CoChat
Sign Up
Cited by 155Ethics and Social Impacts of AIExplainable Artificial Intelligence (XAI)AI and HR Technologies

AI Fairness in Data Management and Analytics: A Review on Challenges, Methodologies and Applications

Open access
Chen, Pu, Wu, Linna, Wang, Lei·2023·Applied Sciences
DOI 10.3390/app131810258
Open this paper in CoChat — free accountOpen PDF

Abstract

This article provides a comprehensive overview of the fairness issues in artificial intelligence (AI) systems, delving into its background, definition, and development process. The article explores the fairness problem in AI through practical applications and current advances and focuses on bias analysis and fairness training as key research directions. The paper explains in detail the concept, implementation, characteristics, and use cases of each method. The paper explores strategies to reduce bias and improve fairness in AI systems, reviews challenges and solutions to real-world AI fairness applications, and proposes future research directions. In addition, this study provides an in-depth comparative analysis of the various approaches, utilizing cutting-edge research information to elucidate their different characteristics, strengths, and weaknesses. The results of the comparison provide guidance for future research. The paper concludes with an overview of existing challenges in practical applications and suggests priorities and solutions for future research. The conclusions provide insights for promoting fairness in AI systems. The information reviewed in this paper is drawn from reputable sources, including leading academic journals, prominent conference proceedings, and well-established online repositories dedicated to AI fairness. However, it is important to recognize that research nuances, sample sizes, and contextual factors may create limitations that affect the generalizability of the findings.

AI snapshot

Written from the paper’s full text.

TL;DR

This review surveys AI fairness in data management and analytics, organizing the field around fairness definitions, bias analysis, fair training, and real-world applications. It synthesizes 310 selected papers from 1083 materials and compares methods, challenges, and application domains.

Key findings

  • The review selected 310 papers from 1083 materials, covering literature from as far back as 1993 to recent contributions, and structures the field around definition/problem formulation, bias analysis, fair training, and applications.
  • It identifies seven fairness definitions—individual fairness, group fairness, equality of opportunity, disparate treatment, fairness through unawareness, disparate impact, and subgroup fairness—and catalogs data, algorithmic, and user-interaction biases.
  • It compares fair-training approaches including pre-processing, in-processing, post-processing, regularization-based, and counterfactual fairness, noting recurring trade-offs between fairness and model accuracy.
  • It reviews applications in healthcare, education, criminal justice, hiring, lending/credit, online advertising, and customer service/chatbots, and cites tools such as IBM AI Fairness 360 with around 70 fairness metrics and 11 bias-reduction algorithms.

Methods

  • Systematic review using PRISMA; searched IEEE Xplore, ScienceDirect, ACM Digital Library, Springer, Wiley Library, conference proceedings, and authoritative websites, compiling 1083 materials and including 310 papers after screening.
  • Inclusion/exclusion criteria emphasized peer-reviewed academic articles, conference papers, and authoritative reports explicitly addressing AI fairness in data management and analytics, grouped by thematic affinity, methodology, and fairness challenge.
  • Comparative synthesis of methods and applications; no primary data collection or new experiments are reported.

Limitations

  • The authors state that limitations stem primarily from nuances in research methodology, sample size, and context, which may affect the generalizability of conclusions drawn from individual studies.
  • The authors also note that AI fairness is a dynamic field and that rapid change and ongoing research may limit current coverage.
  • Evident from the text: as a review, it relies on secondary sources and comparative tables rather than new empirical validation; some table content in the provided text appears garbled or duplicated, which may affect clarity.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/app131810258

Cites

255 of 255 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 162 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The review included 310 papers from an extensive pool of 1083 pieces of materials.

    “A meticulous curation process led to the inclusion of 310 papers from an extensive pool of 1083 pieces of materials.”
    Quote in the paperNot a substantial claim
  • Premisecitation

    Clear explanation of data collection purpose builds trust with participants.

    “Builds trust with participants”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [235] Rubinstein, 2015, NIH/NCATS/GRDR® Common Data Elements: A leading force for standardized data collection, Contemp. Clin. Trials source

      Not checked in the free run.

    • cites [236] Frick, 2009, Micro-costing quantity data collection methods, Med. Care source

      Not checked in the free run.

  • Premisecitation

    Some healthcare AI systems have been found to differ in the diagnosis of certain diseases among different racial groups.

    “as some healthcare AI systems have been found to differ in the diagnosis of certain diseases among different racial groups [52,278].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [52] Chen, 2023, Algorithmic fairness in Artificial Intelligence for medicine and healthcare, Nat. Biomed. Eng. source

      Not checked in the free run.

    • cites [278] Chen, R.J., Chen, T.Y., Lipkova, J., Wang, J.J., Williamson, D.F., Lu, M.Y., Sahai, S., and Mahmood, F. (2021). Algorithm fairness in ai for medicine and healthcare. arXiv.

      Not checked in the free run.

  • Premisecitation

    Implementation of fairness awareness training resulted in a significant reduction in system bias.

    “Implementation of fairness awareness training resulted in a significant reduction in system bias [283].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [283] Berk, 2021, Fairness in criminal justice risk assessments: The state of the art, Sociol. Methods Res. source

      Not checked in the free run.

  • Premisecitation

    Informed consent (obtaining explicit consent from participants) may lead to selection bias.

    “May lead to selection bias”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [232] Ferryman, K., and Pitcan, M. (2018). Fairness in Precision Medicine, Data and Society Research Institute.

      Not checked in the free run.

    • cites [233] Dempsey, 2022, Sharing begins at home: How continuous and ubiquitous FAIRness can enhance research productivity and data reuse, Harv. Data Sci. Rev.

      Not checked in the free run.

    • cites [234] Durand, 2016, Realizing HOPE: The ethics of organ transplantation from HIV-positive donors, Ann. Intern. Med. source

      Not checked in the free run.

  • Classified as probably not a substantial claim. 13
  • May combine two claims. 11

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    The feedback loop between machine learning system outputs and inputs can perpetuate and reinforce biases.

    “Moreover, the feedback loop between machine learning system outputs and inputs can perpetuate and reinforce biases, necessitating the analysis of algorithms in dynamic systems.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Some AI-based risk assessment tools in the criminal justice system have been found to be racially biased, resulting in harsher sentences for some minorities.

    “In the criminal justice system, some AI-based risk assessment tools have been found to be racially biased, resulting in harsher sentences for some minorities [282].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [282] Angwin, J., Larson, J., Mattu, S., and Kirchner, L. (2022). Ethics of Data and Analytics, Auerbach Publications.

      Not checked in the free run.

  • May combine two claims.

    Premisenone given

    Biases in AI models can arise from data collection, labeling, and partitioning, affecting data integrity.

    “Biases in AI models can arise from data collection, labeling, and partitioning, affecting data integrity.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Post-processing fairness techniques might not entirely resolve root biases.

    “While valuable in certain contexts, these techniques might not entirely resolve root biases.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    By incorporating fairness constraints, the platform achieved more balanced ad targeting across all users.

    “By incorporating fairness constraints, the platform achieved more balanced ad targeting across all users [295].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [295] Bateni, A., Chan, M.C., and Eitel-Porter, R. (2022). AI fairness: From principles to practice. arXiv.

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The use of a fairness-aware algorithm improves the performance of the model and provides a fairer diagnosis for all patients.

    “The use of a fairness-aware algorithm improves the performance of the model and provides a fairer diagnosis for all patients [279,280].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [279] Gichoya, 2021, Equity in essence: A call for operationalising fairness in machine learning for healthcare, BMJ Health Care Inform. source

      Not checked in the free run.

    • cites [280] Johnson, 2021, Precision medicine, AI, and the future of personalized health care, Clin. Transl. Sci. source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Many AI-driven recruitment tools have exhibited biases, favoring specific candidates and overlooking job requisites.

    “Many AI-driven recruitment tools have exhibited biases, favoring specific candidates and overlooking job requisites [307,308].”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [307] Hunkenschroer, 2023, Is AI recruiting (un) ethical? A human rights perspective on the use of AI for hiring, AI Ethics source

      Not checked in the free run.

    • cites [308] Dastin, J. (2022). Ethics of Data and Analytics, Auerbach Publications.

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    There is no mature standard for quantifying the risk of AI fairness and little insight into how to make decisions amid consensus and controversy with commonly accepted solutions to the risks.

    “Although the above work provides tools and theoretical analysis, due to the short research time of this problem and still in the preliminary exploratory stage, there is no mature standard for how to quantify the risk of AI fairness and little insights in how to make decisions in the case of consensus and controversy with the commonly accepted solutions to the risks.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Informed consent (obtaining explicit consent from participants) respects individual autonomy.

    “Respects individual autonomy”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [232] Ferryman, K., and Pitcan, M. (2018). Fairness in Precision Medicine, Data and Society Research Institute.

      Not checked in the free run.

    • cites [233] Dempsey, 2022, Sharing begins at home: How continuous and ubiquitous FAIRness can enhance research productivity and data reuse, Harv. Data Sci. Rev.

      Not checked in the free run.

    • cites [234] Durand, 2016, Realizing HOPE: The ethics of organ transplantation from HIV-positive donors, Ann. Intern. Med. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The challenge in fair training lies in striking a balance between fairness and accuracy without compromising model performance.

    “The challenge lies in striking a balance between fairness and accuracy, avoiding the compromise of model performance while enhancing fairness.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    The review adopted PRISMA, which outlines the systematic progression of study identification, screening, eligibility, and inclusion phases, increasing reproducibility and rigor.

    “We adopted PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) in our review, which meticulously outlines the systematic progression of the study identification, screening, eligibility, and inclusion phases, thereby increasing the reproducibility and rigor of the review process.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    Limitations of the evidence from the reviewed sources stem primarily from nuances in research methodology, sample size, and context, which may affect the generalizability of conclusions drawn from individual studies.

    “These limitations stem primarily from nuances in research methodology, sample size, and context, which may affect the generalizability of the conclusions drawn from individual studies.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Implementing bias mitigation techniques enhances model fairness, leading to more impartial lending determinations.

    “Implementing bias mitigation techniques enhances model fairness, leading to more impartial lending determinations [290-292].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [290] Hiller, 2020, Fairness in the eyes of the beholder: Ai; fairness; and alternative credit scoring, W. Va. L. Rev.

      Not checked in the free run.

    • cites [291] Kumar, I.E., Hines, K.E., and Dickerson, J.P. (2021, January 19–21). Equalizing credit opportunity in algorithms: Aligning algorithmic fairness research with us fair lending regulation. Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society, Oxford, UK. source

      Not checked in the free run.

    • cites [292] Moldovan, 2023, Algorithmic decision making methods for fair credit scoring, IEEE Access source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Data anonymization, aggregation, and de-identification protect participant privacy.

    “Protects participant privacy”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [240] Elliot, M., Mackey, E., O’Hara, K., and Tudor, C. (2016). The Anonymisation Decision-Making Framework, UKAN.

      Not checked in the free run.

    • cites [241] Rosner, 2019, De-Identification as Public Policy, J. Data Prot. Priv.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Encryption, access controls, and security audits protect data from unauthorized access.

    “Protects data from unauthorized access”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [252] Force, 2013, Security and privacy controls for federal information systems and organizations, NIST Spec. Publ.

      Not checked in the free run.

    • cites [253] Plans, 2014, Assessing security and privacy controls in federal information systems and organizations, NIST Spec. Publ.

      Not checked in the free run.

    • cites [254] Dempsey, K., Witte, G., and Rike, D. (2014). Summary of NIST SP 800-53, Revision 4: Security and Privacy Controls for Federal Information Systems and Organizations, Technical Report. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Encryption, access controls, and security audits have implementation costs.

    “Implementation costs”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [252] Force, 2013, Security and privacy controls for federal information systems and organizations, NIST Spec. Publ.

      Not checked in the free run.

    • cites [253] Plans, 2014, Assessing security and privacy controls in federal information systems and organizations, NIST Spec. Publ.

      Not checked in the free run.

    • cites [254] Dempsey, K., Witte, G., and Rike, D. (2014). Summary of NIST SP 800-53, Revision 4: Security and Privacy Controls for Federal Information Systems and Organizations, Technical Report. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Disparate impact analysis measures disparate impact ratios.

    “Measures disparate impact ratios”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [6] Barocas, 2016, Big data’s disparate impact, Calif. Law Rev.

      Not checked in the free run.

    • cites [257] Carle, 2011, A social movement history of Title VII Disparate Impact analysis, Fla. L. Rev.

      Not checked in the free run.

    • cites [258] Griffith, 2021, Using Disparate Impact Analysis to Develop Anti-Racist Policies: An Application to Coronavirus Liability Waivers, J. High. Educ. Manag.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Bias detection techniques identify biases in data or model predictions.

    “Identifies biases in data or model predictions”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [71] Bolukbasi, T., Chang, K.W., Zou, J.Y., Saligrama, V., and Kalai, A.T. (2016). Man is to computer programmer as woman is to homemaker? debiasing word embeddings. Adv. Neural Inf. Process. Syst., 29.

      Not checked in the free run.

    • cites [262] Serna, I., DeAlcala, D., Morales, A., Fierrez, J., and Ortega-Garcia, J. (2021). IFBiD: Inference-free bias detection. arXiv.

      Not checked in the free run.

    • cites [263] Li, B., Peng, H., Sainju, R., Yang, J., Yang, L., Liang, Y., Jiang, W., Wang, B., Liu, H., and Ding, C. (2021). Detecting gender bias in transformer-based models: A case study on BERT. arXiv.

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Disparate impact analysis is easy to implement and interpret.

    “Easy to implement and interpret”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [6] Barocas, 2016, Big data’s disparate impact, Calif. Law Rev.

      Not checked in the free run.

    • cites [257] Carle, 2011, A social movement history of Title VII Disparate Impact analysis, Fla. L. Rev.

      Not checked in the free run.

    • cites [258] Griffith, 2021, Using Disparate Impact Analysis to Develop Anti-Racist Policies: An Application to Coronavirus Liability Waivers, J. High. Educ. Manag.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Fairness-aware performance metrics simultaneously evaluate accuracy and fairness.

    “Simultaneously evaluates accuracy and fairness”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [259] Liu, S., Ge, Y., Xu, S., Zhang, Y., and Marian, A. (2022, January 18–22). Fairness-aware federated matrix factorization. Proceedings of the 16th ACM Conference on Recommender Systems, Seattle, WA, USA. source

      Not checked in the free run.

    • cites [260] Gao, 2022, FAIR: Fairness-aware information retrieval evaluation, J. Assoc. Inf. Sci. Technol. source

      Not checked in the free run.

    • cites [261] Zhang, W., and Ntoutsi, E. (2019). Faht: An adaptive fairness-aware decision tree classifier. arXiv. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Continual bias monitoring is ongoing and regular assessment.

    “Ongoing and regular assessment”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [47] Kearns, M., Neel, S., Roth, A., and Wu, Z.S. (2018, January 10–15). Preventing fairness gerrymandering: Auditing and learning for subgroup fairness. Proceedings of the International Conference on Machine Learning, PMLR, Stockholm, Sweden.

      Not checked in the free run.

    • cites [271] Jeffries, 2020, Athlete-reported outcome measures for monitoring training responses: A systematic review of risk of bias and measurement property quality according to the COSMIN guidelines, Int. J. Sport. Physiol. Perform. source

      Not checked in the free run.

    • cites [272] Correia, 2022, Assessing data bias in visual surveys from a cetacean monitoring programme, Sci. Data source

      Not checked in the free run.

21 of 21

  • The paper relies on this without support.

    AI algorithms in college admissions and resource allocation may unintentionally perpetuate biases, impacting opportunities and diversity.

    “AI algorithms in college admissions and resource allocation may unintentionally perpetuate biases, impacting opportunities and diversity.”

  • The paper relies on this without support.

    Personalized learning platforms might worsen educational disparities, particularly for marginalized students.

    “Personalized learning platforms might worsen educational disparities, particularly for marginalized students.”

  • The paper relies on this without support.

    Testing and assessment bias can lead to unfair evaluations, affecting self-esteem and prospects.

    “Testing and assessment bias can lead to unfair evaluations, affecting self-esteem and prospects.”

  • The paper relies on this without support.

    Data-driven disparities may lead to biased arrests, bail decisions, and sentencing.

    “Data-driven disparities may lead to biased arrests, bail decisions, and sentencing.”

  • The paper relies on this without support.

    Biased recidivism prediction algorithms misclassify groups as high risk, perpetuating unfair treatment and higher incarceration rates for marginalized groups.

    “Biased recidivism prediction algorithms misclassify groups as high risk, perpetuating unfair treatment and higher incarceration rates for marginalized groups.”

  • The paper relies on this without support.

    AI algorithms in advertising may unintentionally yield biased ad targeting and content due to biased training data and content generation.

    “AI algorithms in advertising may unintentionally yield biased ad targeting and content due to biased training data and content generation.”

  • The paper relies on this without support.

    Chatbots often unknowingly offer biased or offensive replies due to training data exposure, leading to customer dissatisfaction and reputational harm.

    “Chatbots often unknowingly offer biased or offensive replies due to training data exposure, leading to customer dissatisfaction and reputational harm.”

  • The paper relies on this without support.

    Limitations in comprehending diverse linguistic expressions hinder accurate responses to various language forms, including slang.

    “Additionally, limitations in comprehending diverse linguistic expressions hinder accurate responses to various language forms, including slang.”

  • The paper relies on this without support.

    Inadequacies in addressing sensitive topics and emotional responses lead to inappropriate interactions in some customer service bots.

    “Further, inadequacies in addressing sensitive topics and emotional responses lead to inappropriate interactions in some customer service bots.”

  • The paper relies on this without support.

    AI applications in healthcare face challenges of interpretability, trust, data privacy, and security.

    “AI applications in healthcare face challenges of interpretability, trust, data privacy, and security.”

  • The paper relies on this without support.

    Privacy concerns hinder data sharing, while complex AI models lack interpretability, impacting trust.

    “Privacy concerns hinder data sharing, while complex AI models lack interpretability, impacting trust.”

  • The paper relies on this without support.

    AI support can enhance ad relevance and user experience, while also mitigating inappropriate content.

    “It can be seen that the AI support can enhance ad relevance and user experience, while also mitigating inappropriate content.”

  • The paper relies on this without support.

    AI applications use AI algorithms to analyze various data sources, such as financial records, transaction histories, and alternative data, to make more accurate and efficient lending decisions.

    “These applications use AI algorithms to analyze various data sources, such as financial records, transaction histories, and alternative data, to make more accurate and efficient lending decisions.”

  • The paper relies on this without support.

    AI aids in automating and optimizing the loan approval process, reducing human bias, and increasing access to credit for underserved populations.

    “This aids in automating and optimizing the loan approval process, reducing human bias, and increasing access to credit for underserved populations.”

  • The paper relies on this without support.

    AI assists in fraud detection, predicting default risks, and personalizing loan terms based on individual borrower profiles.

    “AI assists in fraud detection, predicting default risks, and personalizing loan terms based on individual borrower profiles.”

  • The paper relies on this without support.

    AI-driven algorithms streamline CV screening and candidate profiling, while proctored assessments ensure secure remote testing.

    “AI-driven algorithms streamline CV screening and candidate profiling, while proctored assessments ensure secure remote testing.”

  • The paper relies on this without support.

    AI optimizes interview scheduling and personalizes HR training.

    “AI optimizes interview scheduling and personalizes HR training.”

  • The paper relies on this without support.

    Behavior tracking and personality analysis provide insights into candidate dynamics, and AI aids in appraisal monitoring through performance metric analysis.

    “Behavior tracking and personality analysis provide insights into candidate dynamics, and AI aids in appraisal monitoring through performance metric analysis.”

  • The paper relies on this without support.

    These applications collectively reshape HRM practices, enhancing efficiency and informed decision making.

    “These applications collectively reshape HRM practices, enhancing efficiency and informed decision making.”

  • The paper relies on this without support.

    AI systems have a substantial role in online advertising, including targeted ad delivery, content censorship and related design.

    “AI systems have a substantial role in online advertising, including targeted ad delivery, content censorship and related design.”

  • The paper relies on this without support.

    AI technologies both facilitate and necessitate ongoing efforts to maintain fairness and effectiveness in online advertising practices.

    “In this context, AI technologies both facilitate and necessitate ongoing efforts to maintain fairness and effectiveness in online advertising practices.”
21 of 21

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Included-study count differs from the flowchart

    Minor

    The text says 310 papers were included, but the final stage of the PRISMA flowchart totals 311 materials. The text does not explain the difference.

  • Training-data resampling is classified as post-processing

    Minor

    The paper defines post-processing as adjusting predictions after training, but Table 8 classifies a method that modifies the training-data distribution as post-processing.

  • Figure 4 caption identifies the wrong application

    Minor

    The text describes Figure 4 as depicting loan and credit decision making, while its caption identifies it as a hiring and recruiting figure.

  • Data-bias category count does not match the list

    Minor

    The sentence claims five main types of data bias but names twelve.

  • Training-time regularization is classified as post-processing

    Minor

    The paper defines post-processing as adjustment after predictions are made, but Table 8 describes its regularization post-processing method as applying constraints during training.

  • Post-processing method appears in the in-processing table

    Minor

    Table 7 classifies equalized odds post-processing under in-processing fairness, whereas Table 8 classifies the same method under post-processing fairness.

6 of 6

Cites

Flagged first, then by citations

41 of the 296 references couldn’t be matched to a known paper.

Research works printed but not linked: 47

  1. Frincke, D., Tobin, D., McConnell, J., Marconi, J., and Polla, D. (1998, January 2–8). A framework for cooperative intrusion detection. Proceedings of the 21st NIST-NCSC National Information Systems Security Conference, Arlington, VA, USA.
  2. Estivill-Castro, V., and Brankovic, L. (1999). International Conference on Data Warehousing and Knowledge Discovery, Springer.
  3. Kusner, M.J., Loftus, J., Russell, C., and Silva, R. (2017). Counterfactual fairness. Adv. Neural Inf. Process. Syst., 30.
  4. Dwork, C., and Ilvento, C. (2018, January 23–24). Group fairness under composition. Proceedings of the 2018 Conference on Fairness, Accountability, and Transparency (FAT* 2018), New York, NY, USA.
  5. Liao, Y., and Naghizadeh, P. (2022, January 25–29). The impacts of labeling biases on fairness criteria. Proceedings of the 10th International Conference on Learning Representations, ICLR, Virtually.
  6. Weerts, H., Pfisterer, F., Feurer, M., Eggensperger, K., Bergman, E., Awad, N., Vanschoren, J., Pechenizkiy, M., Bischl, B., and Hutter, F. (2023). Can Fairness be Automated? Guidelines and Opportunities for Fairness-aware AutoML. arXiv.
  7. Chai, J., and Wang, X. (2022, January 17–23). Fairness with adaptive weights. Proceedings of the International Conference on Machine Learning, PMLR, Baltimore, MD, USA.
  8. Kleinberg, J., Mullainathan, S., and Raghavan, M. (2016). Inherent trade-offs in the fair determination of risk scores. arXiv.
  9. Pan, W., Cui, S., Wen, H., Chen, K., Zhang, C., and Wang, F. (2021). Correcting the user feedback-loop bias for recommendation systems. arXiv.
  10. Schemmer, M., Kühl, N., Benz, C., and Satzger, G. (2022). On the influence of explainable AI on automation bias. arXiv.
  11. Calmon, F., Wei, D., Vinzamuri, B., Natesan Ramamurthy, K., and Varshney, K.R. (2017). Optimized pre-processing for discrimination prevention. Adv. Neural Inf. Process. Syst., 30.
  12. Petersen, 2021, Post-processing for individual fairness, Adv. Neural Inf. Process. Syst., 34, 25944
  13. Putzel, P., and Lee, S. (2022). Blackbox post-processing for multiclass fairness. arXiv.
  14. Jung, S., Park, T., Chun, S., and Moon, T. (2023). Re-weighting Based Group Fairness Regularization via Classwise Robust Optimization. arXiv.
  15. Lal, G.R., Geyik, S.C., and Kenthapadi, K. (2020). Fairness-aware online personalization. arXiv.
  16. Cheong, J., Kalkan, S., and Gunes, H. (2022). European Conference on Computer Vision, Springer.
  17. Xie, S.M., Santurkar, S., Ma, T., and Liang, P. (2023). Data selection for language models via importance resampling. arXiv.
  18. Le Quy, T., Friege, G., and Ntoutsi, E. (2023). Educational Data Science: Essentials, Approaches, and Tendencies: Proactive Education based on Empirical Big Data Evidence, Springer.
  19. Chierichetti, F., Kumar, R., Lattanzi, S., and Vassilvitskii, S. (2017). Fair clustering through fairlets. Adv. Neural Inf. Process. Syst., 30.
  20. Wong, E., Rice, L., and Kolter, J.Z. (2020). Fast is better than free: Revisiting adversarial training. arXiv.
  21. Andriushchenko, 2020, Understanding and improving fast adversarial training, Adv. Neural Inf. Process. Syst., 33, 16048
  22. Shafahi, A., Najibi, M., Ghiasi, M.A., Xu, Z., Dickerson, J., Studer, C., Davis, L.S., Taylor, G., and Goldstein, T. (2019). Adversarial training for free!. Adv. Neural Inf. Process. Syst., 32.
  23. Darlow, L., Jastrzębski, S., and Storkey, A. (2020). Latent adversarial debiasing: Mitigating collider bias in deep neural networks. arXiv.
  24. Roy, S., and Salimi, B. (2023). Reasoning Web. Causality, Explanations and Declarative Knowledge: 18th International Summer School 2022, Berlin, Germany, 27–30 September 2022, Springer.
  25. Lundin, 2022, Involving teachers in meta-design of AI to ensure situated fairness, Proceedings, 1613, 0073
  26. Hsu, B., Chen, X., Han, Y., Namkoong, H., and Basu, K. (2023). An Operational Perspective to Fairness Interventions: Where and How to Intervene. arXiv.
  27. Salvador, T., Cairns, S., Voleti, V., Marshall, N., and Oberman, A. (2021). Faircal: Fairness calibration for face verification. arXiv.
  28. Pleiss, G., Raghavan, M., Wu, F., Kleinberg, J., and Weinberger, K.Q. (2017). On fairness and calibration. Adv. Neural Inf. Process. Syst., 30.
  29. Kasmi, M.L. (2021). Machine Learning Fairness in Finance: An Application to Credit Scoring. [Ph.D. Thesis, Tilburg University].
  30. Caton, S., and Haas, C. (2020). Fairness in machine learning: A survey. arXiv.
  31. Small, E.A., Sokol, K., Manning, D., Salim, F.D., and Chan, J. (2023). Equalised Odds is not Equal Individual Odds: Post-processing for Group and Individual Fairness. arXiv.
  32. Errica, F., Bacciu, D., and Micheli, A. (2021, January 18–24). Graph mixture density networks. Proceedings of the International Conference on Machine Learning, PMLR, Virtual.
  33. Han, X., Baldwin, T., and Cohn, T. (2022). Towards equal opportunity fairness through adversarial learning. arXiv.
  34. Veitch, 2021, Counterfactual invariance to spurious correlations in text classification, Adv. Neural Inf. Process. Syst., 34, 16196
  35. Meister, M., Sheikholeslami, S., Andersson, R., Ormenisan, A.A., and Dowling, J. (2020, January 2–4). Towards distribution transparency for supervised ML with oblivious training functions. Proceedings of the Workshop MLOps Syst, Austin, TX, USA.
  36. Makhlouf, K., Zhioua, S., and Palamidessi, C. (2020). Survey on causal-based machine learning fairness notions. arXiv.
  37. Gölz, P., Kahng, A., and Procaccia, A.D. (2019). Paradoxes in fair machine learning. Adv. Neural Inf. Process. Syst., 32.
  38. Griffith, 2021, Using Disparate Impact Analysis to Develop Anti-Racist Policies: An Application to Coronavirus Liability Waivers, J. High. Educ. Manag., 36, 104
  39. Serna, I., DeAlcala, D., Morales, A., Fierrez, J., and Ortega-Garcia, J. (2021). IFBiD: Inference-free bias detection. arXiv.
  40. Li, B., Peng, H., Sainju, R., Yang, J., Yang, L., Liang, Y., Jiang, W., Wang, B., Liu, H., and Ding, C. (2021). Detecting gender bias in transformer-based models: A case study on BERT. arXiv.
  41. Kizilcec, R.F., and Lee, H. (2022). The Ethics of Artificial Intelligence in Education, Routledge.
  42. Fenu, G., Galici, R., and Marras, M. (2022). International Conference on Artificial Intelligence in Education, Springer.
  43. Chen, R.J., Chen, T.Y., Lipkova, J., Wang, J.J., Williamson, D.F., Lu, M.Y., Sahai, S., and Mahmood, F. (2021). Algorithm fairness in ai for medicine and healthcare. arXiv.
  44. Nugent, S.E., and Scott-Parker, S. (2022). Towards Trustworthy Artificial Intelligent Systems, Springer.
  45. Hiller, 2020, Fairness in the eyes of the beholder: Ai; fairness; and alternative credit scoring, W. Va. L. Rev., 123, 907
  46. Yuan, D. (2023). Artificial Intelligence, Fairness and Productivity. [Ph.D. Thesis, University of Pittsburgh].
  47. Bateni, A., Chan, M.C., and Eitel-Porter, R. (2022). AI fairness: From principles to practice. arXiv.

Cited by

Sorted by citations

Related

Authors
Chen, P., Wu, L., & Wang, L.
Affiliations
  • PLA Army Engineering University (CN)
  • Aerospace System Engineering Shanghai, Shanghai 201109, China
Corresponding author
Wang, Lei
ORCID
Venue
Applied Sciences
Publisher
MDPI AG
Published
2023-09-13
Volume
13
Issue
18
Pages
10258
ISSN
2076-3417 (ISSN-L)
DOI
10.3390/app131810258
OpenAlex
W4386712902
Semantic Scholar
e9a30a36c5d28fde8c7b159ed8adddf85f92ff1d
Cited by
155
References
296
Citations by year
2026: 37 · 2025: 74 · 2024: 42 · 2023: 2
FWCI
38.87
Type
journal article
Language
en
Open access
Licence
Locations
  • Applied Sciences (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
  • Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
  • Social Sciences › Business, Management and Accounting › Organizational Behavior and Human Resource Management › AI and HR Technologies
Keywords
algorithmic fairness, algorithmic bias, bias assessment, bias mitigation, data management, data analytics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review