Actionable companion · Learning design & outcomes
Learning with agency: Active inference goes to school: the importance of active learning in the age of large language models
AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not claims or recommendations attributed to the source authors. No independent peer review of this guide is asserted.
Evidence and context
Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Age / stage not established. Verify the participant population and task before transferring the findings.
Original authors’ abstract, reproduced under the linked source license:
Human learning essentially involves embodied interactions with the material world. But our worlds now include increasing numbers of powerful and (apparently) disembodied generative artificial intelligence (AI). In what follows we ask how best to understand these new (somewhat ‘alien’, because of their disembodied nature) resources and how to incorporate them in our educational practices. We focus on methodologies that encourage exploration and embodied interactions with ‘prepared’ material environments, such as the carefully organized settings of Montessori education. Using the active inference framework, we approach our questions by thinking about human learning as epistemic foraging and prediction error minimization. We end by arguing that generative AI should figure naturally as new elements in prepared learning environments by facilitating sequences of precise prediction error enabling trajectories of self-correction. In these ways, we anticipate new synergies between (apparently) disembodied and (essentially) embodied forms of intelligence. This article is part of the theme issue ‘Minds in movement: embodied cognition in the age of artificial intelligence’.
What kind of evidence is this?
Supporting AI learning science · Design requires full-text appraisal
Stage: Age / stage not established. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.
Limitations and uncertainty · source passage
In the case of ChatGPT, the training of the model occurs offline, 1 on human-produced textual data taken from the internet. The model is provided with input–output pairs: texts (or other inputs; e.g. [62–64]) from which part of the information is missed and of which is asked to predict the missing part (output). When predictions are accurate, the model is ‘rewarded’ (its accuracy scored through an assigned number) and then fine-tuned with an increasing quantity of texts, missing larger portions. Finally, performances are evaluated, validated and tested on datasets to ensure the model generalizes accurately to unseen data. Through different degrees of supervision, the network is optimized for effectively predicting the most likely output, based on associations and contrasts between inputs and data in its training set. That is, the model can predict, with a high level of accuracy, the portion of the missed text, which corresponds to answering the question asked by a user. LLMs are powerful tools: users leverage the (approximately) 2.25 billion pages contained in the indexed web and, through ‘prompt engineering’, craft their prompts in targeted ways to exact useful responses. The capabilities of LLMs have already been applied to a wide range of cases, such as the ability to decode the meaning of perceived and imagined speech and silent video [65]. Results are impressive, but they also highlight inherent limitations, particularly the flexibility in transferring knowledge from one set of data to another, which also bring about states of ‘artificial hallucinations’ (i.e. generated responses that are either factually incorrect, non-sensical or not grounded in reality (e.g. [66–68]).
Design the system before the pilot
Use this study to inform a learning mechanism or implementation question. It is not direct evidence that autonomous agents improve learning. Keep the proposed agent behavior separate from what the source actually tested.
- Specify a measurable learning objective, a comparator and a task without tool assistance before adopting a system.
A 90-day district pilot
- Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
- Days 16–30 · Design. Treat the paper as background evidence until its population is verified in the full text. Do not assign an age-specific intervention or extend adult results to children. Name an accountable owner and set the baseline, success threshold and budget before implementation.
- Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
- Days 61–90 · Decide. Extract the participant ages or education stage, recruitment, task and comparator before selecting a local pilot. If these cannot be established, retain the paper for design discussion rather than age-band efficacy claims. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.
The board decision brief
Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.
Attribution and reuse
Laura Desirèe Di Paolo; Ben White; Avel Guénin-Carlut; Axel Constant; Andy Clark. Active inference goes to school: the importance of active learning in the age of large language models. Philosophical Transactions of the Royal Society B: Biological Sciences (2024-08-19). https://doi.org/10.1098/rstb.2023.0148
Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.