Behavioural science needs a simple, outcome-focused benchmark that can be used to measure the impact of interventions to debias and improve decision-making. This paper proposes one: calibration. The goal of calibration isn’t to eliminate cognitive bias, but to enable good decisions despite it. The paper provides a framework for understanding and measuring calibration, tips and...

This report describes an exercise to investigate the robustness and reliability of using generative AI to help produce rapid evidence reviews. The AI-assisted output was completed in 23% less time than the human-only output but was judged to be less fluent and required more revisions.

Governments are increasingly using behavioural insights to design, enhance and reassess their policies and services. This report explores how this happens – and how any bias can be addressed or mitigated. To do this, this report focuses on three core activities of policymaking: noticing, deliberating and executing.