The Impact of Pseudo-Science in Financial Loans Risk Prediction

2025-07-21

AI Summary

This paper investigates the coupled effect of pseudoscientific modeling assumptions and survivorship bias on fairness degradation in financial loan risk prediction. Using mainstream machine learning models trained on real-world lending data, we conduct longitudinal dynamic modeling and bias trajectory analysis. Results reveal that localized improvements in recall or precision—while overall accuracy remains stable—can mask systemic fairness deterioration; survivorship bias systematically underestimates default risk, disproportionately increasing false rejections among disadvantaged groups, with unfairness compounding over time. We propose a novel “performance–fairness trade-off” framework, demonstrating that conventional metric improvements may induce a dangerous illusion of progress. Our core contribution is the first empirical identification of the synergistic bias mechanism between pseudoscientific assumptions (e.g., stationarity, independence) and survivorship bias in credit scoring. We further advocate integrating societal cost as a fundamental dimension in model evaluation.

Technology Category

• Machine Learning: Ethics, Bias, and Fairness
• Computer Vision: Bias, Fairness & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments

Abstract

We study the societal impact of pseudo-scientific assumptions for predicting the behavior of people in a straightforward application of machine learning to risk prediction in financial lending. This use case also exemplifies the impact of survival bias in loan return prediction. We analyze the models in terms of their accuracy and social cost, showing that the socially optimal model may not imply a significant accuracy loss for this downstream task. Our results are verified for commonly used learning methods and datasets. Our findings also show that there is a natural dynamic when training models that suffer survival bias where accuracy slightly deteriorates, and whose recall and precision improves with time. These results act as an illusion, leading the observer to believe that the system is getting better, when in fact the model is suffering from increasingly more unfairness and survival bias.

Problem

• Analyzing pseudo-scientific assumptions in financial loan risk prediction
• Investigating survival bias impact on loan return prediction accuracy
• Examining trade-offs between model accuracy and social fairness

Innovation

• Analyzing pseudo-scientific assumptions in risk prediction
• Evaluating social cost versus model accuracy trade-offs
• Detecting survival bias dynamics in loan prediction models