnVVma2tLdOmTWPRokUMGzbMurhNZGdns2fPHiRJwtnZGaPRKPeKNNXroNVqyc7Oxmw24+vr26JAqNVqSU5OpqamhtDQUAYMGAD1k26Sk5PJyMho95Z6YWEhp06dQpIk+vfvj7e3t3WVFtFoNAQGBqLRaCgpKWm2kXDu7NSGGftXWocHQr1eT3Z2NtXV1YSEhDQZCBumK+t0Ory9vfHx8WlUnpmZye7du6mrq2Py5MkMGjSoUbk1lUoljyekp6c3+YeytbXFy8uL3r17NxkoAXr37s2YMWNISEjgxIkT7dY1I0kSer0enU53SUddXR06na7DW26C0FpOTk7ccccdTJw4EYCTJ0+yYcMGUlNTratC/bjgkCFDuPvuu5k0aZJ18WVJSkpi8+bNbNiwgdOnTwOQmprKN998w5dffsmhQ4eaXJZVW1srL33w8PBocqKeNa1WS2JiInq9noEDB+Lh4UF0dDRffPEF7777LqtXr+b777+nrKys3QJFQUEBBQUFODs706tXr0ZpDFvDxsYGb29vbG1tOXv2LCUlJdZVLkihUFyxG7FzdXggLC4uJjk5GYVCQURERJNLHPLz84mOjsZsNtOvX7/zug2ys7OJjo6WtwVyc3NrVG7N29tb7g5NTU1lz549pKamtjqQubu7M378eGxsbIiPjycpKcm6SpvQ6XRy2qJLOX7//XdiYmLkcT1B6MwmTZrE9OnT5Ukwn376Kbt377auBvUXTicnJ0aMGCGPFbaVLVu28Oabb7JhwwYkScLT05Ps7GzWr1/Ptm3bLjg+rtPp5Bmszs7OODg4WFc5T1VVFYmJidjY2ODr60tubi6rVq1izZo1/PTTT7z//vs8+eST7Nu374KTBS+HxWKRGwShoaG4urpaV2kxlUqFm5sbNjY2VFVVNXmz0Nl1eCAsLS2Vm+N9+/Y9r0VYWlrKvn37+Ouvv/D09GTChAmEhoY2qpOfn09KSgphYWF4eHg0KruQa665hgULFqBUKlm5ciXLli3j4MGD1tUuytXVlUGDBmFnZ8fJkyebXdpxqYqLi/n73//O7NmzmTVrFnPmzGnVcfPNN3PXXXcRHR3d5Be3rbXXHWtn0t6/Y3ufv6O19veZNm0aDz74INS3sLZs2cLWrVutq7WrZ555hrVr13LvvfdC/czWhpniO3bs4MEHH2xyRmhdXZ0cCF1cXM6bsd6UyspKkpOTGTJkCDt27GDjxo3MmzeP3377jZ07d7Jo0SLKy8tZu3Ytubm51k+/bKmpqeTn56NUKhk4cOAFe8FaQq1W4+bmhlqtpqqqirNnz1pX6fQ6fEH97t27efHFFzly5AgrVqxg3rx5ODs7ywHyp59+YteuXSgUCl544QXmzZvXqGvUbDbz6quv8vLLLzN16lRWr15Nv379Gv2Mpuj1eo4dO8Z//vMftm/fDvW58m677TbuuOMO3NzckCSp2WZ6amoq48aNQ61Wy9uRtLWKigrWrFkjf6Cae03WjEYjrq6uzJo1i/Dw8Mv6kAMsXryYtWvXolAoCAsLY+PGjedNs/7ggw/4/PPPW9Qt1JVotVqWLVvGbbfdZl10QYmJiZw8eRKNRsOMGTNavE5q9+7dvPDCCzg4OLT6b95ZVFdX89JLLzF79mzrombFxMTw8MMPEx8fj1Kp5Pbbb+f999/H1dW1w96PtLQ03n33XT744AOWL1/Offfd12Sv1bn++OMPnnrqKY4ePcratWuJjIy0rnKenTt3MmvWLJycnBg9ejTz5s3j1ltvlXu3Vq5cydNPP01AQAA//fQTo0aNsj5FkxISErj22msZMGAAy5YtY86cOdZVANi6dSuvvfYaKSkpPPfcc9x3331yi7y1dDodMTEx3HnnnTg7O/P0009z3333WVeTbdiwgSVLllBdXY2rqyvLly9vl+toa7TsG9qGSktLSUtLAyA5OZnNmzezdu1aNmzYwI4dO8jJyWHIkCEsXbqURYsWnTc+WFpaSnl5OdR3ebZkxptUv25nzJgxPPzwwyxevJgePXqwf/9+PvjgAzZt2kRpaWmLvmzu7u7Y2NhQVlbWbgt6XVxcuPfee3nyySd58skneeKJJ1p1PPPMM9x///306dOnRe9PW3BwcMDDwwN3d/er6vDw8JDXurWE2Wxm48aNvP7666xdu5aEhIQWT3qwtbXt8u+hh4fHJd94DRw4kCVLluDr64vFYmH//v18//331NTUWFdtN9nZ2Rw/fhzq93JsSY+TyWSSd15oyTibwWAgNzcXo9GIVqtl2LBhTJ06tdEQT8P/a7XadhneSE9PJy8vDxcXFwYMGHDJfzPqb9Q1Gg1KpRKDwdDkhKJOT+pgK1eulADJ3d1duv7666W5c+dKs2fPlu68807p6aeflr7++mvp9OnT1k+TpaWlSffee6+kVCqlBx54QMrMzLSu0qycnBzplVdekby8vCRAGjhwoPT1119bV2tSZWWlFBQUJAHSE088YV18VbrnnnskQFIoFFJ4eLh07NixRuUWi6XRv69GLf0dq6urpXHjxkmA5O3tLa1bt06qrKy0rnaelp6/q7jU36e6ulqaMWOGpFarJUAaMGCAlJ+fb12t3Xz77beSt7e3ZGNjIx04cMC6uEm7d++WQkNDJUBav369dfF50tLSpKeffloCpGuuuUb6448/rKtIH330kQRIPXr0kGJjY62LL+jEiROSi4uLNHbsWGnz5s3WxbIHH3xQUqvV0qBBg6SMjAzJbDZbV2kxvV4vHT58WAoMDJQCAgKkd955x7pKI+vXr5ecnJwkQHJ1dZVWrVplXaXDdWiLsKysTE7BM2zYMFatWsWmTZtYv349a9as4bXXXmPhwoUX3cjSYDBgMBhQKpXyXUhrBQQE8NBDD/HWW29hZ2dHYmIiBw4caNF4miRJcgtB14HZ7zuzlrSku7qW/o6Ojo4MGjQIjUaDj48Pw4YNa9HkiZaev6u41N/H3t6eRx99lHHjxkH9eGFHjt4UFhZSXFxMSEhIs5PwGpjNZnktXEuy3eTl5ZGamopKpeL2228nLCzMugqV9VsatbTXq7XS0tIwmUx4enoSEhJySdfRBkqlEltbWxQKBUajsVOsC2ytS//tL0FmZiZnzpzBxsaGAQMGyOtWGtb22draolarL/pHaVhbKEkSJpPpkr8kXl5eTJw4Ue57b8h52Nz5FAqF3FVxsRRGl6O8vJzly5fz6KOP8uijj/LYY4+16njwwQd5/vnnOXbsmAjWV8AzzzzD7t27+fzzzwkNDW23z8nVSKlUUlNTQ2lpKe7u7jzzzDMdNu5cVVVFTk4OABERES26gWnQ3HXjXLm5uaSmpqJWqxk3btx5EwZzcnLk1zFo0KCLbufUWpIkUVBQIGeA6d279yXftDSQ6nfxOfffXc2FI047OH36NDk5OTg4ODBgwIAW9adbawiaFouFurq6Cy6BkCQJo9FIcXHxBVt6np6e9O3bF7VajdFobFHfttlslutdTr/6xeh0OuLi4ti/f/8lHVFRUcTGxrbZ8olz32Oz2dwlP+gdKSwsjOuuu46RI0dib29/2Rea7qSoqIgff/yRmpoaZs6cyYIFC1o0C7MtZGRkcObMGTQaDf37929xpiLFOUnMW7J2Nycnh+zsbLy9vZvcOeXkyZMkJSWh0WiYNGlSi8YpW0qv15OSkkJlZSW+vr5tktpNkiT5uqA8J6H2hTTUp/79utA1vCN1eCDMzs7G0dGRAQMGXNIH3NXVFRcXFyRJoqKi4oITEYxGI5mZmWzevFnOa2ft3KwGLi4uLUp6W1NTg16vx87O7rLW3lyMvb091157LTfccANTpkxp9TF16lQmTpyIj49Pi7pqmuPg4ICNjQ0KhQIXF5c2OacgWCsvL5f3KAwPD+eee+7By8ur2e9kW0lPTycnJwdHR0f69evX4hahSqWSuy8vdD06V15eHjqdjqCgoCZvpuPi4khJSSEwMJAJEybgfomzOZui0+lISkpCq9USEBBw3tK0S9HQ6JAkCRsbm2a7cm1tbXFycoL6oYTWTEZrLx0aCNPT0ykuLsbBwYG+ffs2+SFojru7uzyTNDc394KtuMrKSvbu3csLL7xATEyMdTHUr/85ceIEkiQRFhZ2XheFNZPJRH5+PiaTqUXJdS+Vu7s7zz//PG+//TYrVqzg7bffbtWxevVqli9fzpAhQ9rkQxYeHk5AQAA2NjZcc801bdpVIwgN4uPjWb58OR4eHkRGRrZ55pjmNLQIHR0d6d+/f4sDYcMYGfXXiIvJzc2Vd2jo0aNHo2Gghm7Lffv2UV1dzYwZMwgPD7/oUFFrNVzzqqqq8PPzazLvcmtJ9XvKWiwW1Gp1s4HQy8tLTikXERHR5NrMjtZ273AzGvLxUd+9eakJXqnf9sPBwYH09PQLdntqtVri4+MpKyvjjz/+OC9lU25uLu+99x5xcXHMmTOnUYb5C6muriYtLQ29Xk9QUBCBgYHWVdqMoj710OUcbeXcc17ul7JhPedvv/3Gr7/+yrZt29i8eTPbt2/n1KlTnaKbpDOTJInq6mr27dvHL7/8wqFDhygtLZXfN5PJREFBATt37uSXX34hOjqanJycTt+dHR0dzYoVKyguLmbp0qVMmzbNukq7S0tLo6ioSN6iqKVDN2q1Wg6adXV1F2wVSpLEqVOn5N0WCgoKGt2olpWV8dRTTxEfH8/NN9/M448/3uLX0FIWi4WDBw/KE2Xa4mbeYrFQU1OD2WxGo9E0e/N97nWkLa4pbaHdX4HJZOLkyZO88sor8nZKmZmZfPzxxy3asqMp/v7+DB48mPLy8gtuBGk2m9FqtWg0GrZv386TTz7Jq6++ysqVK3n55Zd57rnniImJYf78+dx///0tSuBbWVlJfHw8tbW1DBw4UL6rudpJ9UlyJatB8UuhUqlwdHTkxx9/5MMPP2T//v1UVFSQlZXF559/zn/+859Wrb3rbhQKBSqVivT0dL788kveeustfv31V3lSVEO3dUxMDBs3biQhIaHduvDbSmFhId9//z2HDh1iwYIFTJ8+/aKJq00mE9HR0Rw6dOiCPUKtZTQa5QkqgYGBrer1sLOzk7svq6ur5TWFTYmLi0Or1eLm5kZ1dTWvvfYaH3/8MStWrODZZ58lPT2dhQsX8vDDDzfaDq4tJCQk8Oabb8pJzWNjY/n555+prKy8rBslk8lEVVUVJpMJJyenZj9v515HGq4rV1qHBMKKigoyMzMZMmQIo0ePZuTIkSQmJl5yTrqQkBCmTJkCwJ9//tnkGGBDxob58+cTERFBbm4uO3fuZO/evezatYvs7GxGjhzJP//5T6ZMmdKigfHi4mIOHjyInZ0dQ4cO7RRN+q5GrVYTGhrKkSNHiI2NRVm/BUx4eDhZWVmsXbuWb775pl3yK14t7OzsMBqNnDx5ki1btrBx48ZGPSOOjo4UFhZSVFSESqXCxcWlTXsI2pLJZOK7775j7969BAQE8Pjjjzc5gaSBxWKhuLiYzz//nB9//BGdTnfZF1Kz2UxmZibFxcVoNJpWTyCxs7OTd6zRarUXTQCg0WgYP348kZGRXH/99SQlJfHbb79x4MAB0tPTmTVrFkuXLm12Rx2TydRoe7SWyMnJISsri7FjxzJ69Gj8/f0pKCho9j00m80X3X7NbDbLG3a7uLi06Zhmh7FeWNhVxMbGSm5ubtLIkSOlH3/80bq4EbPZLFVUVEjZ2dnSqVOnpOLiYsloNFpXa9a3334rqdVqaerUqdLhw4eti69aK1eulEJCQiSNRiMtXrz4kpIYnMtisUgjRoyQxo0bJ+3fv19+/Pfff5fmzp0rhYSEXPbPuJqZzWbpvffek2677TYpIiJC6tGjh5SamiqXV1RUSPfdd5/0yiuvSH/99Vej53YmBoNBOnr0qNSvXz8pODi4RQur9Xq9tH37dqlXr17SzJkzJa1We8mL9xvU1tZKP/zwgxQUFCSFhoZKH330UavOmZqaKj355JOSQqGQHnvssUZ/C2tVVVWSTqeT/63X66X8/HypqKioUb3mnD59Wjp48KB0+vRpyWAwyI+3dEF9S+n1eik7O1vas2fPBV9jcXGx9J///EdycHCQ5s6dKx09etS6SiN//PGHNGnSJAmQJk6cKG3bts26Sodr9xZhe+nduzf/93//R0pKCrt3777oXZhSqcTV1ZWAgAB69+6Nl5dXq2c+Hjx4kPXr16NUKrn//vu7TbdoWzObzaSlpXH27Fl8fHwa7YqtVP5v78mLLYsR/tedlJCQwIABA5g9ezbl5eXExcXJi7BNJhNJSUn06NGjTSZDtJfCwkKeffZZCgsLufPOO3n00Uetq5xHpVKxbds2cnNzcXBwaJP9FA0GAwkJCZSWltKrV69mN+K15ujoiL+/PyqVipKSkosmnXZ2dm40SdDW1hYfH59Wz5lYs2YN9957L3/++We7LmDPycnhs88+47HHHiMxMdG6GOrH/Rtmwnp6el7yvoZXUpcNhJ6enjzwwAOMGDGCqKgo3nnnnQs23RsolUrUanWrvzg5OTl88sknxMfHc//99zN+/PhLWvoh/O9Lk5ycjMFgwN/fv9EFoCGpQZ8+fS5pRnF3oVAoSExMxN3dneuvv56ePXsSExNDXl4e1E+RP3PmjJz/szPKz89n3bp1REdHyzeniYmJxMTENHkcOnSIvXv38t5777Fr1y4sFkubBEHqu1v3799PbW0t/fr1Y+jQoa06r6urK6GhoSiVSgoKCigrK7OuclENN4CtMWzYMG655ZZWTeq5FK6urgwZMoQFCxbg7+9vXQz1NxI5OTlYLBZ8fX0vWK8z67KBUKVS0b9/fx588EGCgoL4+eef+fLLL0lJSbnoYHVraLVajh49ygcffEBycjJjxozhwQcf7JJ3PJ2FwWAgMTERR0dHeblKTU0N6enp/Pbbb9TV1ck7kghNq6qqoqqqCjs7O/r378+IESM4fPiwvF1PUVERGo2mxSnCOprRaOTXX39lzZo11NXVYTab+eOPP3jjjTd49913L3i89dZbrF69Wt4093JvlkwmE+Xl5ezfv58jR44waNAgJk6c2OrA4ujoSEhICLa2tuTn57dqY9rWsFgs1NbWUlZWxpAhQ7j33nsZPHhwo+UKUv1i9bZIfNGwZnro0KE88MADF+xd0Ov1ZGZmolKpztskoavosoGwwfz581m+fDnh4eH84x//4LPPPuPkyZOX3bWm1WqJjY1lxYoVfPPNN0yaNIk333yTfv36tfruTfj/9Ho9iYmJeHh44ODgQHFxMSdOnGDFihX8+eefjB07lnvvvbfD0mp1NXq9nrS0NLy9vXFycsLLy4vJkyeTkpJCdnY21dXVpKam0r9//2Zn710pR48eZc+ePWRlZWFjY0NCQgKbN29m/fr1bNy4sclj/fr17N27l9zcXHlYo2FR9qXKyspiw4YNfPTRR6jVau6//36mTp1qXa1FGro3CwoK5JZ5WzObzRQWFhIdHU1cXByFhYXnXYuUSqWcfau5DC/NSU9PJyoqiiNHjlxwmRr118qsrCwCAwNb1Eiwfs2dQZcPhABjxoxh9erVfPnll9TW1vL777+Tnp5uXa1V4uLi2LlzJ3369GHTpk289NJLF00GfjVrWEPYFmt+jEYjiYmJVFVVERMTw8qVK9mxYwcDBw7k/fffZ+XKlS1eyNwdVVdXk5SURFBQEJ6enmg0Gv72t7+hUqnk5PGnT58mIiKi03aLFhYW4unpyYwZM5gxYwYzZ848b3Pppo6GjapnzJjBLbfcwuDBg61P3WIGg4FPP/2UNWvW0KNHD7766isWLFhwya1oW1tbRo4cSXV1NVlZWdbFbaJhveI333zD6tWr+eOPP6yr4OzszLRp0xg/fvxlrxGsrq5my5YtLFu2jMLCQutiqH8fMzIyqKmpYfDgwRed8Xuuc9cRdobA2OEb87anuro6kpOT0Wg0BAQEXPKHmvrxqry8PFxdXds8u0NX88orr/DRRx9RXl7OTTfdxIoVKwgPD7eu1iLHjx/n2muvZf78+cyZM4eePXuiVCrx8PCgZ8+ezWal6O7OnDnD+++/j6OjI7Nnz2bo0KFUVlZy00034evrS+/evdFqtYwZM0Z+rLPJzc2ltLS02TH9i1EoFPj6+jabDepCzGYzf/75J8XFxfj5+dGvX7/LGvevqanhs88+41//+hejR4/mk08+ITAwsM0v8gaDgTlz5lBeXs7ixYvP2wC3rq6OU6dOYW9vj4+Pz2VdA48fP84bb7zB9u3bOX78eJNdo1lZWXz22We89tpr/Pvf/yYyMrLZYLh7926effZZ4uPjGTJkCM8//3yrNr5uF9bTSIVL30vtavXoo49KdnZ2kkqlkkaMGCElJCRYV2kRnU4n/fzzzxIgffLJJ+J9vgRJSUnS9OnTpS+//FIqKCiQpPop7suWLZMmT54sDRkyRJo6daoUExMj1dXVWT/9qtNZPkMGg0H666+/pAEDBkh9+/aVvv/++3Z5bRaLRZowYYI0Z84cKTo62rq4TcXGxkp33XWXNGzYMEmr1VoXS5IkSfv27ZOuv/56yc7OTtq+fXuL9jXcvHmz5OfnJwGSv7+/9PHHH1tX6XDdt5lzEW19F9fVabVadDodFotFXjh7KUpKSkhOTsbW1pZevXqJ9/kSVFdXk5KSQq9eveTxGKVSyfXXX4/JZOL48ePyfnrNpbq6GnSWz1BDHt4+ffpQWFjIb7/91i7ZkcrKysjOzpbTwLWn1NRUKioqGDx48AXf56SkJI4dO0ZERAR9+vRpUc+ZXq+nqqoK6q8tFxt/7CjNv2qh2zv3S3ChL0RLlJSUcOrUKfz9/Tvt+FVnV15eTklJCd7e3vJFR6VSMWHCBDm7ibOzsxhnvUJuuOEGfHx8iIqKkpNrtxWdTseJEyfQ6XQEBAS0eu1ha2VkZFBRUcHAgQObnHiTn5/Pn3/+iV6v5/bbb2/VdlEN15HOMkYoAqHQIcrKyvj111/ZsGEDmZmZ/Pnnn/ICcKFloqKiWLVqFVqtlpiYGPn9UygU2NvbExISQs+ePQkMDGzywiW0v5tvvplrr72W06dP88UXX1BeXg71yxouV21tLUlJSXh5eXXIWr3Tp09TXV3NoEGD5N6Fc3+PjRs3EhUVRVBQELfcckunnaXcEiIQCh2mT58+3HXXXTz22GOEhoZe9hKX7sbBwYExY8bw8MMP06tXr/PevxkzZvD888+zYMGCJtfCVVVVUVhY2Gi3Cr1eT1lZGfn5+ZSVlWE0GrFYLGi1WgoLC8nJyaGiogJJkigtLSU3N5f8/HxxE3MBvXr14sYbbyQiIoJ169YRFRWF2Wxuk1ZPXV0dKSkpBAQE4OfnB/WTVdqjC1av15Ofn49Go6Ffv36kpaWh1Wrl3+PYsWN88803SJLEggULCAsL69I3XyIQCh3C09OTefPm8cknn/Duu+8ybdo00T3aCpIkMWrUKF555RXef/99brzxxkbvnyRJTJ48mUceeYTZs2c3Ofv27NmzHDhwgM2bN5OXl4fFYsFsNpOens4vv/xCfn6+HCDLy8vZsmULBw4ckGd4Hjt2jF27dvHXX3/JYzwtZbFYMBqNmEymNmkdWWtIQn2p49dtacKECURGRlJUVMS6devYtWtXm6RBa9jAQKFQUFhYyPHjx0lPT2/zQChJEpWVlfLfq7CwUP4cUL+Lxdtvv01aWho33HADixcvtj5FlyMCoSB0Ac21KJorp37D5z/++IN//OMfbN++naqqKhwcHKitreWjjz7C3t4ejUaDUqnEycmJ9957j1OnTuHq6orBYCA+Pp49e/ZQWlra6r04KyoqSExMJCcnp822TjpXcnIyx44dIz8/37qow/n5+bFo0SIiIyPZu3cvL774Ihs3brSu1mr29vaEh4fLgWjjxo0MGzasRTvntIZCocDLy4tevXpx9OhRHn/8caqqqnB3d+fw4cMsXbqU77//npkzZ3Lfffe1+rPQGYlAKHRqDSmjqO8aOneGWXV1daN/n9vSqKysJD8/X06319BSqKuro7q6utFzdDpdm6Xl68zKy8sJDAwkIiKCnTt3yltdKZVK+vbt22gdXUNXaGhoKD169CApKYmsrCzuvPNObrnllnPO2jJ79+7ljjvu4Pfff2/0/l+uvLw8li9fzl133cXatWtb3VJtL35+fjz//PM8+eST2NvbExUVZV2l1Tw8PFiyZAlr167liy++4O9//ztubm4tuglqLaVSyfPPP8+OHTt45513mD9/Pm5ubpw8eZKkpCQef/xxnnrqKUaMGGH91C5JtXz58uXWDwrCuX755Rfi4+NRKBR4eHgwf/78dl+oXVNTQ2JiIvv27ePAgQNQv8A3JiaGxMREtFotcXFxHDlyhJKSEsLCwlAoFOTm5hIbG0tqaiqlpaX8+uuvnDx5kkGDBlFWViYncc7JycHb25stW7Zw4sQJPDw82u2i0lkcOXJETl6wb98+Zs+ejb29PWlpaajVasaMGYO9vT2SJJGdnc2uXbuYOHEiKpWKuLg4hgwZwt/+9rdLygNbVFREcXExEydOJCgo6LLzhObn5xMbG8u+ffvIzs5mx44duLm5MWXKlHb/bLaUq6srwcHBBAQEEBoaysCBA62rtIpSqcTFxYXQ0FBCQkJwc3Nr0XKFS+Xt7U1oaCiBgYFyysOGZPm33347gwYNQqlUIklSq743SUlJbNmyBYPBgJ2dHZMmTWLMmDHW1TpU+72LwlXj3JZWe4zvWJMkCb1eT2lpKb///ruchzQvL489e/awYsUK4uPjycjIYPfu3WzatEl+XQ27bpeWlmIwGPjxxx9Zu3YtKpWKqqoqysvLiY2N5ZNPPiE2NpZjx46RlZXVJmM4nd3JkyfRaDRMmDABBwcHcnJyOHr0KMXFxQwcOFAOThUVFaSmphIQEEB8fDzbt2+ntraWWbNm4ejo2KrPgMFgQKvVEhYWxj/+8Q+uu+66Swqk1kpLS0lJSaG4uJjJkycTEBCAWq3udBM2evfuzcKFC6985pQ20DBO/dRTTzXaPq01QbBBw2dI6i471Atd37l3ne15B9qgoeU5depUpk2bRl1dHUOHDmXx4sWMGTMGrVbLzTffzN13383YsWOxsbFBoVCQlpZGQkICLi4u/P3vf2f69OmMHj2a3r17o1arCQ8P584772TWrFmkpKSwc+dOlixZwj//+U8iIiIu6QvdlZw5cwadTsfgwYOZNm0aiYmJ7N27V14rdm4gTE5OxsHBgd27d/PXX38RHh4ud1G39H1q2E3+zz//5MiRIxQVFTX6/DRcBFtzNOjTpw8PPvggb731FjfeeCMuLi7tNhFH+J+W/t1borPlGm3/q5rQ5YWFheHv7y9nz2iLO/qWOHv2LFVVVYwYMYK+fftSUlKCJEnyHnzZ2dkYDAYGDhyIQqEgKiqKuro6eWPVv/76C5VK1ejulfrte1xdXbnppps6ZD1WZ2E0GjEYDHh6ejJlyhROnDhBTEwMdXV1BAUFya2psrIy0tPTmTJlCsuWLSMoKIgffvih1UFGoVDg4uLCr7/+yooVK9i+fbs8VpuXl8eWLVt49913WbVqFatXr77gsWrVKtasWcORI0fkc2s0GnmJSGe4kAot16NHD/r37w9AREQEvXr1sq7S4UQgFJrl6OiInZ2dPEbRsA1Oe8vIyKCsrIwRI0bg6upKWloaNTU1jBo1Cjs7O+Lj4zEYDPL4Qnp6ujzxA2DTpk3odDqGDBkin7OoqIjk5GR8fX1RKBStvrh3VTqdrlHGmWuuuYaSkhJyc3PPW+dWWVlJSkqKvPlrUFAQiYmJrd5nryEQarVaysvLG81u1Ol0FBYWkpmZ2exx+vRpzpw5Q0VFRaNzN7BeTyl0bra2tvIWWo6Ojm0+6/VSiEAoNMtsNmOxWJAkqUO7n1JTUykoKGDIkCHyvnXl5eXypIO8vDwKCwupqamhrKwMe3t7dDodSUlJHD9+XN6fr6amhtzcXA4fPsyuXbvQ6XQMGzaMffv2cfDgwat+cXhdXR0xMTFotVr5JsbDwwNPT0/c3d3PuxClpqZy8uRJ7OzssLW1xcfHB71ez5YtWy4pbZjRaMTDw4Phw4fLrU4XFxf69evHuHHjmj3Gjx/PiBEjmt3VQOgaGtavcs615UoTgVDotIqLi9Hr9fTr1w/q97GTJElONtyjRw/KysrYtWsXmZmZjBo1CqPRyFdffcWOHTvkSSEHDx7k4MGDPPLIIyQmJnLDDTcwc+ZMDh8+zKZNm0hMTLT6yVeXmpoadu3add4au4kTJzJ58uRGXVMZGRmkpKRAfdc09ZvOBgUF8d1335GamirXbQmz2Ux2djYWi4UBAwbIgbBHjx5cd9113Hbbbdx+++0sXLjwgsftt9/O7Nmz5e40QWhrV9V+hEL7WLVqFf/973/Jz89n0aJFvPTSSwQHB1tXa3MN2S1cXV2xsbGhvLwcSZJwd3dHqVRSUVFBXV0dNjY2ODs7Y7FYqKysRKfT4erqilqtltcHOjg4UFhYiIuLC25ubkiSREFBAfb29ri6ul7VOzWYzWZ51xAnJyd5vaBWq8VkMmFrays/ZjQaqayspKamBh8fH+zs7NDr9VRUVGA2m5tsQV6I2WwmOTmZyMhI/Pz82LZtm1zWMA5ZVFSE1Mz0e4vFgoODA8HBwU3uuFBeXs6ECRPw8/Nj5cqVl7Vhr9D+9u/fz8svv0xUVBQTJ05k6dKl3HzzzdbVOpRoEQqdkiRJuLq64uXlJU+KaOjOa5hx5ubmhp+fH15eXmg0Guzt7fH19SU4OBh3d3ecnZ3x9vbG29sbJycnwsLC8Pb2xtbWFo1GQ3BwsHyxv5qpVCo8PT3x8fFptPzB2dkZd3f3RgvpbWxs8PLyIigoSH5fNBoNPj4+9OzZs8VBkPpAmJiYiFqtJigoqFFZeXk5cXFxbN++na1bt7Jt27YLHlu3bmXv3r2kp6c3OkeDhiDaWWYgCl2PCIRCpyQuaO2no95bs9lMSkoKLi4u8u7mhYWFmM1mHB0d8ff3Jzw8nD59+lzwCA8Pp2/fvvTu3fuC2/w03BhJktQhy3uEq4/41AiC0C4kSaKkpAS1Wo3FYuH06dPEx8dTVlZGz549ueWWW3jmmWdYtmwZzzzzTJPH0qVLWbZsGQ888ECj7COSJGE0Gjl79iwZGRnodDq0Wi1nzpyhqqpK3kVDEFpCBEJBENqFUqmkd+/enD59mpUrV/LWW28RFhaGp6enddVWq6mpYefOnTzyyCPcddddnD59mri4OJ588kmeeeYZdu7cSVlZmfXTBKFJIhAKgtAubG1tmTdvHm+88QYrVqxg8eLFBAQEtEkaNDs7O4YOHcr999/P66+/zqZNm/jxxx95++23iYyMZOjQoR2W+EHo+kQgFAShXSiVSoKDg5k7dy533nmnnAihLajVagIDA7nuuuuYOXMmc+fO5dZbb2XmzJmMGzeOwMDANvtZwtVPBEJBEAShWxOBUBAEQejWRCAUBEEQujURCAVBEIRuTQRCQRAEoVsTgVAQBEHo1kQgFARBELo1EQgFQRCEbk0EQkEQBKFbE4FQEARB6DAdtftJa4hAKLSYQqHA1tYWFxcX6yKo3xFAEASBi1wPHB0dUavV1g9fUWKHeqFZr7/+OqtWraKkpAQ/Pz9GjRqFh4cHRqMRhUKBp6cnCxcuZPTo0dZPFQShG/vhhx/YuXMnBoMBhUKBjY0NWVlZxMbGUldXx/Dhw3nxxReZM2eO9VM7lGr58uXLrR8UhHOVlZWhVqvx8PDA29ub6upqysrKOHv2LJWVlZjNZoYOHSpvvioIggBw8OBBYmJiKC0tpaKiQt4aKyQkhD59+nDttdcyatQo/P39rZ/aoUSLUGgzkiR1yv5/QRCEixFjhEKbEUFQEISuSARCQRAEoVsTgVAQBEHo1kQgFARBELo1EQgFQRCEbk0EQkEQBKFbE4FQEARB6NZEIBQEQRC6NREIBUEQhG5NBEJBEAShWxOBUBAEQejWRCAUBEEQujURCAVBEIRuTQRCQRAEoVsTgVAQBEHo1kQgFARBELq1/weW6wpFi/1YUgAAAABJRU5ErkJggg==\" style=\"width: 356px; height: 106.8px;\" width=\"356\" height=\"106.8\"\u003e\u003c/p\u003e\u003cp\u003eWhere \u003cem\u003eI\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e represents the set of \u003cem\u003eK\u003c/em\u003e\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e\u003c/sub\u003e interacting elements and \u003cem\u003ef\u003c/em\u003e are uniformly distributed contribution functions. For simple landscapes \u003cem\u003eK\u003c/em\u003e\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e\u003c/sub\u003e = 0, meaning \u003cem\u003eI\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e is empty and each decision contributes interpedently. For complex landscapes \u003cem\u003eK\u003c/em\u003e\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e\u003c/sub\u003e = 7, meaning each decision's contribution depends on 7 randomly assigned interacting decisions, creating rugged fitness landscapes with multiple local optima. Following Baumann et al. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), payoffs are normalised by \u003cem\u003eP\u003c/em\u003e\u003csub\u003e\u003cem\u003emax\u003c/em\u003e\u003c/sub\u003e and raised to power 8 to create sparse high-quality solutions. \u003cem\u003eK\u003c/em\u003e\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e\u003c/sub\u003e = 7 represents approximately half of all decisions being interdependent, which provides a threshold for meaningful complexity while also balancing computational requirements which grow as \u003cem\u003eK\u003c/em\u003e\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e\u003c/sub\u003e approaches N\u003csub\u003e\u003cem\u003eNK\u003c/em\u003e\u003c/sub\u003e (Baumann et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Network Structure\u003c/h2\u003e\u003cp\u003eCollective problem-solving processes are known to be strongly influenced by communication networks (Lazer, D. \u0026amp; Friedman, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Barkoczi, D. \u0026amp; Galesic, M. 2016; Centola, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). To capture this effect agents are embedded in undirected Erdős\u0026ndash;R\u0026eacute;nyi (ER) random networks (Erdős, P. \u0026amp; R\u0026eacute;nyi, A., 1959) with varying densities following the approach of Baumann et al. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). ER networks provide a baseline topology with homogeneous structural properties (Newman M. E. J., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Baumann et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), allowing density effects to be examined independently of confounding topological features such as clustering or degree heterogeneity. The model implementation allows straightforward selection of alternative network topologies (ER, small-world, Barab\u0026aacute;si-Albert, etc.) via configuration parameters, enabling future work to systematically examine alternative structural assumptions. Networks are generated at initialisation before agents are randomly assigned to network positions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Agent Diversity\u003c/h2\u003e\u003cp\u003eThe n\u0026thinsp;=\u0026thinsp;100 agents are partitioned into \u003cem\u003eG\u003c/em\u003e groups of equal size (see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e for group configurations). Each group g has a distinct payoff function P\u003csub\u003eg\u003c/sub\u003e(\u003cem\u003eS\u003c/em\u003e) generated by unique contribution functions f\u003csub\u003eg\u003c/sub\u003e(b\u003csub\u003ei\u003c/sub\u003e |\u003cem\u003eI\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e)\u0026thinsp;~\u0026thinsp;Uniform(0,1). For the same solution \u003cem\u003eS\u003c/em\u003e, agents in different groups extract different payoffs: P\u003csub\u003e\u003cem\u003eα\u003c/em\u003e\u003c/sub\u003e(\u003cem\u003eS\u003c/em\u003e)\u0026thinsp;\u0026ne;\u0026thinsp;P\u003csub\u003eβ\u003c/sub\u003e(\u003cem\u003eS\u003c/em\u003e) for α\u0026thinsp;\u0026ne;\u0026thinsp;β. This captures how different skills, perspectives, or functional roles influence performance evaluation in collective systems. Agents are evenly distributed across groups, and network attachment probability is independent of group membership.2.5 Agent Behaviour\u003c/p\u003e\u003cp\u003eAt each timestep t \u0026isin; {1,...,T}, agents synchronously update through a two-stage process. First, agents attempt social learning by adopting the best neighbour solution if it improves their group-specific payoff (referred to below as social learning events). If no improvement is found through social learning, agents attempt innovation by randomly flipping one bit in their binary solution vector and adopting the new solution if it yields improvement (innovation events). If neither social learning nor innovation yields improvement, agents retain their current solution. This implements satisficing behaviour balancing exploitation and exploration.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Disruption Mechanism\u003c/h2\u003e\u003cp\u003eDisruption regimes are modelled through stochastic perturbations to fitness landscapes, characterized by three parameters with specific values provided in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Disruption frequency ω determines the probability at each timestep that a disruption occurs, following a Bernoulli process. Moderate impact scenarios reflect typical historical disruptions, while severe scenarios represent the amplified disruptions projected under polycrisis conditions (Lawrence M., et al. 2024)(see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The shape parameters α and β for the Beta distribution are determined by minimising the squared error between target percentiles and those produced by candidate parameter values, using Nelder-Mead optimisation. Beta distributions provide flexible bounded distributions on (0,1) suitable for modeling proportional impacts and are widely used in risk analysis such as earthquake impact modelling (Vose, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Lallemant \u0026amp; Kiremidjian, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). The heavy-tailed parameterisations capture patterns where most disruptions have moderate impacts while severe impacts occur with non-negligible probability (Clauset et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Helbing \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) (See Supplementary Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e-S3 for disruption distributions and characteristics).\u003c/p\u003e\u003cp\u003eThe distribution of impact magnitude across groups follows an exponential decay τ\u003csub\u003eg\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;τ\u0026middot;e\u003csup\u003e\u0026minus;λ(g\u0026minus;1)\u003c/sup\u003e, where λ determines the degree of inequality in impact distribution across groups. When λ\u0026thinsp;=\u0026thinsp;0, all groups experience identical impact. When λ\u0026thinsp;\u0026gt;\u0026thinsp;0, impacts follow an exponential gradient creating substantial differences between most and least affected groups. This captures how distinct functional units of the collective (e.g. organisational departments) are likely to have differential exposure to any given disruption. For each disruption, vulnerability rankings are randomly permuted across groups. Total disruption magnitude is conserved for each disruption event.\u003c/p\u003e\u003cp\u003eWhen disruption occurs, each group's contribution functions are modified through linear interpolation following Siggelkow and Rivkin (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2005\u003c/span\u003e):\u003c/p\u003e\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" style=\"width: 379px; height: 69.3113px;\" width=\"379\" height=\"69.3113\"\u003e\u003c/p\u003e\u003cp\u003ewhere u\u003csub\u003eg,i\u003c/sub\u003e ~ Uniform (0,1) is independently sampled for each group and contribution function. Each of the G groups has its contribution functions updated according to its group-specific impact magnitude τ\u003csub\u003eg\u003c/sub\u003e. For K\u003csub\u003eNK\u003c/sub\u003e = 0, each group has 30 contribution values corresponding to N\u003csub\u003eNK\u003c/sub\u003e positions and 2 states per position. For K\u003csub\u003eNK\u003c/sub\u003e = 7, each group has 3,840 contribution values corresponding to N\u003csub\u003eNK\u003c/sub\u003e positions and 2^(K\u003csub\u003eNK\u003c/sub\u003e+1) sub combinations per position. The interaction matrix M defines the epistatic relationships between positions, the specific pattern of interdependencies where each position's contribution to the payoff depends not only on its own state but also on the states of K other positions. M is preserved across disruptions, reflecting that perturbations typically alter performance within existing structural constraints rather than fundamentally rewiring interdependencies (Siggelkow \u0026amp; Rivkin, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Scheffer et al., 2012).\u003c/p\u003e\u003cp\u003eFrequency and magnitude are modelled as independent factors to enable systematic factorial exploration of how collective configurations perform under varying disruption characteristics. This represents a simplification as many natural hazards and disruption events such as blackouts exhibit frequency-magnitude relationships (Newman, M.E.J., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Carreras, B.A., 2004; Becerra O., et al., 2006), though we believe is suitable to enable streamlined investigation of pure frequency impact size effects.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.6 Performance Measurement and Experimental Design\u003c/h2\u003e\u003cp\u003eCollective performance at time t is measured as median payoff across all N\u0026thinsp;=\u0026thinsp;100 agents, P\u003csup\u003e~\u003c/sup\u003e(t)\u0026thinsp;=\u0026thinsp;median {P\u003csub\u003eg\u003c/sub\u003e(S\u003csub\u003ei\u003c/sub\u003e(t)): i\u0026thinsp;=\u0026thinsp;1,...,N}. Cumulative performance Π=\u0026sum;\u003csup\u003eT\u003c/sup\u003e\u003csub\u003et=1\u003c/sub\u003e P\u003csup\u003e~\u003c/sup\u003e(t) aggregates sustained functionality, capturing resilience as persistence of function under disruption (Folke, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2006\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWe employ a full factorial design crossing network density, agent diversity, landscape complexity, and disruption regimes (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Network density and agent group values use denser sampling at lower values where theoretical predictions and prior empirical findings (Lazer \u0026amp; Friedman, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Baumann et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) indicate that marginal effects are expected to be larger. This parameter spacing focuses simulations on areas where qualitatively important transitions occur i.e. from homogeneous to diverse collectives as the first few groups are added (G\u0026thinsp;=\u0026thinsp;1\u0026ndash;3 groups) and sparse to intermediate connectivity (d\u0026thinsp;=\u0026thinsp;0.04\u0026ndash;0.12) while limiting the factorial to a 7\u0026times;7 configuration for computational tractability. We run simulations for all diversity and density parameter combinations with no disruptions which functions as a control to compare impacts of disruptions against. Each parameter combination is replicated 83 times, a total of 54,544 simulations. This sample size was the result of successive sampling rounds (3,30,50 replicates) to ensure sufficient statistical power for main effects while balancing computational requirements.\u003c/p\u003e\u003cp\u003eMedian cumulative performance across replicates for a given parameter set is reported with bootstrap confidence intervals (5,000 samples) used to quantify uncertainty in median estimates. Wilcoxon rank-sum tests are used to assess cross-landscape differences.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSimulation parameters and factor choices for disruption regimes.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eParameter\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003eValues\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTimesteps (t)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e400\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAgents (n)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNetwork Density (d)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e0.04, 0.08, 0.12, 0.20, 0.32, 0.52, 0.76\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAgent Groups (G)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e1, 2, 3, 5, 8, 13, 21\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLandscape Complexity (K)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e0, 7\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u003cp\u003e\u003cb\u003eDisruption Regimes\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003eLow Factor\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003eHigh Factor\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFrequency (ω)\u003c/p\u003e\u003cp\u003e(infrequent, frequent)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1/100\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1/25\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eImpact Size\u003c/p\u003e\u003cp\u003e(moderate, severe)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5th % = 0.001\u003c/p\u003e\u003cp\u003e95th % = 0.5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5th% = 0.001\u003c/p\u003e\u003cp\u003e95th % = 0.65\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eImpact Inequality (λ)\u003c/p\u003e\u003cp\u003e(uniform, skewed)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.5\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Results","content":"\u003cp\u003eWe examine how network connectivity and agent diversity influence collective performance under disruption across simple (K\u003csub\u003eNK\u003c/sub\u003e = 0) and complex (K\u003csub\u003eNK\u003c/sub\u003e = 7) landscapes. Our core finding is that landscape complexity determines whether disruptions degrade or enhance collective performance, with simple and complex landscapes exhibiting markedly different responses to identical disruption regimes. We first present this cross-landscape comparison, then detail the mechanisms and configuration effects for each landscape type. All cumulative performance values are reported as medians with 95% bootstrap confidence intervals unless otherwise noted.\u003c/p\u003e\u003cp\u003eDisruptions trigger immediate performance changes that vary systematically by landscape complexity (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). In simple landscapes, disruptions consistently reduce performance by interfering with convergence toward the global optimum. Notably, in complex landscapes, disruptions occasionally produce immediate performance increases when landscape changes allow agents to escape local optima where they were previously trapped. This \"creative destruction\" phenomenon is well-documented in search processes and complex adaptive systems (Kauffman \u0026amp; Weinberger, 1989; Allen et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Cross-Landscape Analysis: Complexity Moderates Creative Destruction\u003c/h2\u003e\u003cp\u003eComparative analysis reveals disruptions produce opposite performance effects depending on landscape complexity (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Simple landscapes (K\u003csub\u003eNK\u003c/sub\u003e = 0) show small negative reductions in median cumulative performance across all disruption regimes, with median performance declining between \u0026minus;\u0026thinsp;0.06% and \u0026minus;\u0026thinsp;0.75% relative to the control (no disruptions). In contrast, complex landscapes (K\u003csub\u003eNK\u003c/sub\u003e = 7) show substantial positive effects across all disruption regimes, with performance increasing between 5.62% and 16.79% relative to control. This occurs under identical disruption mechanisms and is highly significant (Wilcoxon rank-sum test, W\u0026thinsp;=\u0026thinsp;0.0, p\u0026thinsp;=\u0026thinsp;0.0078, comparing disruption effects between landscapes across eight regime combinations).\u003c/p\u003e\u003cp\u003eImpact inequality (skew) effects only matter in complex landscapes for severe disruptions (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). In simple landscapes, uniform versus skewed distribution produces nearly identical outcomes within each disruption regime. In complex landscapes, skew appears to only become relevant for severe disruptions irrelevant of frequency. The impact is most pronounced under frequent-severe conditions, uniformly distributed impacts yields a median performance effect of 16.79% (95% CI: [15.67%, 17.89%]) compared to 7.21% (95% CI: [5.71%, 8.74%]) under skewed impacts, a median difference of 9.58 percentage points.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe median cumulative performance ranges conditions differ dramatically between simple and complex landscapes. Median cumulative performance in simple landscapes span 0.036 units across all disruption conditions ranging from 0.958 (G\u0026thinsp;=\u0026thinsp;1, d\u0026thinsp;=\u0026thinsp;0.08, [frequent, severe, skewed]) to 0.994 (various disruption regimes) (Supplementary Figure S4). In complex landscapes the median cumulative performance spans 0.435 units, ranging from 0.325 (G\u0026thinsp;=\u0026thinsp;13, d\u0026thinsp;=\u0026thinsp;0.04, [infrequent, severe, skewed]) to 0.760 (G\u0026thinsp;=\u0026thinsp;21, d\u0026thinsp;=\u0026thinsp;0.052, [frequent, severe, skewed]), a ten-fold difference (Supplementary Figure S5). This reflects that simple landscapes operate near ceiling performance with limited room for variation, while complex landscapes pose greater collective learning challenges where collective configuration choices meaningfully impact outcomes.\u003c/p\u003e\u003cp\u003ePerformance difference heatmaps (disruption minus control, Supplementary Figures S5-S6) reveal distinct mechanisms that increase resilience to disruptions. In simple landscapes, disruption effects concentrate on low-diversity configurations regardless of density: at G\u0026thinsp;=\u0026thinsp;1\u0026ndash;2, severe disruptions reduce performance by 3\u0026ndash;4 percentage points, but by G\u0026thinsp;=\u0026thinsp;5, disruption effects become negligible across all density levels (Figure S5). This demonstrates that diversity, not density, provides most of the resilience. In complex landscapes, the patterns are much more complex. Positive effects appear in diverse configurations including, both high-density-low-diversity (G\u0026thinsp;=\u0026thinsp;1, d\u0026thinsp;=\u0026thinsp;0.76) and low-density-high-diversity (G\u0026thinsp;=\u0026thinsp;21, d\u0026thinsp;=\u0026thinsp;0.04) combinations. Notably, high diversity (G\u0026thinsp;=\u0026thinsp;21) shows consistently positive or neutral effects across all density levels and disruption regimes, indicating robust benefits of maintaining diverse perspectives. The heterogeneity in difference patterns across regimes suggests that disruption-configuration interactions warrant further investigation to identify specific mechanisms underlying performance gains.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThese patterns reflect fundamentally different adaptation challenges. In simple landscapes, disruptions interfere with convergence toward a single global optimum, though high baseline performance limits absolute impacts. In complex landscapes, disruptions enable escape from local maxima, facilitating exploration of superior solution regions. Uniform distribution amplifies this creative destruction effect by preserving diverse perspectives needed to navigate rugged landscapes, while concentrated impacts eliminate specific viewpoints from the collective search process.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Simple Landscapes (K\u0026thinsp;=\u0026thinsp;0)\u003c/h2\u003e\u003cp\u003eNetwork density and agent diversity serve distinct functions in simple landscapes (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eUnder control conditions, performance peaks 0.994 at high density and low diversity (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). Under disruptions, collectives with greater diversity (G\u0026thinsp;\u0026gt;\u0026thinsp;3) and density (d\u0026thinsp;\u0026gt;\u0026thinsp;0.2) performing better. Density appears to drive baseline performance through rapid information diffusion, exhibiting a logarithmic-like increase with steep gains below d\u0026thinsp;=\u0026thinsp;0.2 and a plateau beyond (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). Diversity provides minimal baseline benefits beyond G\u0026thinsp;=\u0026thinsp;2\u0026ndash;3. This pattern replicates Baumann et al. (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and reflects that dense networks enable rapid diffusion of the single global optimum in simple landscapes (Centola, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), while diversity provides minimal additional benefit navigating the smooth landscape.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe diversity-performance relationship shows threshold behaviour: in response to disruptions, performance jumps from G\u0026thinsp;=\u0026thinsp;1 up to G\u0026thinsp;=\u0026thinsp;3, after which additional diversity provides less value. This is most prominent under frequent-severe disruptions where performance jumps from a median\u0026thinsp;\u0026asymp;\u0026thinsp;0.964 (G\u0026thinsp;=\u0026thinsp;1) to \u0026asymp;\u0026thinsp;0.991 (G\u0026thinsp;=\u0026thinsp;3) then saturates through G\u0026thinsp;=\u0026thinsp;21 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD). This threshold reflects the transition from complete homogeneity, where all agents seek identical solutions and disruptions affect everyone similarly, to diversity where additional payoff functions provide redundancy.\u003c/p\u003e\u003cp\u003eCritically, diversity\u0026thinsp;\u0026ge;\u0026thinsp;5 largely eliminates disruption effects entirely. Performance difference heatmaps (Supplementary Figure S6) show that severe disruptions reduce performance by ~\u0026thinsp;1.0\u0026ndash;3.3 percentage points at G\u0026thinsp;=\u0026thinsp;1\u0026ndash;2 but produce near-zero effects at G\u0026thinsp;\u0026ge;\u0026thinsp;5, regardless of network density. Additional diversity beyond G\u0026thinsp;=\u0026thinsp;5 provides neither baseline gains nor resilience benefits, as sufficient redundancy has been achieved.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Complex Landscapes (K\u0026thinsp;=\u0026thinsp;7)\u003c/h2\u003e\u003cp\u003eThe complex landscape requires fundamentally different collective properties to achieve good cumulative performance compared to simple landscapes (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e5\u003c/span\u003e.). Under control conditions, performance peaks 0.577 at moderate density and high diversity (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). Under disruption, performance increases substantially across most configurations, with roughly the same combinations favoured as the control, with a cumulative performance peak of 0.677 (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e5\u003c/span\u003eB).\u003c/p\u003e\u003cp\u003eThe density-performance relationship exhibits an inverted-U with peaks around d\u0026thinsp;=\u0026thinsp;0.2\u0026ndash;0.32, declining at both low (d\u0026thinsp;\u0026lt;\u0026thinsp;0.12) and high (d\u0026thinsp;\u0026gt;\u0026thinsp;0.5) density (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e5\u003c/span\u003eC). This non-monotonic pattern persists across all disruption regimes and suggests that moderate connectivity balances information sharing benefits against costs of excessive coupling. Sparse networks limit collective learning, while dense networks may promote premature convergence or rapid spread of solutions reducing effective diversity limiting future innovation (Levinthal \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1997\u003c/span\u003e; Lazer \u0026amp; Friedman, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Centola, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This type of convergence is also likely to reduce resilience and adaptability to disruptions due to limiting diversity of unique solutions leading to greater effective heterogeneity of agents known to increase potential for critical transitions (Scheffer 2012). This inverted-U persists across all disruption regimes and is also present in the control, with uniform distribution outperforming skewed at all density levels, except infrequent-mild which is roughly equivalent.\u003c/p\u003e\u003cp\u003eDiversity increase performance gradually through G\u0026thinsp;=\u0026thinsp;8\u0026ndash;13 (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e5\u003c/span\u003eD), contrasting with simple landscapes' sharp threshold and plateau (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e4\u003c/span\u003eD). Across the regimes shown, performance continues increasing through the highest diversity level examined (G\u0026thinsp;=\u0026thinsp;21), with different disruption regimes achieving different absolute performance levels.\u003c/p\u003e\u003cp\u003ePerformance difference heatmaps (Supplementary Figure S7) show that disruptions are broadly beneficial across complex landscapes, though patterns are heterogeneous and complex. High diversity (G\u0026thinsp;=\u0026thinsp;21) shows consistently positive or neutral effects across all density levels and disruption regimes. The varied spatial patterns of positive effects across different regime heatmaps suggest that multiple configurations may achieve performance gains through different mechanisms, warranting further investigation of potential mechanism.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4 Discussion","content":"\u003cp\u003eOur analysis reveals that landscape complexity fundamentally determines whether disruption enhances or degrades collective performance. Simple landscapes (K\u0026thinsp;=\u0026thinsp;0) demonstrate universal performance degradation under disruption (performance effects ranging from \u0026minus;\u0026thinsp;0.06% to -0.78%), while complex landscapes (K\u0026thinsp;=\u0026thinsp;7) show universal performance enhancement under disruption (median effects ranging from 5.78% to 16.79%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The difference in effects between landscapes is significant (Wilcoxon W\u0026thinsp;=\u0026thinsp;0.0, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and challenges assumptions about disruption effects suggesting that the relationship between disruption and collective performance is highly dependent on the environment collectives must navigate.\u003c/p\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e4.1 Extending Collective Intelligence Frameworks\u003c/h2\u003e\u003cp\u003eThe reversal between simple and complex landscapes extends existing collective intelligence frameworks in important ways. While previous work has shown that network structure moderates diversity effects in static environments (Baumann et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), our results demonstrate that disruptions fundamentally alter these relationships. In simple landscapes, disruptions interfere with convergence toward global optima, confirming that volatility impedes performance when solutions are straightforward. However, in complex landscapes, disruptions facilitate escape from local optima, functioning similarly to simulated annealing by facilitating exploration of solution spaces.\u003c/p\u003e\u003cp\u003eThis finding aligns with theoretical predictions for rugged fitness landscapes (Kauffman \u0026amp; Weinberger, 1989; Weinberger, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e1991\u003c/span\u003e) and provides computational support for the collective adaptation framework's proposition that collectives must match their structure to problem characteristics (Galesic et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The creative destruction mechanism we observe, where disruptions enable performance gains in complex environments, has been theorised in socio-ecological systems (Allen et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) but with limited observations in computational models and in collective learning contexts.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e4.2 Network Structure and Resilience Mechanism\u003c/h2\u003e\u003cp\u003eFor complex landscapes, network connectivity exhibits an inverted-U relationship with performance (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e5\u003c/span\u003eC), aligning with patterns observed across diverse complex systems from supply chains (Chen \u0026amp; Wen, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) to infrastructure networks (Brummitt et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Sparse connectivity limits information diffusion necessary for collective learning, while excessive connectivity promotes premature convergence or rapid spread of suboptimal solutions. Critically, this inverted-U persists under disruption, demonstrating robustness of the connectivity optimum across environmental regimes and providing empirical grounding for theoretical predictions about connectivity thresholds in complex adaptive systems (Gao et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe difference between disrupted and control conditions (Supplementary Figures S6-S7) illuminates distinct resilience mechanisms across landscape types. In simple landscapes, network density optimizes baseline performance by enabling rapid diffusion of the single optimal solution, but diversity provides resilience by maintaining redundant solution approaches that buffer against disruption. Once sufficient diversity is achieved (G\u0026thinsp;\u0026ge;\u0026thinsp;5), additional diversity provides neither baseline gains nor resilience benefits, as adequate coverage of the solution space has been attained. In complex landscapes, both connectivity and diversity contribute to resilience, with moderate connectivity (avoiding both under-connection and over-coupling) and high diversity (maintaining multiple perspectives for navigating local optima) required for maximal disruption-enabled performance gains. The role of distributional equity emerges clearly in complex landscape difference maps, where uniform distribution amplifies positive effects throughout the parameter space.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e4.3 Distribution of Impacts and Collective Performance\u003c/h2\u003e\u003cp\u003eThe 4\u0026ndash;7 percentage point performance advantage of severe-uniform over severe-skewed impacts indicates that equality of impacts matters for maintaining collective performance (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.). When certain groups face disproportionately severe impacts it is possible for skewed distributions to become τg\u0026thinsp;=\u0026thinsp;1 for certain groups (see Supplementary Information figures S 2\u0026ndash;3). This results in a completely new contribution function being redrawn from the uniform distribution which effectively randomizes the position on the landscape of agents in those groups. In the severe skewed disruptions this likely results in agents being unable to contribute meaningfully to collective learning, functionally reducing the diversity that is critical for performance in complex landscapes (Baumann et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Centola, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Hong \u0026amp; Page, 2004). This implies that minimising potential (or at least managing) potential for highly unequal impacts in turbulent complex environments can have benefits for collective performance and adaptability. Future work should investigate threshold effects and how network homophily might amplify inequality effects by limiting information flow between differentially affected groups.\u003c/p\u003e\u003cp\u003ePractically, this implies that minimising inequality of impacts in turbulent complex environments, or at least managing them below thresholds that completely eliminate groups' knowledge, is beneficial for collective performance. This points toward future work investigating group-level impacts to determine threshold effects and more mechanistic analysis of how agent locations in social learning networks contribute. The current model randomly distributes agents across network positions regardless of group membership, but real-world collectives likely exhibit homophily where similar agents preferentially connect. For instance, complexity science researchers are more likely to learn from other complexity scientists than from individuals in different fields, which could amplify inequality effects by limiting information flow from less-affected to severely-affected groups.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e4.4 Connections to Socio-Ecological Resilience\u003c/h2\u003e\u003cp\u003eOur results provide quantitative support for established resilience principles while revealing how they apply differently across problem complexities. This aligns with socio-ecological resilience findings which show connectivity and diversity play a central role in system resilience that is context dependent (Biggs et al., 2015). Practically this implies the need for more context aware management approaches. The pattern connects to polycentric governance principles, which emphasize distributed authority and diverse decision-making centres for enhancing system resilience (Ostrom, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Polycentric governance increases diversity and autonomy across levels, fostering innovation and flexibility beneficial under changing circumstances (Singh \u0026amp; Kumar, 2024).\u003c/p\u003e\u003cp\u003eOur findings demonstrate that diversity consistently enhances performance under disruption across landscape types, while connectivity requirements vary with environmental context. Collectives navigating complex environments require fundamentally different management approaches than those in simpler environments. Maintained diversity and moderate connectivity yield highest performance under disruption, aligning with polycentric governance principles.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e4.5 Multiple Modes of Adaptation\u003c/h2\u003e\u003cp\u003eThe optimal configuration for collectives to adapt fundamentally depends on both landscape complexity and the characteristics of environmental disruption. Looking at differences in cumulative performance compared to controls for complex landscapes across different regimes, shows that both low diversity, high network density; high network density, high diversity; and high diversity, low network density configurations can all see relatively large (\u0026gt;\u0026thinsp;15%) and significant increases (Supplementary Fig.\u0026nbsp;7.). This may be indicative of different dominant modes of adaptation to new landscapes which have different resilience and response characteristics. For instance, rapid high connectivity low diversity collectives might rapidly share good solutions that result from fortunate disruptions that place agents on high performing locations of the landscape. While low network density high diversity groups may be more slowly and independently accruing a greater distribution of high performing solutions, the slower convergence and greater diversity of landscape solutions would then mean that the likelihood of all solutions massively dropping in payoff value is lower. This aligns with existing evidence in resilience and network science of collective intelligence literature (Centola, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Further mechanistic analysis of social learning and innovation along with time series analysis of performance drops and recoveries is required to confirm different modes contributing to resilience and recovery which we defer to future work.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003e4.6 Limitations and Future Directions\u003c/h2\u003e\u003cp\u003eWhile our analysis advances understanding of collective adaptation under disruption, several limitations warrant acknowledgment. Our focus on random network structures restricts insights into how specific topologies influence adaptation, and the binary comparison between K\u0026thinsp;=\u0026thinsp;0 and K\u0026thinsp;=\u0026thinsp;7 leaves intermediate complexity levels unexplored. Future work examining broader ranges of landscape complexity could reveal phase transitions in the disruption-performance relationship and identify when collectives should shift strategies. Additionally, temporal analysis of performance drops and recoveries could distinguish when rapid information sharing versus sustained exploration provides advantages.\u003c/p\u003e\u003cp\u003eThe basic NK landscape implementation doesn't capture agents' capacity to influence their environment, missing key characteristics like niche construction observed in real-world systems. Incorporating endogenous landscape dynamics through NKES (Suzuki \u0026amp; Arita, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) or NKEZ landscapes (Gavetti et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) would allow agents to deliberately modify their problem spaces, similar to policy interventions or ecosystem engineering. Such extensions could reveal how collectives might strategically complexify or simplify their decision landscapes to leverage or avoid disruption effects, potentially generating endogenous disruption dynamics.\u003c/p\u003e\u003cp\u003eGiven that multiple network-diversity configurations yield performance gains under disruption, investigating adaptive networks represents a crucial next step. Simulations where network connections evolve based on social learning or innovation patterns could reveal how collectives discover appropriate structures for their problem landscapes. This aligns with the collective adaptation framework's emphasis on meta-learning (Galesic et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Baumann et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and earlier conceptual work on Open Systems Theory exploring how organisations need to adapt to changing environments of different complexity (Emery \u0026amp; Trist, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e1965\u003c/span\u003e; Emery, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1999\u003c/span\u003e). Extension of the proposed model through adaptive network approaches (Berner et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Sayama et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) could allow exploration of these meta-learning and organisational adaptation themes. Understanding when and how collectives should switch between rapid-convergence versus distributed-exploration configurations would provide practical insights for designing resilient decision-making structures.\u003c/p\u003e\u003cp\u003eOur results demonstrate that collectives navigating complex environments require fundamentally different structures than those in simple environments. The reversal between landscapes, where disruptions hinder performance in simple problems but enhance it in complex ones, suggests that with appropriate structures, collectives can navigate and even benefit from environmental volatility rather than merely enduring it. The convergence of our findings around moderate connectivity, maintained diversity, and distributional equity aligns with polycentric governance principles while revealing how these principles apply differently across problem complexities. By maintaining or enhancing these key properties in collectives it may be possible to adapt to the disruption laden future entailed by the polycrisis.\u003c/p\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAI was used in the development of manuscript. AI coding tools were used in development of model code: Cursor-small, Claude Sonnet 4.5, Claude Opus 4.1. Generative AI was used improve readability of manuscript text: Claude Sonnet 4.5, Claude Opus 4.1.\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003cstrong\u003eFunding Declaration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eS.J.L. This research has been supported by funding from the Australian Government (Australian Research Council Future Fellowship FT200100381 to S.J.L.).\u003cbr\u003e\u0026nbsp;K.D. receives funding from the OneBasin CRC.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eR.J.T. Developed research concept, developed model and analysis approach, coded the model, ran simulations, performed data analysis, interpreted results, created all figure and wrote manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eS.J.L. Contributed to research concept developed, assisted in model design and analysis approach, assisted in research framing, contributed to interpretation of results and reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003eR.A. and K.D. Provided input on research concept \u0026amp; modelling approach, contributed to interpretation of results, reviewed manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode and Data Availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCode for the simulation model is available at: https://github.com/rosstie/caud\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNone of the authors have competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003cp\u003eAllen, C. R., Angeler, D. G., Garmestani, A. S., Gunderson, L. H., \u0026amp; Holling, C. S. (2014). Panarchy: Theory and Application. \u003cem\u003eEcosystems\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(4), 578–589. https://doi.org/10.1007/s10021-013-9744-2\u003c/p\u003e\n\u003cp\u003eBaumann, F., Czaplicka, A., \u0026amp; Rahwan, I. (2024). Network structure shapes the impact of diversity in collective learning. \u003cem\u003eScientific Reports\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(1), 2491. https://doi.org/10.1038/s41598-024-52837-3\u003c/p\u003e\n\u003cp\u003eBerner, R., Gross, T., Kuehn, C., Kurths, J., \u0026amp; Yanchuk, S. (2023). \u003cem\u003eAdaptive Dynamical Networks\u003c/em\u003e (No. arXiv:2304.05652). arXiv. http://arxiv.org/abs/2304.05652\u003c/p\u003e\n\u003cp\u003eBodin, Ö. (2017). Collaborative environmental governance: Achieving collective action in social-ecological systems. \u003cem\u003eScience (New York, N.Y.)\u003c/em\u003e, \u003cem\u003e357\u003c/em\u003e(6352). https://doi.org/10.1126/science.aan1114\u003c/p\u003e\n\u003cp\u003eBiggs, Reinette, Maja Schlüter, Duan Biggs, et al. 2012. “Toward Principles for Enhancing the Resilience of Ecosystem Services.” Annual Review of Environment and Resources 37 (Volume 37, 2012): 421–48. https://doi.org/10.1146/annurev-environ-051211-123836.\u003c/p\u003e\n\u003cp\u003eBrummitt, C. D., D’Souza, R. M., \u0026amp; Leicht, E. A. (2012). Suppressing cascades of load in interdependent networks. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e, \u003cem\u003e109\u003c/em\u003e(12), E680–E689. https://doi.org/10.1073/pnas.1110586109\u003c/p\u003e\n\u003cp\u003eBurt, R. S. (2004). Structural Holes and Good Ideas. \u003cem\u003eAmerican Journal of Sociology\u003c/em\u003e, \u003cem\u003e110\u003c/em\u003e(2), 349–399. https://doi.org/10.1086/421787 \u003c/p\u003e\n\u003cp\u003eCarreras, B.A., D.E. Newman, I. Dobson, and A.B. Poole. 2004. “Evidence for Self-Organized Criticality in a Time Series of Electric Power System Blackouts.” IEEE Transactions on Circuits and Systems I: Regular Papers 51 (9): 1733–40. https://doi.org/10.1109/TCSI.2004.834513.\u003c/p\u003e\n\u003cp\u003eCentola, D. (2022). The network science of collective intelligence. \u003cem\u003eTrends in Cognitive Sciences\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(11), 923–941. https://doi.org/10.1016/j.tics.2022.08.009\u003c/p\u003e\n\u003cp\u003eChen, J., \u0026amp; Wen, H. (2023). The application of complex network theory for resilience improvement of knowledge-intensive supply chains. \u003cem\u003eOperations Management Research\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(3), 1140–1161. https://doi.org/10.1007/s12063-023-00365-0 \u003c/p\u003e\n\u003cp\u003eClauset, Aaron, Cosma Rohilla Shalizi, and M. E. J. Newman. 2009. “Power-Law Distributions in Empirical Data.” SIAM Review 51 (4): 661–703. https://doi.org/10.1137/070710111.\u003c/p\u003e\n\u003cp\u003eEmery, F. E., and E. L. Trist. 1965. “The Causal Texture of Organizational Environments.” Human Relations 18 (1): 21–32. https://doi.org/10.1177/001872676501800103.\u003c/p\u003e\n\u003cp\u003eEmery, Merrelyn. 1999. Searching: The Theory and Practice of Making Cultural Change. John Benjamins. https://doi.org/10.1075/dowi.4.\u003c/p\u003e\n\u003cp\u003eFolke, C. (2006). Resilience: The emergence of a perspective for social–ecological systems analyses. \u003cem\u003eGlobal Environmental Change\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(3), 253–267. https://doi.org/10.1016/j.gloenvcha.2006.04.002\u003c/p\u003e\n\u003cp\u003eFolke, C., Carpenter, S., Walker, B., Scheffer, M., Chapin, T., \u0026amp; Rockström, J. (2010). Resilience Thinking: Integrating Resilience, Adaptability and Transformability. \u003cem\u003eEcology and Society\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(4). https://doi.org/10.5751/ES-03610-150420\u003c/p\u003e\n\u003cp\u003eGalesic, M., Barkoczi, D., Berdahl, A. M., Biro, D., Carbone, G., Giannoccaro, I., Goldstone, R. L., Gonzalez, C., Kandler, A., Kao, A. B., Kendal, R., Kline, M., Lee, E., Massari, G. F., Mesoudi, A., Olsson, H., Pescetelli, N., Sloman, S. J., Smaldino, P. E., \u0026amp; Stein, D. L. (2023). Beyond collective intelligence: Collective adaptation. \u003cem\u003eJournal of The Royal Society Interface\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e(200), 20220736. https://doi.org/10.1098/rsif.2022.0736\u003c/p\u003e\n\u003cp\u003eGao, J., Barzel, B., \u0026amp; Barabási, A.-L. (2016). Universal resilience patterns in complex networks. \u003cem\u003eNature\u003c/em\u003e, \u003cem\u003e530\u003c/em\u003e(7590), 307–312. https://doi.org/10.1038/nature16948\u003c/p\u003e\n\u003cp\u003eGavetti, G., Helfat, C. E., \u0026amp; Marengo, L. (2017). Searching, Shaping, and the Quest for Superior Performance. \u003cem\u003eStrategy Science\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(3), 194–209. https://doi.org/10.1287/stsc.2017.0036 \u003c/p\u003e\n\u003cp\u003eHelbing, Dirk. 2013. “Globally Networked Risks and How to Respond.” Nature 497 (7447): 51–59. https://doi.org/10.1038/nature12047.\u003c/p\u003e\n\u003cp\u003eKauffman, S., \u0026amp; Levin, S. (1987). Towards a general theory of adaptive walks on rugged landscapes. \u003cem\u003eJournal of Theoretical Biology\u003c/em\u003e, \u003cem\u003e128\u003c/em\u003e(1), 11–45. https://doi.org/10.1016/S0022-5193(87)80029-2 \u003c/p\u003e\n\u003cp\u003eLallemant, David, and Anne Kiremidjian. 2015. “A Beta Distribution Model for Characterizing Earthquake Damage State Distribution.” Earthquake Spectra 31 (3): 1337–52. https://doi.org/10.1193/012413EQS013M.\u003c/p\u003e\n\u003cp\u003eLazer, D., \u0026amp; Friedman, A. (2007). The Network Structure of Exploration and Exploitation. \u003cem\u003eAdministrative Science Quarterly\u003c/em\u003e, \u003cem\u003e52\u003c/em\u003e(4), 667–694. https://doi.org/10.2189/asqu.52.4.667\u003c/p\u003e\n\u003cp\u003eLevinthal, D. A. (1997). Adaptation on Rugged Landscapes. \u003cem\u003eManagement Science\u003c/em\u003e, \u003cem\u003e43\u003c/em\u003e(7), 934–950.\u003c/p\u003e\n\u003cp\u003eMarch, J. G. (1991). Exploration and Exploitation in Organizational Learning. \u003cem\u003eOrganization Science\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(1), 71–87.\u003c/p\u003e\n\u003cp\u003eNewman, M. E. J. 2003. “The Structure and Function of Complex Networks.” SIAM Review 45 (2): 167–256. https://doi.org/10.1137/S003614450342480.\u003c/p\u003e\n\u003cp\u003eNewman, M. E. J. 2005. “Power Laws, Pareto Distributions and Zipf’s Law.” Contemporary Physics 46 (5): 323–51. https://doi.org/10.1080/00107510500052444.\u003c/p\u003e\n\u003cp\u003eOstrom, E. (2010). Beyond Markets and States: Polycentric Governance of Complex Economic Systems. \u003cem\u003eAmerican Economic Review\u003c/em\u003e, \u003cem\u003e100\u003c/em\u003e(3), 641–672. https://doi.org/10.1257/aer.100.3.641\u003c/p\u003e\n\u003cp\u003eSayama, H., Pestov, I., Schmidt, J., Bush, B. J., Wong, C., Yamanoi, J., \u0026amp; Gross, T. (2013). Modeling complex systems with adaptive networks. \u003cem\u003eComputers \u0026amp; Mathematics with Applications\u003c/em\u003e, \u003cem\u003e65\u003c/em\u003e(10), 1645–1664. https://doi.org/10.1016/j.camwa.2012.12.005\u003c/p\u003e\n\u003cp\u003eScheffer, M., Bascompte, J., Brock, W. A., Brovkin, V., Carpenter, S. R., Dakos, V., Held, H., van Nes, E. H., Rietkerk, M., \u0026amp; Sugihara, G. (2009). Early-warning signals for critical transitions. \u003cem\u003eNature\u003c/em\u003e, \u003cem\u003e461\u003c/em\u003e(7260), 53–59. https://doi.org/10.1038/nature08227\u003c/p\u003e\n\u003cp\u003eSiggelkow, N., \u0026amp; Rivkin, J. W. (2005). Speed and Search: Designing Organizations for Turbulence and Complexity. \u003cem\u003eOrganization Science\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(2), 101–122. https://doi.org/10.1287/orsc.1050.0116 \u003c/p\u003e\n\u003cp\u003eSuzuki, R., \u0026amp; Arita, T. (2005). How Niche Construction Can Guide Coevolution. In M. S. Capcarrère, A. A. Freitas, P. J. Bentley, C. G. Johnson, \u0026amp; J. Timmis (Eds.), \u003cem\u003eAdvances in Artificial Life\u003c/em\u003e (pp. 373–382). Springer. https://doi.org/10.1007/11553090_38\u003c/p\u003e\n\u003cp\u003eVose, David. 2008. Risk Analysis : A Quantitative Guide. With Internet Archive. Chichester, England ; Hoboken, NJ : Wiley. http://archive.org/details/riskanalysisquan0000vose.\u003c/p\u003e\n\u003cp\u003eWeinberger, E. D. (1991). Local properties of Kauffman’s N-k model: A tunably rugged energy landscape. \u003cem\u003ePhysical Review A\u003c/em\u003e, \u003cem\u003e44\u003c/em\u003e(10), 6399–6413. https://doi.org/10.1103/PhysRevA.44.6399\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":"npj-complex","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjcomplex","sideBox":"Learn more about [npj Complexity](https://www.nature.com/npjcomplex/)","snPcode":"44260","submissionUrl":"https://submission.springernature.com/new-submission/44260/3","title":"npj Complexity","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"collective intelligence, collective adaptation, NK landscape, resilience, social learning, agent-based model, networks","lastPublishedDoi":"10.21203/rs.3.rs-8115470/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8115470/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe polycrisis presents a future characterized by multiple co-occurring crises, demanding human collectives capable of navigating complex challenges simultaneously. This requires understanding how human collectives should be structured to improve their capacity to find and maintain effective solutions despite ongoing and frequent disruptions. We model collective problem-solving using an NK landscape impacted by intermittent disruptions to examine how network connectivity and agent diversity influence performance under ongoing disruptions. We measure the capacity for different collective structures to find and maintain high-performing solutions through various disruption regimes (frequency of disruption, impact size, impact distribution). Results reveal that disruptions degrade collective cumulative performance in simple landscapes (K\u0026thinsp;=\u0026thinsp;0) but enhance it in complex ones (K\u0026thinsp;=\u0026thinsp;7). In complex landscapes, moderate connectivity (d\u0026thinsp;\u0026asymp;\u0026thinsp;0.2\u0026ndash;0.32) and high diversity maximize cumulative performance, with disruptions amplifying these benefits. Additionally, we observe that impact distribution across agent groups affects cumulative performance, with skewed distributions. Our analysis demonstrates that in complex landscapes, disruptions can actually improve collective problem-solving capacity, but only with appropriate collective structures. This suggests that effective collective design requires matching structure to both problem complexity and disruption environment rather than applying universal principles.\u003c/p\u003e","manuscriptTitle":"Landscape Complexity Shapes the Role of Network Density and Diversity in Collective Adaptation Under Disruption","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-04 05:25:42","doi":"10.21203/rs.3.rs-8115470/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-05T17:29:24+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-23T10:57:18+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-10T14:22:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"41423423119596070457122158628516980561","date":"2026-01-24T09:16:23+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"259267137097911528628452487499843698199","date":"2026-01-20T18:26:03+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-19T07:58:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"168416768843002767007801014797781073141","date":"2026-01-10T16:11:52+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-08T18:52:28+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-11-27T04:45:06+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-11-24T05:39:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Complexity","date":"2025-11-14T13:19:32+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-complex","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjcomplex","sideBox":"Learn more about [npj Complexity](https://www.nature.com/npjcomplex/)","snPcode":"44260","submissionUrl":"https://submission.springernature.com/new-submission/44260/3","title":"npj Complexity","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a00fcde6-1347-4ae8-ac6a-830354d5a367","owner":[],"postedDate":"December 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[{"id":58686058,"name":"Physical sciences/Mathematics and computing"},{"id":58686059,"name":"Physical sciences/Physics"}],"tags":[],"updatedAt":"2026-03-05T17:39:47+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-04 05:25:42","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8115470","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8115470","identity":"rs-8115470","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}