Explaining Humour Style Classifications: An XAI Approach to Understanding Computational Humour Analysis

Mental health applications require transparency: The paper emphasizes that "clarity and transparency are vital in the field of mental health," highlighting why black-box AI models are problematic when used to assess humour styles that directly impact psychological well-being.

Current AI systems struggle to distinguish between similar humour types: The research identifies "particular emphasis on the challenges in distinguishing affiliative humour from other styles," suggesting that automated systems may misclassify humour in ways that could lead to incorrect mental health assessments.

Multiple factors create classification errors: The study found that key sources of model mistakes include "emotional ambiguity, context misinterpretation, and target identification," demonstrating that humour analysis is complex and requires understanding nuanced linguistic and emotional cues that AI systems may miss.

Explainable AI can reveal how mental health tools make decisions: By applying XAI techniques to the humour classification model, the researchers achieved "interpretable insights into the complex interplay of features that define different humour styles," showing that we can make AI systems more trustworthy by understanding their reasoning.

Feature analysis matters for responsible deployment: The framework's detailed examination of "feature importance, error patterns, and misclassification cases" demonstrates that before deploying AI in mental health contexts, organizations should rigorously audit what linguistic and emotional patterns their systems rely on.