Review Anchoring

Anchoring is a judgment bias in which an initial value, claim, or frame exerts too much influence on a later estimate or decision, even when the starting point is incomplete or irrelevant.

The first answer creates a mental starting point. People may adjust away from it, but often not far enough. In an AI conversation, an early estimate or framing can pull both the user and the model toward later answers that stay near it.

An arbitrary starting point can shape budgets, timelines, risk estimates, and interpretations. Recognizing the anchor helps a team generate an independent estimate, reconsider the frame, or compare several starting points before committing.

An AI assistant suggests that a migration will take two weeks. Even after the team discovers several unexamined systems, its revised estimates remain close to two weeks because the first number became the reference point.

Anchoring does not mean people never change their minds. They may make several adjustments and still remain too close to the initial value. The effect can also occur when people know that the anchor is arbitrary.

The first number pulls the next one.

* [[confirmation-bias|Confirmation Bias]] * [[evidence-selection|Evidence Selection]] * [[calibration|Calibration]] * [[epistemic-humility|Epistemic Humility]]

* [Amos Tversky and Daniel Kahneman, “Judgment under Uncertainty: Heuristics and Biases”](https://pubmed.ncbi.nlm.nih.gov/17835457/) — the classic account of anchoring and insufficient adjustment from an initial value.

* [RS AI Learning Lab, “Escaping the Conversation Gravity Well”](../../../../modules/escaping-the-conversation-gravity-well/module.md) * [RS AI Learning Lab, “When AI Starts Arguing Instead of Thinking”](../../../../modules/when-ai-starts-arguing-instead-of-thinking/module.md)