Research
Sol
Sol is a conversational system that helps autistic adolescents describe how they learn, what they are good at, and what support actually helps — in their own words.
It runs on WhatsApp or Telegram, in ordinary conversation. There are no forms and no questionnaires. Across a few short chats, Sol talks with the young person about school, interests, what a good day looks like and what a hard one does. Their caregiver has a separate conversation of their own.
From those conversations Sol drafts an integrated learner profile: a portrait of how this young person learns, built from what they said, with the source of every statement recorded.
The study
What is being studied
Sol is the subject of doctoral research by Isaac S. Lee at the University of Cambridge, examining whether a generative-AI system can help adolescents produce learner-led profiles that are accurate, practically useful, and acceptable to the families and professionals who would rely on them.
The study asks three things:
- Can autistic adolescents create their own integrated profiles?
- Can a language model do this in a technically viable and secure way?
- Do learners, caregivers and professionals find the result trustworthy and useful?
The study is conducted by Isaac S. Lee as doctoral research at the University of Cambridge, under its own ethical approval. Edra, Inc. is a separate company and the University is not a partner in it.
What came before
What we heard first
Before building anything, we surveyed the three groups who would have to live with the result: adolescents, caregivers, and the educators and clinicians who work with them. Four findings shaped everything since.
Learner voice is thin, and the people closest to the work know it.
The most experienced special educators and community practitioners rated meaningful learner involvement lowest of any group. Not scepticism about the principle — a report from people who have watched it fail in practice.
Unfamiliarity with AI is not the obstacle.
Every group reported high familiarity already. Whatever is holding this back, it is not novelty.
Governance is the precondition, and it tightens as learners get older.
Control over who sees information about their child was the single strongest preference caregivers expressed — and it grew stronger, not weaker, for parents of sixteen and seventeen-year-olds, even as comfort with AI fell. Older adolescents’ families need the most explicit assurances, not the least.
The gap is widest exactly where interest is highest.
The groups reporting the least satisfaction with current support reported no less willingness to try something new. The equity case for this work is not an assumption; it is what the responses said.
Taking part
The study is currently recruiting
Participation involves a few short conversations on an app you already use, with nothing to prepare and nothing to install. You set the pace, and you can stop at any point without giving a reason.
For questions about the study, contact the researcher at isl29@cam.ac.uk.
Take part →How it works →For learners →For caregivers →Ethics and data →