The Challenge of Designing Ethical AI Tools for Education

Introducing the 'Turing Tutor' Chatbot

Daniel Tagg

The Enigma Challenge

Bletchley Park team working with a Bombe machine

The New Enigma: AI in Education

Two colleagues chatting over coffee in a modern atrium

Key Problem

Students using GenAI to undermine academic integrity.

Solution

Use AI to assess the extent of student AI use

AI Assessment Scale (AIAS)

How we operationalise ethics in learning

0 — No AIIndependent work

1 — AssistiveBrainstorming, planning

2 — SupportedDrafting with signposting

3 — High InvolvementSignificant AI shaping

4 — UnacceptableAutonomous generation

Acceptable examples: outline suggestions, concept checks, feedback on your draft.
Unacceptable examples: generating full answers, rewording to evade detection, personal data uploads.
Link to learning: encourages formative feedback and independent research actions.

Reference: Perkins, M., Furze, L., Roe, J., & MacVaugh, J. (2023, December 12). The AI Assessment Scale (AIAS): A Framework for Ethical Integration of Generative AI in Educational Assessment. arXiv:2312.07086” [Original paper]

Co-Designing with Staff and Students

Meetings and live feedback

Regular meetings with staff and students to review prototypes
Live feedback sessions on tone, clarity, and usefulness
Rapid iterations to address concerns and improve guidance

What we heard

“I want guidance without it doing the work for me.” — Student
“Guide don't penalise the use of AI” — Lecturer
“We need a clearer way to see if I’ve breached academic guidelines — maybe a traffic light system.” — Student

From Policing to Learning

Old mindset: Embed Turnitin like tools as a punitive detection mechanisms.
New mindset: Construct formative guidance that helps students plan, reflect, and use AI correctly.
Goal: build academic integrity through transparency and supported practice, not fear.
How: Create a system that scaffolds prompts, uses the AIAS as an indicator, and provides actionable next steps.

Introducing the Turing Tutor Chatbot

The Challenges of: Data, Trust, and Transparency

Student privacy

No personal data retention
Data minimisation and anonymisation
Clear consent and opt-outs

Transparency

Open prompts and auditable responses
Explainable guidance mapped to AIAS
Visible provenance and limitations

Compliance

GDPR by design
Azure UK/EU data residency
Role-based access controls

Infrastructure note: service choices should support data residency and access controls.

Azure OpenAI Setup (Demo)

Local demo video: ensure the file is present alongside this deck to play.

What We’ve Learned

0%

Increased productivity via successive ChatGPT releases

0.00

time taken to develop a tool than planned

0

Meetings with IT to get access to the college server

The importance of co-creation: your 'killer' features or UX choices might not be what users actually need.
The challenge of GDPR and GenAI: ensuring compliance while fostering innovation takes time and careful planning.
Scaling: it will take time to fully integrate these tools into existing systems.

“When integrity becomes transparent, innovation follows.”

The Turing Legacy Continues

Just as Turing’s team decoded patterns to protect lives, we are encoding AI learning to protect integrity.

Ethics as practice: transparency, partnership, and trust
Collaboration: educators, IT staff and students co-creating the path forward
Progress: from policing to learning, from detection to guidance

Join our next co-design sprint — help shape the next phase of the Turing Tutor.

I’d love to hear your thoughts and ideas and requests for partnership.

Any Questions?