Where Does My Model Underperform? A Human Evaluation of Slice Discovery Algorithms
This work addresses the challenge for ML practitioners in identifying model biases or safety issues in domains lacking group annotations, though it is incremental in evaluating existing tools rather than introducing new methods.
The study tackled the problem of machine learning models underperforming on specific data slices by evaluating whether slice discovery algorithms help humans form correct hypotheses about these underperformances, finding that these tools provide some benefit over a naive baseline in a user study with 15 participants and 40 slices.
Machine learning (ML) models that achieve high average accuracy can still underperform on semantically coherent subsets ("slices") of data. This behavior can have significant societal consequences for the safety or bias of the model in deployment, but identifying these underperforming slices can be difficult in practice, especially in domains where practitioners lack access to group annotations to define coherent subsets of their data. Motivated by these challenges, ML researchers have developed new slice discovery algorithms that aim to group together coherent and high-error subsets of data. However, there has been little evaluation focused on whether these tools help humans form correct hypotheses about where (for which groups) their model underperforms. We conduct a controlled user study (N = 15) where we show 40 slices output by two state-of-the-art slice discovery algorithms to users, and ask them to form hypotheses about an object detection model. Our results provide positive evidence that these tools provide some benefit over a naive baseline, and also shed light on challenges faced by users during the hypothesis formation step. We conclude by discussing design opportunities for ML and HCI researchers. Our findings point to the importance of centering users when creating and evaluating new tools for slice discovery.