cognitive-bias-checklist-skill

To check for cognitive bias in AI reasoning outputs, apply a structured bias checklist after slow-mode analysis to surface risks like anchoring and confirmation bias, then guide corrective actions across engineering and product decisions.

Slow-mode reasoning in AI safety is a deliberate analysis phase where outputs are generated before undergoing explicit cognitive bias checks to ensure risk-aware decision making and mitigate high-consequence errors.

To mitigate the planning fallacy and overconfidence bias in software engineering estimates, enforce explicit checks using a bias checklist template after slow-mode analysis to record final confidence levels and remaining uncertainties.

Yes, you can integrate a bias checklist into operations and product analysis workflows by applying it after slow-mode reasoning, using the provided workflow definitions and pairing guidance to enforce risk-aware decision making.

Yes, this bias check approach covers the availability heuristic and narrative fallacy, alongside anchoring, confirmation, scope insensitivity, and substitution bias, by enforcing structured checks before finalizing slow-mode outputs.

A limitation of using a cognitive bias checklist for AI safety is that it must be applied after slow-mode reasoning completes, meaning it surfaces biases post-analysis and records remaining uncertainties rather than preventing them during initial generation.