Dataset resulting from the user study on comprehensibility of explainable AI algorithms

Different expertise levels understand AI explanations very differently: The study found that domain experts (mycologists), data science students, and humanities students had distinctly different abilities to comprehend the same AI explanations, highlighting that "one-size-fits-all" explanations may not work across diverse user groups.

Qualitative research is essential for evaluating explainability: Rather than relying solely on technical metrics, this dataset demonstrates the value of in-depth interviews and thematic analysis to understand how real people actually interpret AI decision explanations, addressing an "emerging topic" in responsible AI evaluation.

Users can identify specific improvements to explanations: Study participants provided concrete recommendations for how AI explanations could be made clearer, suggesting that involving actual users in the design process—rather than only having AI developers create explanations—yields actionable insights for better comprehensibility.

Explainability requires multidisciplinary evaluation: The inclusion of experts from mycology, data science, and social sciences/humanities reflects that assessing whether AI systems are truly understandable cannot be done by technologists alone and benefits from diverse perspectives on what makes explanations meaningful.

Reproducible datasets advance responsible AI practice: By making interview transcripts, visualizations, and analysis publicly available, this work enables other researchers to study explainability rigorously and consistently, supporting the development of more trustworthy AI systems.