Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Disciplining Deliberation: A Sociotechnical Perspective on Machine Learning Trade-offs

Sina Fazelpour

TL;DR

This paper examines two prominent formal trade-offs in artificial intelligence -- between predictive accuracy and fairness, and between predictive accuracy and interpretability -- and introduces a sociotechnical approach to examining the value implications of trade-offs.

Abstract

This paper examines two prominent formal trade-offs in artificial intelligence (AI) -- between predictive accuracy and fairness, and between predictive accuracy and interpretability. These trade-offs have become a central focus in normative and regulatory discussions as policymakers seek to understand the value tensions that can arise in the social adoption of AI tools. The prevailing interpretation views these formal trade-offs as directly corresponding to tensions between underlying social values, implying unavoidable conflicts between those social objectives. In this paper, I challenge that prevalent interpretation by introducing a sociotechnical approach to examining the value implications of trade-offs. Specifically, I identify three key considerations -- validity and instrumental relevance, compositionality, and dynamics -- for contextualizing and characterizing these implications. These considerations reveal that the relationship between model trade-offs and corresponding values depends on critical choices and assumptions. Crucially, judicious sacrifices in one model property for another can, in fact, promote both sets of corresponding values. The proposed sociotechnical perspective thus shows that we can and should aspire to higher epistemic and ethical possibilities than the prevalent interpretation suggests, while offering practical guidance for achieving those outcomes. Finally, I draw out the broader implications of this perspective for AI design and governance, highlighting the need to broaden normative engagement across the AI lifecycle, develop legal and auditing tools sensitive to sociotechnical considerations, and rethink the vital role and appropriate structure of interdisciplinary collaboration in fostering a responsible AI workforce.

Disciplining Deliberation: A Sociotechnical Perspective on Machine Learning Trade-offs

TL;DR

This paper examines two prominent formal trade-offs in artificial intelligence -- between predictive accuracy and fairness, and between predictive accuracy and interpretability -- and introduces a sociotechnical approach to examining the value implications of trade-offs.

Abstract

This paper examines two prominent formal trade-offs in artificial intelligence (AI) -- between predictive accuracy and fairness, and between predictive accuracy and interpretability. These trade-offs have become a central focus in normative and regulatory discussions as policymakers seek to understand the value tensions that can arise in the social adoption of AI tools. The prevailing interpretation views these formal trade-offs as directly corresponding to tensions between underlying social values, implying unavoidable conflicts between those social objectives. In this paper, I challenge that prevalent interpretation by introducing a sociotechnical approach to examining the value implications of trade-offs. Specifically, I identify three key considerations -- validity and instrumental relevance, compositionality, and dynamics -- for contextualizing and characterizing these implications. These considerations reveal that the relationship between model trade-offs and corresponding values depends on critical choices and assumptions. Crucially, judicious sacrifices in one model property for another can, in fact, promote both sets of corresponding values. The proposed sociotechnical perspective thus shows that we can and should aspire to higher epistemic and ethical possibilities than the prevalent interpretation suggests, while offering practical guidance for achieving those outcomes. Finally, I draw out the broader implications of this perspective for AI design and governance, highlighting the need to broaden normative engagement across the AI lifecycle, develop legal and auditing tools sensitive to sociotechnical considerations, and rethink the vital role and appropriate structure of interdisciplinary collaboration in fostering a responsible AI workforce.
Paper Structure (15 sections, 3 equations, 1 figure)

This paper contains 15 sections, 3 equations, 1 figure.

Table of Contents

  1. Introduction
  2. Trade-offs in AI-based Decision-making
  3. Prediction-based decision-making and the cost of constraints
  4. Fairness as a constraint
  5. Interpretability as a constraint
  6. Direct correspondence and a convenient division of labor
  7. A Sociotechnical Perspective On Interpreting the Trade-offs
  8. Validity and relevance
  9. Compositionality
  10. Deployment dynamics
  11. Broader Implications
  12. Expanding the scope of normative engagement
  13. "Business necessity", transparency, and implications for law and policy
  14. The meaning of "AI talent" and the need for interdisciplinarity
  15. Conclusion

Figures (1)

  • Figure 1: Figure (a) represents a characterization of the two trade-offs in terms of a possible relation in the hypothesis space $\mathcal{H}$ between the set of accurate models $\mathcal{H}_A$ and the set of models that satisfy an additional formal constraint $\mathcal{H}_C$. The figure is inspired by Figure 1 in dziugaite2020enforcing. (b) This figure, adapted from kearns2019ethical, offers a toy illustration of the accuracy-fairness Pareto frontier, while Figure (c), adapted from gunning2019xai (itself adapted from DARPA XAI presentation) offers a widely shared illustration of the accuracy-interpretability trade-off.