Discovering and Validating AI Errors With Crowdsourced Failure Reports

TL;DR

This paper introduces crowdsourced failure reports and a visual analytics tool called Deblinder to help developers identify, validate, and address systematic AI errors more efficiently, ultimately improving model performance.

Contribution

The paper presents a novel approach combining crowdsourced failure reports with Deblinder for systematic AI error detection and validation, demonstrated through practitioner studies.

Findings

Deblinder aids in discovering AI failures effectively.

Failure reports help validate systematic errors.

Collecting targeted data improves model performance.

Abstract

AI systems can fail to learn important behaviors, leading to real-world issues like safety concerns and biases. Discovering these systematic failures often requires significant developer attention, from hypothesizing potential edge cases to collecting evidence and validating patterns. To scale and streamline this process, we introduce crowdsourced failure reports, end-user descriptions of how or why a model failed, and show how developers can use them to detect AI errors. We also design and implement Deblinder, a visual analytics system for synthesizing failure reports that developers can use to discover and validate systematic failures. In semi-structured interviews and think-aloud studies with 10 AI practitioners, we explore the affordances of the Deblinder system and the applicability of failure reports in real-world settings. Lastly, we show how collecting additional data from the groups identified by developers can improve model performance.