Designing LLM Agents to Ask Questions When Facing Uncertainty
aestheticcode · reddit · 2026-08-17
A developer explores how to design an LLM agent that decides whether to ask a follow-up question when facing incomplete information, rather than making a blind guess. The use case is pediatric symptom assessment where parents might not provide full details initially. The agent should identify missing info (e.g., breathing difficulty) and ask. The author is considering Bayesian inference and HMM for uncertainty handling and question selection. The post discusses whether the LLM or a separate decision layer should handle this logic, and asks for advice on designing the loop, preventing endless questions, acting under uncertainty, and identifying potential failure modes.
More from coding & agent
- Using Codex for remote cluster control from phone — iScienceLuvr · 2026-08-17
- Agent Decision Logging Framework for Better Observability — blaizedsouza · 2026-08-17
- Dokie AI MCP Integration Automates Brand-Consistent Slide Generation — Aiden_Tech_Ai · 2026-08-17
- Multica open-sources a harness for your harnesses: assign issues to 20 coding agents — jiayuan_jy · 2026-08-17
- Reverse-engineered recipe for AI 3D games: threejs scaffold plus AI-generated .glb models — filiksyos · 2026-08-17
- One command setup: Claude Cowork configures itself via /setup-cowork — rubenhassid · 2026-08-17