ACME: an evidence-based behavioural, neurofunctional and bioethical assessment of the impact of present and future generative AIs on cognitive development in ecological settings
ProgettoBy 2026, Generative AI (GenAI) has transitioned from an experimental tool to a fundamental infrastructure in education, offering support to learners across
all levels. However, empirical evidence regarding the impact of AI chatbot (AIC) tutoring on the cognitive development of children and pre-adolescents
remains sparse and anecdotal. A significant concern is that premature or improper interaction with GenAI may lead to cognitive offloading, potentially
resulting in AIC Induced Cognitive Atrophy, where foundational cognitive skills are weakened through over-reliance on external guidance. Furthermore,
existing systems often suffer from sycophancy and inherent biases, which can widen learning inequalities.
ACME addresses these knowledge gaps by aiming to develop a neuro-cognitively grounded and empirically validated roadmap for the design of
responsible educational technology and curricula, specifically tailored for children aged 8-14, including vulnerable groups and individuals with
disabilities.
ACME employs an innovative, highly interdisciplinary methodology that bridges the gap between ecological classroom use and biological mechanisms.
The project’s approach includes:
Human cohort studies: conducting behavioral trials in classroom and home settings with 180+ Italian students to collect digitised evidence of task execution.
This includes neuropsychological and neurofunctional assessments (using EEG, fMRI, and fNIRS) to identify functional brain signatures associated with AI-
assisted learning.
Preclinical research: utilising mouse models to experimentally manipulate external guidance versus autonomous exploration. This provides a mechanistic
framework to interpret the neurophysiological changes observed in human learners.
Domain-controlled chatbot prototypes: implementing four distinct AI configurations — Direct Solver, Socratic/Metacognitive guidance, Adaptive Practice, and
Epistemic Verification — to assess their relative effects on engagement, retention, and cognitive offloading.
Stakeholder analysis: investigating factors affecting the adoption of GenAI, such as technology anxiety and perceived literacy among 100+ teachers and
educators.
ACME will deliver 3 output assets to foster a safe and effective digital learning environment:
Scientific assets: open-access datasets and validated neurocognitive markers of developmental trajectories related to digital use.
Educational assets: a Moodle-based infrastructure for age-appropriate digital literacy and adaptive assessment protocols for technology-specific tutoring.
Governance assets: legally and ethically compliant technological guidelines and policy briefs for developers, educators, and parents.
The project will establish the scientific foundations for trustworthy AI in education, directly supporting the objectives of the EU AI Act and the Digital
Education Action Plan by identifying when AI supports learning and when it risks cognitive atrophy.
all levels. However, empirical evidence regarding the impact of AI chatbot (AIC) tutoring on the cognitive development of children and pre-adolescents
remains sparse and anecdotal. A significant concern is that premature or improper interaction with GenAI may lead to cognitive offloading, potentially
resulting in AIC Induced Cognitive Atrophy, where foundational cognitive skills are weakened through over-reliance on external guidance. Furthermore,
existing systems often suffer from sycophancy and inherent biases, which can widen learning inequalities.
ACME addresses these knowledge gaps by aiming to develop a neuro-cognitively grounded and empirically validated roadmap for the design of
responsible educational technology and curricula, specifically tailored for children aged 8-14, including vulnerable groups and individuals with
disabilities.
ACME employs an innovative, highly interdisciplinary methodology that bridges the gap between ecological classroom use and biological mechanisms.
The project’s approach includes:
Human cohort studies: conducting behavioral trials in classroom and home settings with 180+ Italian students to collect digitised evidence of task execution.
This includes neuropsychological and neurofunctional assessments (using EEG, fMRI, and fNIRS) to identify functional brain signatures associated with AI-
assisted learning.
Preclinical research: utilising mouse models to experimentally manipulate external guidance versus autonomous exploration. This provides a mechanistic
framework to interpret the neurophysiological changes observed in human learners.
Domain-controlled chatbot prototypes: implementing four distinct AI configurations — Direct Solver, Socratic/Metacognitive guidance, Adaptive Practice, and
Epistemic Verification — to assess their relative effects on engagement, retention, and cognitive offloading.
Stakeholder analysis: investigating factors affecting the adoption of GenAI, such as technology anxiety and perceived literacy among 100+ teachers and
educators.
ACME will deliver 3 output assets to foster a safe and effective digital learning environment:
Scientific assets: open-access datasets and validated neurocognitive markers of developmental trajectories related to digital use.
Educational assets: a Moodle-based infrastructure for age-appropriate digital literacy and adaptive assessment protocols for technology-specific tutoring.
Governance assets: legally and ethically compliant technological guidelines and policy briefs for developers, educators, and parents.
The project will establish the scientific foundations for trustworthy AI in education, directly supporting the objectives of the EU AI Act and the Digital
Education Action Plan by identifying when AI supports learning and when it risks cognitive atrophy.