Explaining AI Without Code: A User Study on Explainable AI

2026-02-13 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Advanced, long

Summary

A user study evaluated a human-centered Explainable AI (XAI) module integrated into DashAI, an open-source no-code Machine Learning (ML) platform. The module incorporates three XAI techniques: Partial Dependence Plots (PDP), Permutation Feature Importance (PFI), and KernelSHAP, providing global and local explanations for tabular classification models. The study, involving 20 participants (10 ML novices, 10 experts), assessed usability, satisfaction, and trust. Results showed high task success rates (>=80%) across all explainability tasks. Novices rated explanations as useful, accurate, and trustworthy on the Explanation Satisfaction Scale (ESS, Cronbach's α = 0.74), while experts were more critical regarding sufficiency and completeness. Explanations also improved perceived predictability and confidence on the Trust in Automation (TiA, α = 0.60), with novices exhibiting higher trust than experts. The findings highlight the challenge of designing XAI that is both accessible to novices and sufficiently detailed for experts in no-code environments.

Key takeaway

For AI Scientists developing or deploying ML models in no-code environments, you should prioritize integrating diverse XAI methods directly into the workflow. Focus on adaptive explanation strategies that can adjust depth and interpretability based on user expertise, ensuring novices gain trust and understanding while experts receive sufficient detail for diagnosis and validation. This approach will enhance model transparency and user confidence across your diverse user base.

Key insights

Integrating XAI into no-code ML platforms improves usability and trust, but must balance novice accessibility with expert diagnostic depth.

Principles

• XAI should be human-centered and integrated into ML workflows.
• Multiple complementary XAI methods enhance understanding.
• User expertise dictates explanation needs: simple for novices, detailed for experts.

Method

The study integrated PDP, PFI, and KernelSHAP into DashAI for tabular classification, evaluating usability, satisfaction (ESS), and trust (TiA) via a user study with 20 participants (novices and experts).

In practice

• Implement PDP for global feature impact.
• Use PFI to rank feature importance.
• Apply KernelSHAP for instance-level predictions.

Topics

• Explainable AI
• No-code ML Platforms
• Partial Dependence Plots
• Permutation Feature Importance
• KernelSHAP

Best for: AI Scientist, AI Researcher, Research Scientist, Machine Learning Engineer

Related on AIssential

• Why Should I Trust You? (and Can I?) — AI models are black boxes because their complex internal reasoning is difficult to interpret, not because their underlying mechanics are unknown.
• Using Learning Theories to Evolve Human-Centered XAI: Future Perspectives and Challenges — Infusing learning theories into XAI can shift focus from explanation complexity to human learning and agency.
• Improving AI models’ ability to explain their predictions — Extracting inherent model concepts yields more accurate and interpretable AI explanations than predefined human concepts.
• Interpretable and Explainable Surrogate Modeling for Simulations: A State-of-the-Art Survey and Perspectives on Explainable AI for Decision-Making — Integrating XAI with surrogate models enhances transparency and diagnostic capabilities for complex simulations, bridging efficiency and interpretability.
• An Explainable AI Assistant for Introductory Programming Education: Improving Feedback Reliability with Instructor-AI Collaboration — Instructor-AI collaboration with explainable models and constrained LLMs provides scalable, reliable, and pedagogically sound programming feedback.
• They Requested It. I Built It. Nobody Ever Used It. — Data models fail adoption due to explainability, slow delivery, and poor integration into user workflows.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.