RECAST: Enabling User Recourse and Interpretability of Toxicity Detection Models with Interactive Visualization

TL;DR

RECAST is an interactive visualization tool that helps users understand and modify toxic language predictions from NLP models, promoting transparency, user recourse, and improved moderation effectiveness.

Contribution

It introduces an open-source web tool that visualizes toxicity predictions, offers alternative phrasings, and evaluates user interactions to enhance model transparency and user agency.

Findings

RECAST effectively helps users reduce detected toxicity.

Users gain better understanding of toxicity criteria.

Models become less effective when users optimize language for them.

Abstract

With the widespread use of toxic language online, platforms are increasingly using automated systems that leverage advances in natural language processing to automatically flag and remove toxic comments. However, most automated systems -- when detecting and moderating toxic language -- do not provide feedback to their users, let alone provide an avenue of recourse for these users to make actionable changes. We present our work, RECAST, an interactive, open-sourced web tool for visualizing these models' toxic predictions, while providing alternative suggestions for flagged toxic language. Our work also provides users with a new path of recourse when using these automated moderation tools. RECAST highlights text responsible for classifying toxicity, and allows users to interactively substitute potentially toxic phrases with neutral alternatives. We examined the effect of RECAST via two large-scale user evaluations, and found that RECAST was highly effective at helping users reduce toxicity as detected through the model. Users also gained a stronger understanding of the underlying toxicity criterion used by black-box models, enabling transparency and recourse. In addition, we found that when users focus on optimizing language for these models instead of their own judgement (which is the implied incentive and goal of deploying automated models), these models cease to be effective classifiers of toxicity compared to human annotations. This opens a discussion for how toxicity detection models work and should work, and their effect on the future of online discourse.