Design AI learning systems for inspectable, contestable use: provenance, represented uncertainty, user control and epistemic plurality¶

Strategy · All strategies
Evidence · no claims cited

Description¶

The review's design implications prescribe that systems support "claim-level provenance where feasible, distinguish retrieved evidence from model-generated synthesis, and make source limitations visible". Uncertainty should be represented through alternatives, ranges and explicit unknowns rather than unsupported confidence scores; interfaces should preserve user control to revise prompts, compare outputs and reject recommendations; and systems should surface disagreement and diverse sources. These requirements demand sufficient traceability for educational judgement, not complete model transparency.

Design Implications¶

Context¶

Requirements¶

• Sufficient traceability for educational judgement, rather than complete model transparency

Constraints¶

• Claim-level provenance is recommended where feasible, recognising full transparency is not demanded

Target Learners¶

• Learners using AI-integrated learning management systems, writing environments and feedback workflows

Target Learning Goals¶

• Maintaining verification practice and ability to audit provenance and limitations of AI outputs

Affordances¶

• Six Diagnostic Criteria Epistemic Dependence

Related Strategies¶

• Teach AI-mediated judgement and make process, judgement and transfer visible in assessment
• Embed AI literacy as a credit-bearing module and design adaptive systems with transparent, user-controllable recommendation interfaces

Examples¶

-

Key Sources¶

• Du Y, Yuan Y (2026) Epistemic dependence in AI-mediated learning. AI & SOCIETY. https://doi.org/10.1007/s00146-026-03294-1