Every Assessment Needs an AI Use Contract

A university publishes a responsible-artificial-intelligence policy. It tells learners to uphold integrity, disclose substantial use and remain accountable for submitted work. A lecturer then releases an assessment that says nothing about AI.

One student uses a chatbot to clarify a concept. Another asks it to restructure a draft. A third generates most of the answer and edits the language. Some disclose every interaction because they fear being accused later. Others disclose nothing because they fear that any admission will be treated as misconduct.

The policy exists, but the educational decision has been transferred to students at the moment of greatest uncertainty.

Responsible assessment needs more than an institution-wide statement. Every consequential task should contain a short AI use contract: a clear, proportionate account of the competence being assessed, the assistance permitted, the work that must remain the learner’s, the evidence required and the route to human review.

On 30 July 2026, the University of the Free State’s AI Community of Practice held an online session on using the AI Assessment Scale to support science-teacher development. The official event page describes a discussion of the scientific and ethical questions raised by AI and of assessment design in African teacher education.

This is an event record, not an evaluation. It does not show that participants changed their assessment practice, that the scale improved learning or that one…