Detecting and Reducing Bias in a High Stakes Domain

TL;DR

This paper investigates bias in deep learning models predicting aggression in social media posts by gang-involved youth, revealing unexpected reliance on stop words and proposing interpretability and bias reduction methods for high-stakes applications.

Contribution

It introduces a systematic interpretability approach, domain-annotated rationales, and bias mitigation techniques tailored for sensitive social media analysis.

Findings

Model often relies on stop words for predictions

Annotated rationales enable bias measurement

Proposed methods significantly reduce bias

Abstract

Gang-involved youth in cities such as Chicago sometimes post on social media to express their aggression towards rival gangs and previous research has demonstrated that a deep learning approach can predict aggression and loss in posts. To address the possibility of bias in this sensitive application, we developed an approach to systematically interpret the state of the art model. We found, surprisingly, that it frequently bases its predictions on stop words such as "a" or "on", an approach that could harm social media users who have no aggressive intentions. To tackle this bias, domain experts annotated the rationales, highlighting words that explain why a tweet is labeled as "aggression". These new annotations enable us to quantitatively measure how justified the model predictions are, and build models that drastically reduce bias. Our study shows that in high stake scenarios, accuracy alone cannot guarantee a good system and we need new evaluation methods.