Cognitive Grounding Before AI Scaling: Why AI-in-Education Systems Need Learning Theory First

Syaamantak Das
Proceedings of the Impactful and Responsible AI Systems for Education Workshop, PMLR 339:129-139, 2026.

Abstract

AI-powered educational tools proliferate globally, reaching hundreds of millions of learners through national education platforms, EdTech products, and AI tutoring systems. A critical gap persists between the pace of AI deployment and the depth of cognitive grounding that underlies these systems. This paper argues that responsible AI for education requires a \emph{cognitive grounding phase} before scaling: a systematic, empirically validated assessment of what existing educational systems actually demand cognitively, and whether AI tools are designed to enhance or merely replicate those demands. Drawing on a decade of empirical research classifying over 3,000 assessment items from national school board examinations using the revised Bloom’s Taxonomy, we demonstrate that the dominant examination systems in a major Global South context overwhelmingly test the lowest cognitive levels, precisely the levels at which large language models perform most capably. We introduce the concept of digital rote to describe the phenomenon in which AI systems technologically reproduce pedagogically shallow learning at scale, and propose a three-stage Cognitive Grounding Framework — Cognitive Audit, Cognitive Specification, and Cognitive Validation — developed in the context of LLM-based assessment generation and extensible in principle to other AI-in-education applications. We demonstrate the framework’s practical viability through a patented AI-based assessment system whose design was directly informed by cognitive-level analysis. Finally, we discuss policy implications for national-scale AI deployment in education and outline open questions for the responsible AI-in-education community.

Cite this Paper

BibTeX
@InProceedings{pmlr-v339-das26a, title = {Cognitive Grounding Before AI Scaling: Why AI-in-Education Systems Need Learning Theory First}, author = {Das, Syaamantak}, booktitle = {Proceedings of the Impactful and Responsible AI Systems for Education Workshop}, pages = {129--139}, year = {2026}, editor = {Basu Mallick, Debshila and Woodhead, Simon and Wang, Zichao and Ananda, Muktha and Burstein, Jill and Murphy, April}, volume = {339}, series = {Proceedings of Machine Learning Research}, month = {28 Jun}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v339/main/assets/das26a/das26a.pdf}, url = {https://proceedings.mlr.press/v339/das26a.html}, abstract = {AI-powered educational tools proliferate globally, reaching hundreds of millions of learners through national education platforms, EdTech products, and AI tutoring systems. A critical gap persists between the pace of AI deployment and the depth of cognitive grounding that underlies these systems. This paper argues that responsible AI for education requires a \emph{cognitive grounding phase} before scaling: a systematic, empirically validated assessment of what existing educational systems actually demand cognitively, and whether AI tools are designed to enhance or merely replicate those demands. Drawing on a decade of empirical research classifying over 3,000 assessment items from national school board examinations using the revised Bloom’s Taxonomy, we demonstrate that the dominant examination systems in a major Global South context overwhelmingly test the lowest cognitive levels, precisely the levels at which large language models perform most capably. We introduce the concept of digital rote to describe the phenomenon in which AI systems technologically reproduce pedagogically shallow learning at scale, and propose a three-stage Cognitive Grounding Framework — Cognitive Audit, Cognitive Specification, and Cognitive Validation — developed in the context of LLM-based assessment generation and extensible in principle to other AI-in-education applications. We demonstrate the framework’s practical viability through a patented AI-based assessment system whose design was directly informed by cognitive-level analysis. Finally, we discuss policy implications for national-scale AI deployment in education and outline open questions for the responsible AI-in-education community.} }
Endnote
%0 Conference Paper %T Cognitive Grounding Before AI Scaling: Why AI-in-Education Systems Need Learning Theory First %A Syaamantak Das %B Proceedings of the Impactful and Responsible AI Systems for Education Workshop %C Proceedings of Machine Learning Research %D 2026 %E Debshila Basu Mallick %E Simon Woodhead %E Zichao Wang %E Muktha Ananda %E Jill Burstein %E April Murphy %F pmlr-v339-das26a %I PMLR %P 129--139 %U https://proceedings.mlr.press/v339/das26a.html %V 339 %X AI-powered educational tools proliferate globally, reaching hundreds of millions of learners through national education platforms, EdTech products, and AI tutoring systems. A critical gap persists between the pace of AI deployment and the depth of cognitive grounding that underlies these systems. This paper argues that responsible AI for education requires a \emph{cognitive grounding phase} before scaling: a systematic, empirically validated assessment of what existing educational systems actually demand cognitively, and whether AI tools are designed to enhance or merely replicate those demands. Drawing on a decade of empirical research classifying over 3,000 assessment items from national school board examinations using the revised Bloom’s Taxonomy, we demonstrate that the dominant examination systems in a major Global South context overwhelmingly test the lowest cognitive levels, precisely the levels at which large language models perform most capably. We introduce the concept of digital rote to describe the phenomenon in which AI systems technologically reproduce pedagogically shallow learning at scale, and propose a three-stage Cognitive Grounding Framework — Cognitive Audit, Cognitive Specification, and Cognitive Validation — developed in the context of LLM-based assessment generation and extensible in principle to other AI-in-education applications. We demonstrate the framework’s practical viability through a patented AI-based assessment system whose design was directly informed by cognitive-level analysis. Finally, we discuss policy implications for national-scale AI deployment in education and outline open questions for the responsible AI-in-education community.
APA
Das, S.. (2026). Cognitive Grounding Before AI Scaling: Why AI-in-Education Systems Need Learning Theory First. Proceedings of the Impactful and Responsible AI Systems for Education Workshop, in Proceedings of Machine Learning Research 339:129-139 Available from https://proceedings.mlr.press/v339/das26a.html.

Related Material