Evidence
What we know, and how sure we are.
We present evidence in three labelled tiers, so you can see what is established, what is emerging, and what applies to AI tutoring in general rather than to Nested Learning.
Three tiers
Evidence, labelled by strength
Each piece of evidence sits in one of three tiers, so its strength is clear before you read it.
Tier 1
Foundations
Established
The research behind each teaching method: Bloom on tutoring, Rosenshine's principles, retrieval practice, Gibbs' reflective cycle, simulation and scaffolding.
Research behind the modesTier 2
Early impact evidence
Emerging
What learners report from our own pilots. Small-scale and self-reported: an early signal, not proof of outcomes.
See the early evidenceTier 3
Published research on AI tutoring
Peer reviewed
Peer-reviewed studies of AI and intelligent tutors in general, summarised with what each does not show.
Published research
Tier 2: early impact evidence
What learners told us
In a learner survey from an early pilot, learners agreed that the AI tutor helped them learn something new and build an existing skill, and reported no difficulties using it.
4 out of 5Agree
“The AI tool helped me gain knowledge or understanding I didn't have before.”
4 out of 5Agree
“The AI tool helped me develop an existing skill (for example exam technique, structuring an answer).”
4 out of 5Agree
“I am confident I can complete the coursework required in my course this week.”
NoNo difficulties reported
“Has there been a difficult part of using the platform?”
Result: After the pilot, the organisation doubled and extended its use of the tutor.
How to read this
- Source: Learner survey, early pilot. The organisation is not named.
- Scale: 1 (strongly disagree) to 5 (strongly agree).
- Size: small-scale. We report the quality of learners' responses, not a headline number.
- What it does not show: these are learners' own views. They are not a measured learning outcome, and one pilot cannot show the tutor will work the same way elsewhere.
What comes next
Building a fuller picture
We will run the same learner survey across pilots and pair it with the per-turn analytics, so future reports show both what learners say and what the analytics measure.
Questions
Questions about evidence
Is there evidence that Nested Learning's AI tutor works?
There is early, small-scale evidence. In a learner survey from an early pilot, learners agreed (4 out of 5) that the tutor helped them gain new understanding and develop an existing skill, and reported no difficulties using it. After the pilot, the organisation doubled and extended its use. This is what learners reported, not a measured learning outcome.
How does Nested Learning present its evidence?
In three tiers, each labelled: the established research behind each teaching method, early impact evidence from our own pilots, and published research on AI tutoring in general. We do not publish outcome figures for Nested Learning until we have them.
Will Nested Learning publish an impact report?
Yes, when there is enough data. We plan to pair the learner survey with the per-turn analytics across pilots, so the report shows both what learners say and what the analytics measure.