Abstract AI bias is no longer a hidden statistic; it's now a single 'composite face' that anyone can see.
April 15, 2026
Original Paper
GLEaN: A Text-to-image Bias Detection Approach for Public Comprehension
arXiv · 2604.09923
The Takeaway
GLEaN distills the systemic biases of text-to-image models into a 'median' composite portrait, making algorithmic bias instantly visible. Before this, auditing bias required complex data tables and statistical knowledge that non-experts couldn't parse. Now, if a model is biased against a certain demographic, it shows up clearly in a single generated 'average' image. This turns bias auditing from a math problem into a visual one. It allows product managers and policy makers to 'see' the bias in their models without needing a PhD in statistics.
From the abstract
Text-to-image (T2I) models, and their encoded biases, increasingly shape the visual media the public encounters. While researchers have produced a rich body of work on bias measurement, auditing, and mitigation in T2I systems, those methods largely target technical stakeholders, leaving a gap in public legibility. We introduce GLEaN (Generative Likeness Evaluation at N-Scale), a portrait-based explainability pipeline designed to make T2I model biases visually understandable to a broad audience.