Data Skeptic Healthy Friction in Job Recommender Systems
11 snips
Feb 2, 2026 Roan Schellingerhout, a fourth-year PhD student at Maastricht University studying explainable multi-stakeholder recommender systems. He talks about explainable job matching, comparing textual, bar chart, and graph explanations. The “healthy friction” study testing real versus random explanations and how people use explanations as info rather than decision rules. Also building knowledge graphs, LLMs for friendly explanations, and plans for fairness and real-world tests.
AI Snips
Chapters
Transcript
Episode notes
Text Beats Technical Visuals For Lay Users
- Lay users prefer simple textual explanations over technical visualizations like SHAP values or bar charts.
- Text gives a readable, recruiter-like rationale that non-technical users find easiest to understand.
Generate Text From Graphs With LLMs
- Use knowledge graphs as the internal explanation representation and convert them to natural text via LLMs for user-facing explanations.
- Feed structured graph JSON to an LLM and prompt it to generate easy-to-understand recruiter-style text.
Bar Charts Often Confuse Job Seekers
- Bar charts confused many lay users and mainly served as supportive summaries rather than primary informative elements.
- Users often skipped charts and returned to textual or graph explanations for clarity.