Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Structural Interventions and the Dynamics of Inequality

Aurora Zhang, Annette Hosoi

TL;DR

The paper addresses persistent inequality in algorithmic decision-making by proposing a formal dynamic model of mortgage lending with two groups and a universal decision threshold. It analyzes long-run outcomes, proving that fixed thresholds often fail to eliminate disparities and introducing three structural interventions to modulate downstream effects, evaluated via a policymaker's equity-efficiency utility $U_t$ with parameter $\alpha$; an empirical demonstration using HMDA and FNMA data illustrates how intervention choices depend on initial conditions and resource constraints. The contributions include (i) characterization of conditions under which threshold policies cannot achieve parity, (ii) a tripartite intervention taxonomy (\beta-only, group-blind, group-conscious) and a utility-based framework to compare them, and (iii) a data-driven illustration of how structural changes can outperform purely technical fairness fixes in housing markets. The work emphasizes pairing technical fairness with context-sensitive structural policy changes to achieve durable, equitable, and efficient social outcomes.

Abstract

Recent conversations in the algorithmic fairness literature have raised several concerns with standard conceptions of fairness. First, constraining predictive algorithms to satisfy fairness benchmarks may lead to non-optimal outcomes for disadvantaged groups. Second, technical interventions are often ineffective by themselves, especially when divorced from an understanding of structural processes that generate social inequality. Inspired by both these critiques, we construct a common decision-making model, using mortgage loans as a running example. We show that under some conditions, any choice of decision threshold will inevitably perpetuate existing disparities in financial stability unless one deviates from the Pareto optimal policy. Then, we model the effects of three different types of interventions. We show how different interventions are recommended depending upon the difficulty of enacting structural change upon external parameters and depending upon the policymaker's preferences for equity or efficiency. Counterintuitively, we demonstrate that preferences for efficiency over equity may lead to recommendations for interventions that target the under-resourced group. Finally, we simulate the effects of interventions on a dataset that combines HMDA and Fannie Mae loan data. This research highlights the ways that structural inequality can be perpetuated by seemingly unbiased decision mechanisms, and it shows that in many situations, technical solutions must be paired with external, context-aware interventions to enact social change.

Structural Interventions and the Dynamics of Inequality

TL;DR

The paper addresses persistent inequality in algorithmic decision-making by proposing a formal dynamic model of mortgage lending with two groups and a universal decision threshold. It analyzes long-run outcomes, proving that fixed thresholds often fail to eliminate disparities and introducing three structural interventions to modulate downstream effects, evaluated via a policymaker's equity-efficiency utility with parameter ; an empirical demonstration using HMDA and FNMA data illustrates how intervention choices depend on initial conditions and resource constraints. The contributions include (i) characterization of conditions under which threshold policies cannot achieve parity, (ii) a tripartite intervention taxonomy (\beta-only, group-blind, group-conscious) and a utility-based framework to compare them, and (iii) a data-driven illustration of how structural changes can outperform purely technical fairness fixes in housing markets. The work emphasizes pairing technical fairness with context-sensitive structural policy changes to achieve durable, equitable, and efficient social outcomes.

Abstract

Recent conversations in the algorithmic fairness literature have raised several concerns with standard conceptions of fairness. First, constraining predictive algorithms to satisfy fairness benchmarks may lead to non-optimal outcomes for disadvantaged groups. Second, technical interventions are often ineffective by themselves, especially when divorced from an understanding of structural processes that generate social inequality. Inspired by both these critiques, we construct a common decision-making model, using mortgage loans as a running example. We show that under some conditions, any choice of decision threshold will inevitably perpetuate existing disparities in financial stability unless one deviates from the Pareto optimal policy. Then, we model the effects of three different types of interventions. We show how different interventions are recommended depending upon the difficulty of enacting structural change upon external parameters and depending upon the policymaker's preferences for equity or efficiency. Counterintuitively, we demonstrate that preferences for efficiency over equity may lead to recommendations for interventions that target the under-resourced group. Finally, we simulate the effects of interventions on a dataset that combines HMDA and Fannie Mae loan data. This research highlights the ways that structural inequality can be perpetuated by seemingly unbiased decision mechanisms, and it shows that in many situations, technical solutions must be paired with external, context-aware interventions to enact social change.
Paper Structure (15 sections, 8 theorems, 17 equations, 6 figures, 1 table)

This paper contains 15 sections, 8 theorems, 17 equations, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background
  3. Fairness and the consequentialist and structural critique
  4. Related Work
  5. Model
  6. Single Time Step Model
  7. Long Term Behavior
  8. Interventions
  9. Empirical Demonstration
  10. Results
  11. Discussion and Policy Recommendations
  12. Conclusion
  13. Ethics and Positionality Statement
  14. Proofs
  15. Empirical

Key Result

proposition 1

Assume that for a given threshold $\beta \in [0,1]$, $\pi^A_t$ dominates $\pi^D_t$ for the interval $[\beta, 1]$, and that $\mu^A_t \geq \mu^D_t$. Then if $P(0 < \pi^A_{t+1},\pi^D_{t+1} < 1) = 1$, then $\mu^A_{t+1} - \mu^A_{t} \geq \mu^D_{t+1} - \mu^D_{t}$.

Figures (6)

  • Figure 1: Structural interventions in the algorithmic decision-making pipeline change the effects of decisions on downstream consequences.
  • Figure 2: The top graph displays the distribution of $\pi^i_0$ for the advantaged (pink) population and the disadvantaged (blue) population. The three bottom graphs depict the recommended intervention for different levels of $\alpha$. The y-axis is $r$, as described in the earlier sections that enumerate each policy. We choose the policy that maximizes $\lim_{t \to \infty} U_t$. Higher opacity levels correspond to a stronger recommendation for a certain intervention over the other two (all effect sizes are scaled to 1). The lower left coordinate of each box is the anchor point for each box; for example, a yellow square with lower left coordinate (a,b) means that at starting $c=a$ and intervention % $r = b$, the group-blind intervention is recommended.
  • Figure 3: Each quadrant depicts starting distributions and recommended interventions at $\alpha=0.2,0.5,0.8$. Blue corresponds to the group-conscious intervention, yellow corresponds to the group-blind, and gray corresponds to beta-only. The lower left coordinate of each box is the anchor point for each box; for example, a blue square with lower left coordinate (a,b) means that at starting $c=a$ and intervention % $r=b$, the group-conscious intervention is recommended. Opacity corresponds to the marginal utility of this intervention over the next best intervention type.
  • Figure 4: The top panel shows the empirical kernel density estimate of the distribution predicted risk of late payment in the HMDA Massachusetts 2021 data. The bottom three graphs show the recommended interventions for different $\alpha$, $c$, and intervention size.
  • Figure 5: Empirical kernel density estimate of the distribution predicted risk of late payment in the HMDA Massachusetts 2021 data.
  • ...and 1 more figures

Theorems & Definitions (14)

  • definition 1
  • proposition 1
  • corollary 1
  • proposition 2
  • proposition 3
  • definition 2
  • proposition 4
  • proof
  • proposition 5
  • proof
  • ...and 4 more