COGNITIVE BIASES: UNDERSTANDING AND DESIGNING FAIR AI SYSTEMS FOR SOFTWARE DEVELOPMENT

COGNITIVE BIASES: UNDERSTANDING AND DESIGNING FAIR AI SYSTEMS FOR SOFTWARE DEVELOPMENT | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL This is a preprint and has not been peer reviewed. Data may be preliminary. 7 April 2025 V1 Latest version Share on COGNITIVE BIASES: UNDERSTANDING AND DESIGNING FAIR AI SYSTEMS FOR SOFTWARE DEVELOPMENT Authors : Sheriff Adefolarin Adepoju 0009-0006-6741-8518 [email protected] and Mildred Aiwanno-Ose Adepoju Authors Info & Affiliations https://doi.org/10.22541/au.174406322.29071173/v1 Published International Journal of Advanced Computer Science and Applications Version of record Peer review timeline 907 views 224 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Nowadays, innovative systems incorporate machine learning processes to shape important decisions requiring absolute fair treatment for all users to mimic human motives. Cognitive biases function as systematic patterns of irrational decision-making in human judgment, which can unintentionally appear within AI systems and result from biased data accumulation, algorithmic creation, and human supervision. The research investigates AI system bias interactions with analysis of software development processes, which increases social inequities throughout the development cycle. The analysis includes three typical cognitive biases that affect AI model development: confirmation bias, anchoring Bias, and automation bias. The research seeks to reduce AI system bias, including utilizing varied data, performing bias audits, and deploying algorithms that focus on fairness. This document works to instruct developers and stakeholders about building equitable, trustable AI systems through biased education. COGNITIVE BIASES: UNDERSTANDING AND DESIGNING FAIR AI SYSTEMS FOR SOFTWARE DEVELOPMENT Sheriff Adefolarin Adepoju 1 , Mildred Aiwanno-Ose Adepoju 2 1 Department of Computer Science, College of Engineering, Prairie View A&M University, Texas, United States 2 Department of Computer Information Systems, College of Engineering, Prairie View A&M University, Texas, United States. ORCID 0009-0006-6741-8518 (Sheriff Adefolarin Adepoju) Abstract Nowadays, innovative systems incorporate machine learning processes to shape important decisions requiring absolute fair treatment for all users to mimic human motives. Cognitive biases function as systematic patterns of irrational decision-making in human judgment, which can unintentionally appear within AI systems and result from biased data accumulation, algorithmic creation, and human supervision. The research investigates AI system bias interactions with analysis of software development processes, which increases social inequities throughout the development cycle. The analysis includes three typical cognitive biases that affect AI model development: confirmation bias, anchoring Bias, and automation bias. The research seeks to reduce AI system bias, including utilizing varied data, performing bias audits, and deploying algorithms that focus on fairness. This document works to instruct developers and stakeholders about building equitable, trustable AI systems through biased education. Keywords Cognitive Biases, Fair AI Systems, Algorithmic Bias, Software Development, Bias Mitigation, Fairness in Software Development , Bias Mitigation in AI Systems Introduction Steaming from an argument that artificial intelligence simulation of human decision-making can lead to selfish decisions, we noticed that sources of data play a significant role in the output bias of any system. AI systems process substantial database information to recognize patterns through which they execute complex decisions at a superior pace and precision than human capabilities. AI weighs heavily into decision-making procedures, but this advancement triggers doubts regarding fairness, transparency, and accountability requirements, according to Mehrabi et al. (2021). AI developers must address cognitive biases, which represent systematic human judgment errors that unintentionally enter AI systems (Tversky & Kahneman, 1974). These unaddressed biases build an ongoing cycle that deepens and intensifies social disparities (Crawford, 2021). Human cognition includes behavioral biases as a standard functioning element that influences how individuals understand and interpret information (Tversky & Kahneman, 1974). Multiple components within artificial intelligence systems enable biases to appear, including biased programming data, developer-made choices, and the methods through which algorithms function (Mehrabi et al., 2021). Human developers frequently experience confirmation bias while making assumptions, so they select datasets that confirm their beliefs, according to Nickerson (1998). Users make errors in judgment due to automation bias by trusting computers too much instead of verifying their output integrity or equity (Crawford, 2021). Artificial Intelligence has begun to move into HVAC systems (Adepoju, 2025) and has embedded social and ethical implications because of biases that developers introduce (Barocas et al., 2019). Cognitive biases in AI systems create noticeable effects across many different fields of operation. Crime risk assessment programs that run inside criminal justice systems show systematic errors in how they assign high-risk status to minority populations, resulting in biased institutional treatment (Angwin et al., 2016). Medical algorithms demonstrate discriminatory behavior against Black patients by providing reduced access to proper medical care relative to white patients (Obermeyer et al., 2019). The urgent requirement for fairness-aware practices and rigorous bias mitigation exists in all AI development projects (Mitchell et al., 2018). A solution for cognitive biases requires researchers and practitioners to use technical methods and ethical frameworks. Data representation from diverse sources helps reduce biases that stem from data (Buolamwini & Gebru, 2018). Fairness-aware algorithms function to fix biases that exist during model development, according to Mehrabi et al. (2021). Conducting bias audits regularly helps organizations detect and manage biases that appear across AI system development stages (Raji et al., 2020). A team collaboration between technical experts, ethicists, and policymakers will create integrated solutions to resolve artificial intelligence systems and human prejudice issues (Crawford, 2021). The fundamental requirement for public confidence alongside fair service for all communities depends on transparency and accountability (Mitchell et al., 2018). This article extensively assesses bias occurrences present within AI systems used during software development processes. The article studies the origin of these biases and their influence on automated decision systems while examining effective strategies for suppressing their influence. Developers and stakeholders should work together to make AI systems fairer through deepened cognitive bias understanding and ethical design practice promotion. LITERATURE REVIEW The existing literature about cognitive biases acting upon AI systems shows that bias exists in many forms and affects how AI systems perform in ethical decisions. Developing fair AI systems requires researchers to comprehend and reduce cognitive biases according to scholars. This part of the review examines scholarly research which discusses AI system cognitive biases while studying their sources and presenting recommended solutions to minimize their impact. Tversky & Kahneman 1974 Defined cognitive biases as systematic deviations from rational judgment. Anchoring bias, availability heuristic Barocas et al. 2019 Explored how bias in training data affects AI outcomes. Data bias, algorithmic bias Mehrabi et al. 2021 Analyzed the various sources of bias in AI and proposed mitigation techniques. Confirmation bias, selection bias Obermeyer et al. 2019 Identified racial bias in healthcare algorithms and its impact on patient outcomes. Racial bias, outcome bias Mitchell et al. 2018 Suggested fairness-aware algorithms and bias auditing as key mitigation strategies. Automation bias, data bias Crawford 2021 Critiqued the social and ethical implications of AI bias. Automation bias, confirmation bias The 1974 Tversky-Kahneman research established the psychological roots of cognitive biases which became important for studies about AI system bias production. The researchers at Barocas et al. (2019) maintained their analysis within AI systems to demonstrate how prejudice in training datasets maintains institutional discriminatory practices. The research by Mehrabi et al. (2021) created a classification structure for biases as well as an approach for reducing their occurrence. The work of Obermeyer et al. (2019) shows how healthcare algorithms using racial bias produce health inequality results. The practical consequences of defective AI systems on disadvantaged populations become obvious through their findings. The authors Mitchell et al. (2018) along with Crawford (2021) proposed fairness-aware algorithms and stringent bias auditing as practical methods to address bias but Crawford stressed the requirement for increased AI development accountability. Scientific research dismisses the need for immediate action on cognitive bias detection in AI to develop ethical solutions for technology. METHODOLOGY This section outlines the systematic approach to analyzing cognitive biases in AI systems and the strategies for designing fair and equitable AI models. The methodology combines qualitative research methods with a comprehensive framework for identifying, evaluating, and mitigating biases. By employing a multi-dimensional approach, this study ensures a thorough investigation of how cognitive biases manifest in AI systems and how to reduce their impact effectively. Research Design The research methodology uses qualitative methods to extensively evaluate literature observations and real-world case studies by authentic sources. Because of its suitability, the chosen research design effectively explores detailed topics between cognitive biases and AI systems (Creswell, 2014). By merging theoretical information with practical examples, the research project reveals standard patterns in bias appearance and methods for its mitigation. The research design contains three primary stages that follow one another. 1. Literature Review: Analyzing scholarly articles, technical reports, and industry guidelines on cognitive biases and their influence on AI systems. During this stage, researchers explore available knowledge about existing gaps and established mitigation methods (Barocas et al., 2019; Mehrabi et al., 2021). 2. The project examines documented AI system studies from different sectors including healthcare, criminal justice and recruitment to determine how biases manifest and what mitigation approaches are utilized based on Obermeyer et al. (2019) and Raji et al. (2020). 3. The proposed framework creates a system for classifying biases and strategy assessment capabilities. A methodical method enables an organized assessment of AI model susceptibility to biases and their effective correction (Mitchell et al., 2018). 4. Data Collection This inestigation used secondary data from conferences, peer-reviewed journals, technical reports, and case studies. We established the following conditions as we navigated our choice of resource materials: Relevance to cognitive biases in AI and software development. Empirical evidence of bias impact and mitigation outcomes. Also included are recent publications from the past ten years that encompass modern breakthroughs and developing challenges. The research relies on primary cognitive bias literature from Tversky and Kahneman (1974) together with present-day AI fairness examinations by Barocas et al. (2019) and Mehrabi et al. (2021). Practical industry reports about algorithm audits and bias assessment became part of the research through Raji et al., 2020. Bias Identification Framework The identification of cognitive biases in AI systems followed a systematic approach to classification as well as analysis. The framework understands cognitive biases into three main sections. 1. The classification of data bias occurs when there are insufficient or improperly organized datasets (Buolamwini & Gebru, 2018). 2. The fundamental process of model optimization and design selection at any stage causes Bias referred to as Algorithmic Bias (Mehrabi et al., 2021). 3. Human Bias originates from the subjective choices made by developers and end-users according to Crawford (2021). The defined classification system enables researchers to analyze how biases develop throughout the AI development process as well as the effects they have on system results. Evaluation Criteria for Bias Mitigation Strategies The evaluation of bias mitigation strategies happened through an assessment of four vital components: 1. A strategy proves effective when it decreases or eliminates biases from artificial intelligence output. 2. The practicality of deploying the strategy throughout the artificial intelligence development pattern defines feasibility. 3. Internal and external researchers must have accessible insight into the methods used to mitigate biases and understand them clearly throughout the whole process. 4. The strategy can scale across different AI applications and domains according to its scalability measures (Mitchell et al., 2018). The set criteria help professionals achieve an equitable rating of bias reduction strategies and their associated implementation difficulties. Bias Mitigation Techniques The research provides analysis of chosen bias mitigation methods from four specified categories: 2. A comprehensive data collection strategy employs diverse participants from target groups to reduce potential Bias according to Buolamwini and Gebru (2018). 3. A training approach for fair algorithms integrates explicit fairness rules that prevent bias production (Mehrabi et al., 2021). 4. Regular audits through the AI lifecycle for identifying and fixing hidden biases happen through Bias Audits and Fairness Assessments (Raji et al., 2020). 5. The collaborative approach brings together experts from developer roles with ethicists and social scientists to resolve team and human cognitive biases by using methods described by Crawford (2021). The effectiveness along with obstacles and practical compatibility of these methods gets evaluated specifically for their role in real-world AI systems. Data Analysis Approach The study analyzes recurring themes through thematic analysis to examine both bias manifestation patterns and methods to reduce it. The authors first coded the data from literature investigations and case studies before categorizing them following the benchmark criteria of the bias identification framework. This method enables researchers to gain both extensive understanding of cognitive bias origin and an evaluation of various intervention methods (Braun & Clarke, 2006). The assessment spans three different stages. 1. Researchers analyzed the key data about bias varieties, their effects and defense approaches within the obtained materials. 2. The analysis unitates coded data into designated themes which include data-related Bias alongside algorithmic Bias and human-related Bias and it detects new patterns that arise. 3. This section unites the thematic information into a unified explanation about cognitive bias understanding and management in AI systems. 4. Validity and Reliability The research team takes several measures to achieve both valid and reliable results. 1. The study implements data triangulation by combining academic literature analysis with case study results along with industry report content for finding confirmation. 2. Industry experts in both AI ethics and software development conduct peer evaluation of the developed analysis framework along with its final conclusions. 3. The method includes maintaining a detailed record of data origins along with programming choices and analytical processes through an Audit Trail system which improves research repeatability. 4. Ethical Considerations The study implements ethical research procedures through these following measures: 1. The study adopts methods to show data accurately while reducing interpretive errors. 2. SOURCES must be referenced accurately because the research team respects intellectual property rights. 3. Analytical processes should be completely documented to achieve transparency throughout the work. Figure 1: The below diagram shows the AI Bias Research on Methodology and Ethical framework RESULTS The findings of this study reveal the pervasive influence of cognitive biases on AI systems and the effectiveness of various mitigation strategies. Through the analysis of existing literature and case studies, the results highlight how cognitive biases manifest in AI development, the challenges associated with addressing these biases, and the impact of mitigation techniques in promoting fair and equitable AI systems. This section presents the key outcomes of the study, supported by empirical evidence and a comprehensive evaluation framework. 4.1 Manifestation of Cognitive Biases in AI Systems The study identifies three primary channels through which cognitive biases infiltrate AI systems: • Data Bias: Bias introduced through unrepresentative or incomplete datasets. For example, research by Buolamwini and Gebru (2018) shows that facial recognition systems exhibit higher error rates for darker-skinned individuals due to the lack of diverse training data. This bias can result in discriminatory outcomes in applications like law enforcement and hiring systems. • Algorithmic Bias: Bias that emerges from the design and structure of AI models. Algorithms trained on biased data or optimized for specific performance metrics without fairness constraints may produce systematically biased outputs (Mehrabi et al., 2021). For instance, risk assessment algorithms in criminal justice have been shown to overestimate recidivism rates for minority groups (Angwin et al., 2016). • Human Bias: Bias stemming from the subjective decisions of AI developers and users. Confirmation bias, for example, may cause developers to prioritize data that aligns with their expectations, while automation bias leads users to trust AI outputs without critical examination (Crawford, 2021). 4.2Effectiveness of Bias Mitigation Strategies The evaluation of mitigation strategies suggests that a multi-faceted approach is necessary to address cognitive biases effectively. The following techniques emerged as the most impactful: • Diverse and Representative Data Collection: Ensuring datasets are inclusive reduces the likelihood of biased outcomes. This study confirms previous findings that data diversity enhances model fairness (Mitchell et al., 2018). • Fairness-Aware Algorithms: Implementing fairness constraints during model training can mitigate algorithmic bias. For example, re-weighting data or applying adversarial debiasing has been shown to improve outcome parity across demographic groups (Mehrabi et al., 2021). • Bias Audits and Fairness Assessments: Regular audits help identify and address hidden biases. Raji et al. (2020) advocate for third-party audits to enhance accountability and transparency in AI deployment. The table below summarizes the effectiveness of these strategies across key evaluation criteria: Diverse and Representative Data High – Reduces data-driven biases (Mitchell et al., 2018) Moderate – Requires significant data collection and curation High – Transparent if dataset characteristics are disclosed Moderate – Feasible with appropriate resources and policies Fairness-Aware Algorithms High – Mitigates algorithmic disparities (Mehrabi et al., 2021) Moderate – Involves additional algorithm design complexity Moderate – Depends on documentation of model adjustments Low – Requires case-by-case customization Bias Audits and Fairness Checks Moderate – Identifies hidden biases (Raji et al., 2020) High – Practical with audit tools and protocols High – Facilitates external and internal review processes High – Applicable across diverse AI systems 4.3 Challenges in Implementing Bias Mitigation Despite the effectiveness of these strategies, several challenges limit their implementation: 1. The process of making data more fair usually leads to reduced model precision. High-value domains including healthcare and criminal justice experience the most intense difficulties between fairness improvement and prediction accuracy according to Corbett-Davies and Goel (2018). 2. The lack of budgetary funds prevents smaller companies from executing complete audits and fairness-aware algorithm deployments (Barocas et al. 2019). 3. The process to eliminate human biases inherent in AI decision-making systems becomes harder due to their complexity. Applying cultural and organization changes stands as an essential requirement to handle this matter (Crawford, 2021). 4.4 Impact of Bias Mitigation on AI Outcomes When AI models use bias mitigation frameworks they generate results which minimize inequities. The implementation of healthcare algorithms using demographic parity adjustments eliminated disparities in patient treatment order (Obermeyer et al., 2019). Predictive accuracy maintained its high level while fairness-aware models achieved better hiring equity in their results (Mitchell et al., 2018). An integration of technical methods alongside organizational measures alongside ethical guidelines helps AI developers to reduce harmful cognitive biases which leads to better trust in AI systems. A complete approach enables AI system developers to construct efficient technology that delivers both accuracy alongside fairness together with social responsibility. Top of Form Bottom of Form DISCUSSION The research demonstrates how cognitive biases strongly affect AI systems operating in software development while identifying successful mitigation techniques. This section analyzes the obtained results by explaining their meaning and connects them to previous research in the field. The Influence of Cognitive Biases on AI Systems Different stages throughout the operation of AI systems enable cognitive biases to affect data collection and algorithmic processing along with human oversight. AI models developed using datasets tend to reproduce the biases present in societal and structural inequalities because their training relies on these datasets (Barocas et al., 2019). AI systems develop biased outputs because developers tend to choose information that confirms their initial beliefs which reinforces faulty preconceptions (Crawford 2021). Algorithmic bias creates severe problems during hiring and criminal justice operations since biased systems contribute to maintaining social disparities (Mehrabi et al., 2021). Obermeyer et al. (2019) matched this study’s findings about bias in algorithms which are primarily caused by sparse datasets. The underlying bias within facial recognition technology produces higher identification errors among population groups because training algorithms adopt prejudiced information (Buolamwini & Gebru, 2018). The research validates Mitchell et al. (2018) position by showing data quality combined with diversity serves as a solution for bias prevention. The evaluation demonstrates multiple strategies must be employed to minimize bias since any single approach provides inadequate elimination results. The collection of diverse representative data proved to be an exceptionally powerful strategy for managing data bias according to Mehrabi et al. (2021). Data diversity ensures the collection of diverse populations and contextual characteristics which minimizes the occurrence of exclusionary outcomes according to Mitchell et al., 2018. The process of obtaining diverse datasets faces hindrances according to previous research because of privacy restrictions and resource scarcity (Barocas et al., 2019). The integration of fairness constraints during model training enables fair machine learning algorithms according to Mehrabi et al. (2021). The investigators at Obermeyer et al. (2019) demonstrated that healthcare applications achieved higher fairness levels through data training methods that adjusted weights for balancing demographic characteristics. The deployment of fair models demands strategic adjustment because doing so might decrease predictive capability (Corbett-Davies & Goel, 2018). The process of conducting bias audits together with fairness assessments makes it possible to detect and resolve secretive biases according to Mitchell et al. (2018). Sending models for regular audits enables the detection of unfairness through enhanced transparency. The authors support Raji et al. (2020) in recommending independent third-party auditing procedures to maintain both objectivity and accountability within AI systems. Challenges in Bias Mitigation While these strategies achieve their intended goals substantial difficulties remain in the present situation. The key drawback emerges from the compromise that happens between achieving fairness goals and maintaining model performance quality. According to Corbett-Davies & Goel (2018) the implementation of fairness constraints has adverse effects on model performance especially when datasets are unbalanced. The need for simultaneous accuracy and fairness poses great challenges mainly in high-consequence sectors such as healthcare and criminal justice systems. The challenges stem from human involvement which results in biased outputs. The encoding process of developer biases occurs when developers make choices about data selection during AI system design (Crawford, 2021). The effectiveness of bias awareness training gets limited by cultural and organizational resistance to change but allows critical reflection to reduce biases (Mitchell et al., 2018). Implications for AI System Design The study findings establish critical value for using multiple strategies to fight bias in computing systems. The complete lifecycle development of AI demands software developers along with stakeholders to maintain transparency and accountability according to Raji et al. (2020). Testing systems for fairness should be performed at different points during the AI development process from data gathering to building the model and post-deployment assessment to guarantee equitable results (Mehrabi et al., 2021). The resolution of social and ethical dimensions in AI bias needs interdisciplinary cooperation between computer scientists together with ethicists and social scientists (Crawford, 2021). Interdisciplinary joint efforts between experts provide both comprehensive perspectives along with guidance to manage the challenges that emerge from technical enhancements versus social fairness goals (Barocas et al., 2019). Future Research Directions Refined research is necessary to establish scalable automatic bias reduction methods which should integrate into current AI operational procedures (Raji et al., 2020). The investigation of bias audit effects on fairness-aware algorithms through time should be a future research goal together with evaluations of their sustained equitable outcomes (Mitchell et al., 2018). The analysis of new AI applications demands additional research because they introduce novel sources of bias which include generative models along with autonomous decision systems (Buolamwini & Gebru, 2018). The correct approach to handle AI system cognitive biases consists of integration between technical solutions and organizational frameworks and ethical considerations. AI system developers can generate fairer and more transparent system designs by acknowledging and accomplishing these biases according to Mehrabi et al. (2021). Figure2: The below diagram shows the Cognitive Biases in AI: Influence, Migration, and Future Directions CONCLUSION AI technology adoption at high speed throughout healthcare and financial institutions along with criminal justice operations causes society to question both fairness and transparency dimensions. The systematic judgment errors known as cognitive biases severely influence AI unfairness by getting absorbed in the systems through unrepresentative data and flawed algorithms and human oversight processes (Tversky & Kahneman, 1974; Barocas et al., 2019). Marked prejudices exist in society because they create enduring disparities between populations and they damage societal confidence (Obermeyer et al., 2019). Three primary paths through which AI develops bias involve data bias from unbalanced datasets according to Buolamwini & Gebru (2018) as well as algorithmic bias from faulty training systems according to Mehrabi et al. (2021) and human bias which stems from subjective developer choices according to Crawford (2021). These sources involve different obstacles in achieving fairness. The process of data collection mitigation uses methods like Mitchell et al. (2018) while fairness-aware algorithm implementation with Corbett-Davies & Goel (2018) adds fairness constraints alongside bias auditing techniques from Raji et al. (2020) to improve transparency. Model accuracy diminishes when developers try to enhance fairness which represents a key problem for critical applications according to Corbett-Davies & Goel (2018). Large organizations need significant resources to implement fairness-aware systems which further restricts smaller organizations from accessing and applying these systems (Barocas et al, 2019). Only technical solutions do not remove human prejudice from systems. The effort must include training about bias detection along with cross-disciplinary teamwork between computer technology experts and ethical standards professionals together with social science researchers to develop AI systems which reflect social values (Crawford, 2021; Mehrabi et al., 2021). Future investigations need to create automatic detection methods which scale up for bias detection while studying the extended impact of bias reduction techniques (Raji et al., 2020; Mitchell et al., 2018). The study of emerging AI technologies like generative models needs more investigation to stop bias from intensifying (Buolamwini & Gebru, 2018). Declaration Of Competing Interest We declare that they have no known competing financial interests or personal relationships that could have influenced the work reported in this study. Acknowledgments I would like t to thank God for giving us the opportunity to start this project. Author Contributions Sheriff Adefolarin Adepoju 1 : Conceptualization, data curation, Formal Analysis, Funding, Investigation, Methodology, Project administration, software, Validation, Visualization, Writing- original draft, writing- review, and editing Mildred Aiwanno-Ose Adepoju 2 : literature review, Writing- original draft, Writing- review & editing. Funding This work is self-funded after noticing the gaps in the slowness of the new API development frameworks. Data Availability Statement The data is available from the corresponding author upon reasonable request. Conflicts of Interest There was no conflict of interest in the progression of this research. 1. References 2. Bernault, C., Juan, S., Delmas, A., Andre, J. M., Rodier, M., & Chraibi Kaadoud, I. (2023, July). Assessing the impact of cognitive biases in AI project development. In International Conference on Human-Computer Interaction (pp. 401-420). Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-35891-3_24 3. Scatiggio, V. (2020). Tackling the issue of bias in artificial intelligence to design ai-driven fair and inclusive service systems. How human biases are breaching into ai algorithms, with severe impacts on individuals and societies, and what designers can do to face this phenomenon and change for the better. https://hdl.handle.net/10589/186118 4. Vakali, A., & Tantalaki, N. (2024). Rolling in the deep of cognitive and AI biases. arXiv preprint arXiv:2407.21202 . https://doi.org/10.48550/arXiv.2407.21202 5. Schwartz, R., Schwartz, R., Vassilev, A., Greene, K., Perine, L., Burt, A., & Hall, P. (2022). Towards a standard for identifying and managing bias in artificial intelligence (Vol. 3, p. 00). Gaithersburg, MD: US Department of Commerce, National Institute of Standards and Technology. https://doi.org/10.7717/peerj-cs.1630 6. Varona, D., & Suárez, J. L. (2022). Discrimination, bias, fairness, and trustworthy AI. Applied Sciences , 12 (12), 5826. (https://creativecommons.org/licenses/by/4.0/ 7. Chen, Y., Clayton, E. W., Novak, L. L., Anders, S., & Malin, B. (2023). Human-centered design to address biases in artificial intelligence. Journal of medical Internet research , 25 , e43251. https://doi.org/10.2196/43251 8. AlMakinah, R., Goodarzi, M., Tok, B., & Canbaz, M. A. (2024). Mapping artificial intelligence bias: a network-based framework for analysis and mitigation. AI and Ethics , 1-20. 1684–1692 (2023). https://doi.org/10.1093/jamia/ocad118 9. Bellamy, R. K., Dey, K., Hind, M., Hoffman, S. C., Houde, S., Kannan, K., … & Zhang, Y. (2019). AI Fairness 360: An extensible toolkit for detecting and mitigating algorithmic bias. IBM Journal of Research and Development , 63 (4/5), 4-1. DOI: 10.1147/JRD.2019.2942287 10. Fuad, N. R. Cognitive Bias and AI Technology: The Human Element in Automated Recruitment and Selection. 6. https://doi.org/10.1038/d41586-018-05707-8 11. Tejani, A. S., Ng, Y. S., Xi, Y., & Rayan, J. C. (2024). Understanding and mitigating bias in imaging artificial intelligence. Radiographics , 44 (5), e230067. https://doi.org/10.1148/rg.230067 12. Kordzadeh, N., & Ghasemaghaei, M. (2022). Algorithmic bias: review, synthesis, and future research directions. European Journal of Information Systems , 31 (3), 388-409. https://doi.org/10.1080/0960085X.2021.1927212 13. Mensah, G. B. (2023). Artificial intelligence and ethics: a comprehensive review of bias mitigation, transparency, and accountability in AI Systems. Preprint, November , 10 (1). https://doi.org/10.2196/23218 14. Militello, L. G., Diiulio, J., Wilson, D. L., Nguyen, K. A., Harle, C. A., Gellad, W., & Lo-Ciganic, W. H. (2025). Using human factors methods to mitigate bias in artificial intelligence-based clinical decision support. Journal of the American Medical Informatics Association , 32 (2), 398-403. https://doi.org/10.1093/jamia/ocae291 15. Oguntibeju, O. O. (2024). Mitigating artificial intelligence bias in financial systems: A comparative analysis of debiasing techniques. Asian Journal of Research in Computer Science , 17 (12), 165-178. https://doi.org/10.9734/ajrcos/2024/v17i12536 16. Pant, A., Hoda, R., Tantithamthavorn, C., & Turhan, B. (2024). Navigating Fairness: Practitioners’ Understanding, Challenges, and Strategies in AI/ML Development. arXiv preprint arXiv:2403.15481 . https://doi.org/10.48550/arXiv.2403.15481 16 17. Balayn, A., Lofi, C., & Houben, G. J. (2021). Managing bias and unfairness in data for decision support: a survey of machine learning and data engineering approaches to identify and mitigate bias and unfairness within data management and analytics systems. The VLDB Journal , 30 (5), 739-768. https://doi.org/10.1145/1810295.1810326 18. Zou J, Schiebinger L (2018) AI can be sexist and racist—it’s time to make it fair. Nature. https://doi.org/10.1038/d41586-018-05707-8 19. Hall, P., & Ellis, D. (2023). A systematic review of socio-technical gender bias in AI algorithms. Online Information Review , 47 (7), 1264-1279. https://doi.org/10.1108/OIR-08-2021-0452 20. Cirillo, D., & Rementeria, M. J. (2022). Bias and fairness in machine learning and artificial intelligence. In Sex and gender bias in technology and artificial intelligence (pp. 57-75). Academic Press. https://doi.org/10.1016/B978-0-12-821392-6.00006-6 21. Sheriff Adefolarin Adepoju, “How machine learning can revolutionize building comfort: Accessing the impact of occupancy prediction models on HVAC control system,” World J. Adv. Res. Rev., vol. 25, no. 1, pp. 2315–2327, Jan. 2025, doi: 10.30574/wjarr.2025.25.1.0161. 22. Soleimani, M. (2022). Developing unbiased artificial intelligence in recruitment and selection: a processual framework: a dissertation presented in partial fulfilment of the requirements for the degree of doctor of philosophy in management at Massey University, Albany, Auckland, New Zealand (Doctoral dissertation, Massey University). http://hdl.handle.net/10179/17686 23. Devillers, L., Fogelman-Soulié, F., & Baeza-Yates, R. (2021). AI & human values: Inequalities, biases, fairness, nudge, and feedback loops. Reflections on artificial intelligence for humanity , 76-89. https://doi.org/10.1145/3213765 24. Cary Jr, M. P., Bessias, S., McCall, J., Pencina, M. J., Grady, S. D., Lytle, K., & Economou‐Zavlanos, N. J. (2025). Empowering nurses to champion Health equity & BE FAIR: Bias elimination for fair and responsible AI in healthcare. Journal of Nursing Scholarship , 57 (1), 130-139. https://doi.org/10.1111/jnu.13007 25. Desouza, K. C., Dawson, G. S., & Chenok, D. (2020). Designing, developing, and deploying artificial intelligence systems: Lessons from and for the public sector. Business Horizons , 63 (2), 205-213. https://doi.org/10.1016/j.bushor.2019.11.004 26. Smith, C. J. (2019). Designing trustworthy ai: A human-machine teaming framework to guide development. arXiv preprint arXiv:1910.03515 . https://doi.org/10.48550/arXiv.1910.03515 27. Top of FormBottom of Form Sheriff Adefolarin Adepoju is a graduate student of Prairie View A&M University, Computer Science Department. He completed his master’s degree in computer science in Summer of 2024 and his master’s in architecture from University of Lincoln in United Kingdom. He currently works as a software Engineer in one of the top Engineering Companies in the US. Mildred Aiwanno-Ose Adepoju is currently a student of Prairie View A&M University, Computer Information Systems Department. Research Field Sheriff Adefolarin Adepoju: Sustainable Energy -1, Time-Series Forecasting and Analysis -2, Machine Learning -3, Graph-Based Neural Network Applications -4, Internet of Things (IoT -5, Ubiquitous Computing -6 Mildred Aiwanno-Ose Adepoju: Human-Computer Interaction (HCI) -1, Technology Innovation and Product Strategy -2, Cybersecurity -3, Data Governance and IT compliance -4, Software Engineering Management -5 Information & Authors Information Version history V1 Version 1 07 April 2025 Peer review timeline Published International Journal of Advanced Computer Science and Applications Version of Record 1 Jan 2025 Published Copyright This work is licensed under a Non Exclusive No Reuse License. Keywords algorithmic bias bias mitigation cognitive biases fair ai systems fairness in software development software development Authors Affiliations Sheriff Adefolarin Adepoju 0009-0006-6741-8518 [email protected] Prairie View A&M University View all articles by this author Mildred Aiwanno-Ose Adepoju Prairie View A&M University View all articles by this author Metrics & Citations Metrics Article Usage 907 views 224 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Sheriff Adefolarin Adepoju, Mildred Aiwanno-Ose Adepoju. COGNITIVE BIASES: UNDERSTANDING AND DESIGNING FAIR AI SYSTEMS FOR SOFTWARE DEVELOPMENT. Authorea . 07 April 2025. DOI: https://doi.org/10.22541/au.174406322.29071173/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.174406322.29071173/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a004f7fdbcb84193',t:'MTc3OTU0ODgzMA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();