Beyond Predictive Fairness: Quantifying Attribution Consistency Across Demographic Groups in Diabetic Retinopathy Screening

Kerol Djoumessi; Philipp Berens

cs.LGAug 19, 2026

Beyond Predictive Fairness: Quantifying Attribution Consistency Across Demographic Groups in Diabetic Retinopathy Screening

Authors: Kerol Djoumessi, Philipp Berens

Organizations: Hertie Institute for AI in Brain Health, University of Tübingen, Germany · Tübingen AI Center, University of Tübingen, Germany

Abstract

Fairness in medical imaging is commonly evaluated through subgroup performance metrics, yet it remains unclear whether models rely on consistent visual evidence across demographic groups. This work introduces the Explanation Consistency Score (ECS), a fairness-aware metric based on Jensen-Shannon divergence that quantifies the similarity of attribution maps across subgroups. Using diabetic retinopathy screening as a case study, ECS is evaluated globally and within disease severity. Experiments reveal that while predictive performance differs across ethnic groups, explanation consistency remains relatively high and shows no significant association with performance disparities. These findings suggest that predictive fairness and explanation consistency capture complementary dimensions of model behavior, motivating fairness evaluations that extend beyond predictive performance.

Figures & tables

Ethnicity	# Samples	DR0	DR1	DR2	DR3	DR4	DR1-4	Prev.
Latin American	$53,960$	$42,155$	$5,787$	$5,502$	$285$	$231$	$11,805$	$21.9\%$
African descent	$5,097$	$3,597$	$747$	$663$	$40$	$50$	$1,500$	$29.4\%$
Indian origin	$3,341$	$1,822$	$472$	$930$	$51$	$66$	$1,519$	$45.5\%$
Caucasian	$9,542$	$7,818$	$794$	$884$	$24$	$22$	$1,724$	$18.1\%$
Asian	$3,995$	$3,335$	$321$	$314$	$4$	$21$	$660$	$16.5\%$
Not specified	$12,253$	$9,825$	$1,228$	$1,114$	$60$	$26$	$2,428$	$19.8\%$

Table 1: Demographic distribution and DR grade composition of the EyePACS cohort.

	Predictive performance			Explanation consistency
Ethnicity	AUC	Sens.	Spec.	ECS-SC	WG-ECS	ECS-SG	DR1-ECS
Latin American	$.85\pm.002$	$.57\pm.013$	$.97\pm.004$	$.92\pm.002$	$.53\pm.08$	$.88\pm.010$	$.87\pm 006$
African descent	$.88\pm.002$	$.64\pm.023$	$.97\pm.002$	$.91\pm 005$	$.56\pm.08$	$.88\pm.008$	$.85\pm.014$
Indian origin	$.92\pm.001$	$.77\pm.022$	$.94\pm.007$	$.90\pm 006$	$.57\pm.08$	$.86\pm.010$	$.85\pm 009$
Caucasian	$.83\pm.003$	$.54\pm.018$	$.97\pm.004$	$.91\pm 004$	$.57\pm.07$	$.86\pm.014$	$.85\pm 008$
Asian	$.90\pm.006$	$.68\pm.023$	$.97\pm.005$	$.89\pm.004$	$.54\pm.08$	$.85\pm.008$	$.79\pm.011$

Table 2: Predictive performance and attribution consistency across demographic groups. Results are reported as mean

\pm

standard deviation over three random seeds.

Figure 1: Explanation consistency across demographic groups. Average ScoreCAM and SmoothGradCAM++ attribution maps for correctly classified DR-positive samples across demographic groups. Similar attention patterns are observed across ethnicities, indicating high explanation consistency.

Figure 2: Conditional attribution consistency across demographic groups. Average attribution maps for DR grades 1 (top two rows) and 2 (bottom two rows) across demographic groups using ScoreCAM and SmoothGradCAM++. Attribution patterns remain broadly consistent within each disease grade, supporting the robustness of the proposed ECS analysis after controlling for disease severity.

Group Fairness

Group fairness measures and reduces disparities in how algorithms predict, rank, allocate resources, and make decisions for different groups.

See all 140 papers

MESD: A Risk-Sensitive Metric for Explanation Fairness Across Intersectional Subgroups
Mar 13, 2026Gideon Popoola, John SheppardAlgorithmic Fairness Intersectional Fairness
Gianforte School of Computing Montana State University Bozeman, MT, USA, 59715
When Fairness Metrics Disagree: Evaluating the Reliability of Demographic Fairness Assessment in Machine Learning
Apr 16, 2026Khalid Adnan AlsayedAlgorithmic Fairness Group Fairness
School of Computing, Engineering & Digital Technologies, Teesside University, UK
GESD: Beyond Outcome-Oriented Fairness
May 14, 2026Gideon Popoola, John SheppardAlgorithmic Fairness Group Fairness
Montana State University Bozeman, MT, USA

Mar 13, 2026J/K move · Enter open · S save

MESD: A Risk-Sensitive Metric for Explanation Fairness Across Intersectional Subgroups

Gideon Popoola, John Sheppard

Gianforte School of Computing Montana State University Bozeman, MT, USA, 59715

Algorithmic Fairness Intersectional Fairness Explainability Evaluation

Open paper AskPDF arXiv

Fairness in machine learning is predominantly evaluated through outcome-oriented metrics, such as Demographic parity, which measure whether predictions are statistically consistent across protected groups. However, these metrics cannot detect whether a model uses systematically different reasoning for different demographic groups, which violates procedural fairness principles. This problem is compounded by intersectionality, where models may appear fair on individual attributes (e.g., race) while exhibiting significant disparities for intersectional subgroups (e.g., race

\times

gender), a phenomenon known as fairness gerrymandering. In this work, we introduce Multi-category Explanation Stability Disparity (MESD), a procedural fairness metric that quantifies disparities in explanation quality across intersectional subgroups formed by the Cartesian product of multiple protected attributes. MESD integrates three components, which are label-aware aggregation aligned with outcome-conditional fairness, empirical-Bayes shrinkage to stabilize estimates for small intersectional groups, and Conditional Value-at-Risk (CVaR) weighting to emphasize worst-case subgroup disparities. We integrate MESD within a multi-objective optimization framework (UEF) that jointly optimizes utility, outcome fairness, and procedural fairness using NSGA-II. We evaluated MESD and UEF on three benchmark datasets along with four state-of-the-art methods in several experiments, and we demonstrate that MESD reveals procedural disparities invisible to outcome metrics alone. We position our contribution within procedural justice theory and discuss implications for regulatory compliance and intersectional equity.

Abstract

Group Fairness

Explore similar work

MESD: A Risk-Sensitive Metric for Explanation Fairness Across Intersectional Subgroups

When Fairness Metrics Disagree: Evaluating the Reliability of Demographic Fairness Assessment in Machine Learning

GESD: Beyond Outcome-Oriented Fairness