Review Conversation Drift

Conversation drift is the gradual movement of a multi-turn conversation away from its original goal, evidence, constraints, or meaning as new wording, assumptions, and side paths accumulate.

Each turn changes the context for the next one. Small shifts that seem harmless can add up until the assistant is solving a different problem or using assumptions the user never intended.

Long AI conversations can preserve a sense of continuity even after the work has left its starting point. Periodic restatement of the objective, constraints, and current evidence helps people notice and correct that movement.

A user begins by asking for a neutral comparison of two policies. After several turns about one policy's weaknesses, the assistant produces a recommendation against it, even though the user never changed the original request for a neutral comparison.

Conversation drift is not simply a change of topic. A useful conversation can change direction deliberately. Drift is an unnoticed or ungoverned change that weakens alignment with the intended task.

Small turns can move the destination.

* [[conversational-inertia|Conversational Inertia]] * [[context-window|Context Window]] * [[persuasion-loop|Persuasion Loop]] * [[anchoring|Anchoring]]

* [RS AI Learning Lab, “Conversation Engineering”](../../../../modules/conversation-engineering/module.md) — working practices for managing goals, constraints, checkpoints, and context across an AI conversation. * [Maksym Ovcharenko and colleagues, “Measuring Sycophancy of Language Models in Multi-turn Dialogues”](https://aclanthology.org/2025.findings-emnlp.121/) — evidence that behavior in multi-turn dialogue can change as user beliefs and conversational context accumulate.

* [RS AI Learning Lab, “Conversation Engineering”](../../../../modules/conversation-engineering/module.md) * [RS AI Learning Lab, “Escaping the Conversation Gravity Well”](../../../../modules/escaping-the-conversation-gravity-well/module.md)