What question did this study set out to answer?

This study aims to evaluate the effectiveness of explainability techniques for a misogyny detection AI model in code-mixed languages.

March 13, 2026Applied AI LettersOpen Access

User‐Based Evaluation of Explainability Techniques for Misogyny Detection in Code‐Mixed Hindi–English

View Full Paper

Ask AI

Bookmark

Authors

User evaluation reveals preferences for explainability techniques in misogyny detection systems, indicating importance in AI transparency.

This study aims to evaluate the effectiveness of explainability techniques for a misogyny detection AI model in code-mixed languages.
User-based evaluation of three explainability techniques: simulatability, what-if explainability, and qualitative feedback.
Participants justified their predictions and provided feedback on the explanations.
Analysis involved comparing user preferences for different explanation formats.
Integrated Gradients (IG) technique was found to be more consistent and plausible compared to others.
Participants preferred token attribution heatmaps and bar plots over natural language explanations.
Feedback indicated varying levels of satisfaction with the provided explanations.

Yadav et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

User evaluation reveals preferences for explainability techniques in misogyny detection systems, indicating importance in AI transparency.