SALTISE 2026 · June 2

Eye to Eye
with AI

Adopting a five-level AI assessment scale at college level:
steps, obstacles, and expected gains

15 min
Talk
Context

The challenge

Clarity is key when communicating to students what they can and cannot do with Generative AI use during assessments. Establishing coherent descriptions with easy-to-recognise visuals may reduce stress and confusion among students; making and communicating such choices can itself be stressful and confusing for teachers.

Following in the footsteps of other institutions, our college and its international network are adopting a five-level scale for indicating, within the instructions of each graded assessment, what type of GenAI use is permitted. Levels range from forbidding all AI use to free use — topping off with a level that incorporates AI as an ingredient of the work itself.
🏫

Collège LaSalle Montréal · LCI Education

Elisa Schaeffer — Dean, Technology & Design. The scale was developed by LCI Melbourne, translated to French by Collège LaSalle Montréal (motivated by the Université de Sherbrooke task grids), and to Spanish by LCI Barcelona. It draws on Leon Furze's AI Assessment Scale (2024).

Five levels

The AI Assessment Scale

Each graded assessment carries one of these five levels — communicated to students in the course outline and within the assessment instructions.

Click a level to expand its definition.

Three non-negotiables in all AI-assisted work

  • attribution: cite the tool & prompt
  • validation: verify and correct outputs
  • assessing process over product: value decisions and approach, not just the final result

Collège LaSalle Montréal · 2025–2026

Prep work by teachers

During a pedagogical day, teams of full-time teachers (grouped by program) analysed their assessments through the lens of AI, encountering four recurring issues and envisioning four recurring solutions.

Causes for concern

Proposed solutions

Emergent convergence the same three responses arose repeatedly: make the process visible, add a live oral component, redesign the task.
LCI Education · Spring 2026

Level Distribution

Initial planning data from Collège LaSalle Montréal

The planning done by teaching teams does not commit them to implementing it as such.

Discussion · 3 minutes

Over to you

Adoption

  • Do you have or want to have a scale like this in your institution?
  • Who needed/needs to be on board first?

Friction

  • Which of the issues resonates most with your experience?
  • Which solutions feel actionable for your context?

Students

  • How do you currently communicate AI expectations to students?
  • What would clarity look like for your assessments?

Explore the workshop

The interactive AQPC 2026 workshop includes discipline-specific sorting activities, scenario exercises, and the full LCI network survey data — in French, English, and Spanish.

Open full workshop →
QR code — full workshop