Explaining AI Without Code: A User Study on Explainable AI

No-code ML platforms can successfully integrate explainability tools: The study demonstrates that three complementary explanation techniques (Partial Dependence Plots, Permutation Feature Importance, and KernelSHAP) can be integrated into user-friendly interfaces, with novices achieving ≥80% task success rates, showing that technical expertise is not required to understand AI explanations.

Explanations increase user trust and confidence in AI systems: According to the Trust in Automation scale, explanations improved users' perceived predictability and confidence in the AI model's decisions, suggesting that transparency directly addresses concerns about automated decision-making in sensitive domains like healthcare and finance.

There's a critical gap between what novices and experts need from explanations: While novices found explanations useful, accurate, and trustworthy, experts were more critical about whether explanations were sufficiently complete and detailed. This reveals a central challenge: making explanations accessible to non-technical users without oversimplifying them for those with deeper technical knowledge.

Democratizing AI requires more than just removing code barriers: The study highlights that simply making ML platforms "no-code" is insufficient for true democratization—explainability features must be deliberately designed and integrated into these platforms, as they "rarely include explainability" currently, limiting users' ability to understand and trust automated decisions.