CogSAFE scoring rubric v1.0: a six-domain instrument for assessing cognitive safety in digital products

Abstract

CogSAFE is a governance instrument for assessing how digital products, including AI systems, affect human cognitive function. This document is the full scoring rubric: six domains - attentional integrity, cognitive load, autonomy and agency, temporal well-being, memory and learning, and social cognition - each scored 0 to 3 against fixed, published criteria. Every criterion describes an observable property of how a product is built: defaults, interruption architecture, friction symmetry, offloading sequence, record ownership. The instrument assesses the default experience rather than what a determined user can configure. The output is a six-value profile with evidence requirements per domain. There is no total score, because weighting the domains against one another would require empirical justification that does not currently exist. Known limitations are published alongside the rubric: inter-rater reliability has not been tested, level thresholds are reasoned rather than empirically derived, and it supports structured self-assessment rather than independent audit. A four-step validation roadmap is included. Published in response to a documented gap. Policy mapping across 74 documents from 27 countries found addictive application design and neurological development among the least-addressed risks in existing regulation (Castro Lopes, Holly & Kickbusch, Global Policy 2025). Existing instruments regulate who may access a product; none assess how it is built.