Prescribing drug X is correct 99% of the time for symptom Y. For the 1% where it is fatal, statistical reasoning does not help you. Andreas Kollegger calls this reference class validation: before the agent acts, it has to know which group it is in.

Context graphs give agents the why. Not just knowledge and tools but the policies, rules, and prior decisions that explain why a certain action is right in a given context. The decision making framework in this talk has five stages: frame the problem with its causality and environment, pull in global rules and past precedent, run a risk value analysis, either act or escalate to someone with authority, and write the full reasoning chain back into the graph. That last step is the point. Every decision becomes precedent. Future agents inherit it.

Speaker info:
- https://x.com/akollegger
- https://www.linkedin.com/in/akollegger
- https://github.com/akollegger

AI Engineer

A conference talk from Neo4j engineers introducing 'context graphs' as a framework for making AI agents more decision-aware. Beyond storing knowledge, context graphs encode rules, policies, and reasoning memory so agents understand not just what to do but why. The talk presents a structured decision-making workflow covering problem framing, global context (prior decisions and business rules), risk-value analysis, authority-based action or escalation, and self-learning by recording reasoning traces back into the graph. The framework is designed for multi-agent systems using tools like LangGraph or ADK, with Neo4j as the underlying graph database and Cypher for querying.

Context Graphs for Explainable, Decision-Aware AI Agents — Andreas Kollegger & Zaid Zaim, Neo4j