Principled AI in Education

The framework rests on three interconnected anchors that must be addressed before selecting tools:

Rejecting the binary promise-vs-peril discourse and the rush to immediate implementation, Finkelstein (2025) proposes a principled framework anchored in three questions: What are our educational goals? What do we know about human learning? How can technologies serve those goals and models?^Finkelstein Principled AI Education 2025

v2 update (Jul 2026). A new preamble contextualizes the paper one year after initial sharing. Finkelstein argues that the critical question is not whether machines will match humans at every task, but which skills, habits of mind, and practices must remain within the human purview to preserve our humanity and societies. Three objectives for education stand out in the modern era: discernment (framing questions, validating and contextualizing answers, acting appropriately on results), empathy (understanding and sharing others' perspectives, the basis of communication and collaboration), and sense of self (understanding one's role, purpose, and belonging within a course, field, or culture). The original arguments and principles of action, he notes, remain relevant one year later.

The Three Anchors

The framework rests on three interconnected anchors that must be addressed before selecting tools:

1. Goals — What should education accomplish? Meaningful learning, democratic participation, preparation for dynamic futures.

2. Models of human learning — How do people actually learn? Drawing on decades of learning sciences scholarship: active construction, social mediation, transfer, metacognition, motivation.

3. Use of technologies — How can AI serve the goals and models, rather than dictating them?

Roles Clarified

The framework defines distinct responsibilities across four domains:

Domain	Educators	Learners	Technologies
Shaping curricula	Define learning goals, sequence concepts, assess alignment	Express interests, co-design pathways	Recommend resources, flag gaps
Designing instruction	Craft activities, scaffold progressively, adjust in real time	Engage actively, seek help strategically	Generate variations, personalize pacing
Assessing learning	Design authentic tasks, interpret patterns, give feedback	Self-assess, reflect, revise	Score at scale, track patterns, suggest interventions
Cultivating community	Build norms, facilitate dialogue, mediate conflict	Collaborate, support peers, participate	Connect learners, moderate asynchronously

Core Mandate: Augment, Don't Displace

"AI must augment, not displace, human capacities."

This principle is the through-line of the framework. Technology use must be aligned with enduring educational values:

Advancing meaningful learning (not just efficient content delivery)

Supporting democratic societies (not just individual skill acquisition)

Preparing students for dynamic futures (not just today's job market)

Relationship to Existing Research

Finkelstein principle	Supporting wiki evidence	Tension
Goals before tools	Stanford Evidence Base AI K12 2026 (few causal studies because goals were ill-defined)	Industry pressure to deploy fast
Augment, not displace	AI Tutor Safety Harms (displacement harms catalogued)	LLM Fallacy Misattribution (users willingly displace themselves)
Meaningful learning	Self Regulated Learning (SRL as meaningful engagement)	Convenience tempts toward shortcut design
Democratic societies	AI Literacy (distributed AI literacy for citizenship)	Platform concentration concentrates power
Assess authentically	Authentic Assessment (six-dimensional framework)	Automated grading scales cheaply

Using the Framework

The framework can serve as a diagnostic for any AI-in-education initiative:

1. State the goal clearly — Not "use AI in my course" but "improve students' ability to evaluate evidence"

2. Identify the learning model — What research on learning supports this goal? (e.g., productive failure, deliberate practice, social learning)

3. Match technology to model — Which AI affordance serves this model? (e.g., Socratic dialogue for productive failure, peer matching for social learning)

4. Evaluate against displacement — Does the tool do something the learner should do themselves? If yes, redesign.

Connected Concepts

Self Regulated Learning

AI Education

Higher Ed

Scaffolding

Connected Articles

Agentic AI Education Scoping Review — Agentic AI in Education: A Scoping Review of Research Landscape, Capabilities, and the Frontier Agent Paradigm

Agentic Workflows Education — Agentic Workflows in Education

AI Ethics Education Public Discourse — A Longitudinal Analysis of Public Discourse on AI Ethics in Education Using Twitter Data

AI Higher Ed Bridge Gap — Higher Education Must Bridge the AI Gap

Stanford Evidence Base AI K12 2026 — AI in K-12 Evidence Base

Transfer Of Learning — AI Learning Transfer

AI Tutor Safety Harms — AI Tutor Safety and Pedagogical Harms

AIED Carbon Footprint Reporting — The Environmental Cost of LLMs in AIED: Reporting and Practices

Authentic Assessment — Authentic Assessment

Critical Thinking GenAI Scaffolding — Scaffolding Critical Thinking with Generative AI

Educational LLM Alignment — Educational LLM Alignment

Finkelstein Principled AI Education 2025 — Principled AI Education Framework

Institutional Change Framework AI — A Framework for Institutional Change in the Age of AI

Lata Ferpa Compliant Local LLM Autograder — LaTA: A Drop-in, FERPA-Compliant Local-LLM Autograder for Upper-Division STEM Coursework

LLM Fallacy Misattribution — The LLM Fallacy and Misattribution of Competence

Multi Agent Instructional Design — Multi-Agent Systems for Instructional Design

Multimodal Learning GenAI — Multimodal Learning with Generative AI

Teacher Control AI Generation Math Visuals — When Should Teachers Control AI Generation for Mathematics Visuals?

Universities AI Era Rethinking — The University AI Didn't Replace: Rethinking Universities in the AI Era

A4l Analytics Pipeline — Generalizing a Highly Configurable Analytics Pipeline to Replicate and Support Educational Research Across Multiple D...

Aaai2026 Prompting Literacy K12 — Learning to Use AI for Learning: Teaching Responsible Use of AI Chatbot to K-12 Students Through an AI Literacy Module

Academiclaw Student Agent Benchmark — AcademiClaw: When Students Set Challenges for AI Agents

Adaptive Pretesting Retention — Do Gains from Generative AI-Enabled Adaptive Pretesting Persist? Evidence from a Retention Study

Affective Text Wearable Student Health — A Formative Study of Brief Affective Text as a Complement to Wearable Sensing for Longitudinal Student Health Monitoring

Agency Gap AI Writing — The agency gap in AI-supported writing: how reactive and proactive agent designs shape multimodal reasoning

Citation

Finkelstein, N. (2025). A principled way to think about AI in education: guidance for educators and policy makers on action based on goals, models of human learning, and use of technologies. arXiv:2510.01467.