Paper: arXiv 2403.10652

Authors: Cecilia Ying, Stephen Thomas

Abstract

In an effort to improve the accuracy of credit lending decisions, many financial intuitions are now using predictions from machine learning models. While such predictions enjoy many advantages, recent research has shown that the predictions have the potential to be biased and unfair towards certain subgroups of the population. To combat this, several techniques have been introduced to help remove the bias and improve the overall fairness of the predictions. We introduce a new fairness technique, called \textit{Subgroup Threshold Optimizer} (\textit{STO}), that does not require any alternations to the input training data nor does it require any changes to the underlying machine learning algorithm, and thus can be used with any existing machine learning pipeline. STO works by optimizing the classification thresholds for individual subgroups in order to minimize the overall discrimination score between them. Our experiments on a real-world credit lending dataset show that STO can reduce gender discrimination by over 90%.

Complexity vs Empirical Score

  • Math Complexity: 4.5/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper presents a practical post-processing technique (STO) to reduce machine learning bias in credit lending by optimizing group-specific decision thresholds. While the mathematical formulations are straightforward rather than deep, the approach offers strong utility and clear readability for financial applications.

Research Flowchart

  flowchart TD
    A["Research Goal: Improve Credit Model Fairness<br/>Without Altering Data or Algorithms"] --> B["Input: Real-World Credit Lending Dataset<br/>(Subgroup & Gender Data)"]
    B --> C["Baseline ML Model Pipeline<br/>(Generates Standard Risk Predictions)"]
    C --> D["Subgroup Threshold Optimizer (STO)<br/>(Post-Processing Step)"]
    D --> E["Optimize Individual Subgroup Classification Thresholds<br/>(Minimize Overall Discrimination Score)"]
    E --> F["Outcome: >90% Reduction in Gender Discrimination<br/>(Maintains Compatibility with Existing Pipelines)"]

Cited References (27)

  1. A comparative study of fairness-enhancing interventions in machine learning (2019)
  2. The measure and mismeasure of fairness: A critical review of fair machine learning (2018) — link
  3. EU anti-discrimination law (2012)
  4. Man is to Computer Programmer as Woman is to Homemaker ? Debiasing Word Embeddings (2016) — link
  5. Unsupervised K-means clustering algorithm (2020)
  6. General Data Protection Regulation ( GDPR ) – Official Legal Text , https://gdpr-info.eu/
  7. Counterfactual explanations without opening the black box: Automated decisions and the GDPR (2017)
  8. How do fairness definitions fare? Testing public attitudes towards three algorithmic definitions of fairness in loan allocations (2020) — link
and 19 more
  1. Designing Committees for Mitigating Biases (2020) — link
  2. Protecting the Protected Group : Circumventing Harmful Fairness (2021) — link
  3. Preventing Fairness Gerrymandering : Auditing and Learning for Subgroup Fairness (2018)
  4. Fairness Through Computationally - Bounded Awareness (2018) — link
  5. Learning Non - Discriminatory Predictors (2017) — link
  6. Optimized Pre - Processing for Discrimination Prevention (2017) — link
  7. Chartered banks: Classification of non-mortgage loans (formerly C7 ) https://www.bankofcanada.ca/rates/banking-and-financial-statistics/chartered-banks-classification-of-non-mortgage-loans-formerly-c7/ (2021)
  8. A snapshot of the frontiers of fairness in machine learning (2020) — link
  9. Data preprocessing techniques for classification without discrimination (2012) — link
  10. https://www.homecredit.net/about-us/our-vision-and-business-model.aspx
  11. Home Credit Default Risk , https://kaggle.com/c/home-credit-default-risk
  12. A Framework for Understanding Unintended Consequences of Machine Learning (2020) — link
  13. Fairness through awareness (2012) — link
  14. Equality of Opportunity in Supervised Learning (2016) — link
  15. A Survey on Bias and Fairness in Machine Learning (2021) — link
  16. Amazon scraps secret AI recruiting tool that showed bias against women (2018) — link
  17. Apple Card Investigated After Gender Discrimination Complaints (2019) — link
  18. Anti-discrimination Laws, AI , and Gender Bias in Non-mortgage Fintech Lending (2020) — link
  19. AI and Machine Learning - Based Credit Underwriting and Adverse Action under the ECOA (2019) — link

0 of 27 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.