Wizard of Errors: Introducing and Evaluating Machine Learning Errors in Wizard of Oz Studies

TL;DR

This paper introduces Wizard of Errors (WoE), a tool for simulating machine learning errors in Wizard of Oz studies to improve the design process of ML-enabled solutions.

Contribution

It presents a novel tool for simulating ML errors during user experience testing, addressing a gap in realistic error representation in WoZ studies.

Findings

Designers struggle to realistically simulate ML errors.

Descriptive error types are more effective than confusion matrices.

Manual error control has limitations in WoZ studies.

Abstract

When designing Machine Learning (ML) enabled solutions, designers often need to simulate ML behavior through the Wizard of Oz (WoZ) approach to test the user experience before the ML model is available. Although reproducing ML errors is essential for having a good representation, they are rarely considered. We introduce Wizard of Errors (WoE), a tool for conducting WoZ studies on ML-enabled solutions that allows simulating ML errors during user experience assessment. We explored how this system can be used to simulate the behavior of a computer vision model. We tested WoE with design students to determine the importance of considering ML errors in design, the relevance of using descriptive error types instead of confusion matrix, and the suitability of manual error control in WoZ studies. Our work identifies several challenges, which prevent realistic error representation by designers in such studies. We discuss the implications of these findings for design.