Why this belongs in Chitta

The question connects receiving an explanation, examining it and making something of our own. AI is one setting in which to investigate it—not the definition of awareness and not the centre of every Chitta inquiry.

What the first evidence opens up

The mathematics study by Bastani and colleagues found that assistance during practice and performance without that assistance could diverge. The teacher-informed GPT Tutor safeguards largely mitigated the negative exam effect observed with GPT Base. The authors did not find a positive exam effect for GPT Tutor relative to the control group. This distinction concerns the study’s unaided exam, not every kind of independent learning.

Separate the outcomes

• Quality and speed while assistance is available.
• Performance immediately after assistance is removed.
• Retention after a delay.
• Transfer to a different task.
• Ability to recognise and correct an error.

What remains unresolved here

Which findings extend beyond the original learners and tasks? How do prior knowledge, tool design and assessment timing affect the comparison? Our seed sources do not yet support a general prescription. A full dossier needs independent evidence and appraisal of the relevant methods.

How the dossier would develop

Each evidence card would show its question, setting, comparison, outcome, limitations and source. New evidence could revise the synthesis. Personal observations could raise questions, but would be labelled separately from controlled research.

A contrasting study to examine

Kestin and colleagues (2025) studied two introductory physics lessons at Harvard using a crossover design. Students experienced a structured AI tutor and in-class active learning in different weeks. Assignment respected existing peer groups; this was not simple individual randomization.

The tutor condition had higher post-lesson scores. However, the comparison also differed in location, pacing and interaction format. It does not isolate a chatbot alone or establish long-term retention.

Main-article results, methods and limitations inspected. Supplementary analyses and underlying data have not been independently audited.

Comparing this with the mathematics study requires attention to the learners, tutor design, comparison condition and assessment timing. Different results do not automatically mean the studies contradict each other.

Current status

Proposed evidence dossier. No Chitta participant experiment has been conducted, no conclusion is preselected, and no submission service is activated in this preview.