From Curiosity to Caution: How Expertise Shapes the Use of Interpretable Machine Learning
Interpretability tools are increasingly used to make ML models more transparent, yet their effectiveness has been limited in practice despite significant prior work on improving their design. Recent work has presented collaborative settings as a solution, noting the potential value of people with different roles and domain expertise coming together to better utilize interpretability outputs. We extend these collaborative settings to include proficiency differential (novices vs. experts) as another facet of effective team composition. To investigate the value of this proficiency differential, we conducted contextual inquiries and semi-structured interviews with novices and experts (N=23) who completed the same data science task with an interpretability tool. Our work contributes empirical evidence of how proficiency shapes interpretability use, with novices driven by curiosity and experts by efficiency and caution. We present a framework for understanding expert-novice differences, and identify design implications for XAI that scaffold both expert-like reasoning and novice-like exploration.