From Observation to Intervention
What student–AI interaction patterns tell us, and what we can do about them.
Slides (PDF)A learning & teaching research project · Curtin University
What if one line could change how students talk to AI?
Students often use AI like a vending machine: one question in, one answer out, straight into the assignment. Keep Asking tests whether a single, well-timed conversational nudge shifts that pattern toward following up, challenging the answer, and extending the idea.
The study
A randomised controlled trial running across Semester 2, 2026 in four units spanning four disciplines: Business Information Systems, Supply Chain, Marketing, and People, Culture and Organisations. Students complete an authentic AI-assisted task in their usual lab; half see an occasional one-line nudge in the chat, half don't. Everything else is identical: the model, the interface, and the task.
We're asking three things:
- Does the nudge increase critical engagement (follow-up, challenge, extension) compared with a silent control?
- Does that engagement show up in the quality of what students produce?
- Is any benefit equitable? Does it reach students who arrive with fewer AI habits, or only those who already have them?
The trial is pre-registered, and approved by the Curtin University Human Research Ethics Committee (HREC 83897). It is funded by a 2026 Curtin Needs-Based Funding Learning & Teaching grant. Links to the registration and papers will be posted here once peer review concludes.
Talks
Keep Asking
A one-line nudge toward deeper student–AI conversations.
Coming sooniSOLT Presentation
Details and slides to follow.
Coming soonFBL Research Symposium
Details and slides to follow.
Coming soonTeam
Four investigators across four disciplines, so we can see whether the effect travels beyond any one field.
- Michael BorckProject lead · Business Information Systems
- Marcela MoraesCo-investigator · Marketing
- Torsten ReinersCo-investigator · Supply Chain
- Renee RalphCo-investigator · People, Culture and Organisations
Contact: michael.borck@curtin.edu.au
Tools
keep-asking-app
The open-source chat platform that delivers the study: randomised assignment, nudge injection, and a model instructed to answer plainly, so the only structured cue in the trial is ours.
GitHub ↗TaskForge
A task designer for building the authentic, discipline-specific AI-assisted tasks the study runs on: the kind a generic AI answer would get wrong.
taskforge.borck.education ↗