Formative friction: require goal statements, alternative comparison, and justification before high-stakes delegation, kept proportionate and accessible¶

Strategy · All strategies
Evidence · no claims cited

Description¶

This strategy distinguishes exclusionary friction from formative friction in AI-mediated workflows. "Formative friction asks users to state a goal before optimizing, compare alternatives before accepting a recommendation, confront a counterargument before submitting a claim, or explain a choice before delegating a high-stakes action." The article specifies that such friction should be proportionate and accessible: the burden should match the stakes, must not exclude disabled users, multilingual users, overburdened workers, or people with limited resources, and users should be able to challenge or adapt friction when it is misplaced. The aim is to keep necessary difficulty from becoming a polite name for exclusion while preserving occasions to exercise judgment.

Design Implications¶

Context¶

Requirements¶

• Friction must be proportionate to stakes and accessible, with users able to challenge or adapt it when misplaced

Constraints¶

• Empty friction blocks access without improving agency; exclusionary friction such as needless paperwork and opaque appeals burdens people with less time, money, or institutional power

Target Learners¶

• workers
• students
• public-service users interacting with AI-mediated workflows

Target Learning Goals¶

• exercising judgment before delegation, including goal formulation, comparison of alternatives, and justification of choices

Affordances¶

• Five Post Instrumental Capacities

Related Strategies¶

• Post-instrumental assessment: certify the learner's accountable relation to AI-mediated work via a design sequence of naming the capacity, classifying AI use, and requiring evidence at the point of judgment
• Four-question public template for high-stakes AI assessment tasks
• Develop AI literacy frameworks that build metacognitive awareness distinguishing system-assisted performance from independent competence

Examples¶

-

Key Sources¶

• Kai Yao. (2026). When AI Does the Work, What Is Learning For? Post-Instrumental Learning and the Risk of Capacity Dissolution. arXiv. https://arxiv.org/abs/2607.28041