Research & Studies — Socratic AI

🎓 2026

LLMs Can Fuel Extremist Attitudes Using Universal Moral Framings

Rosamunde C. Hendricks, Helena Gil-Buitrago, Clara Pretus

Abstract

This study evaluates whether short, LLM-generated arguments using universal moral framings impact extremist attitudes. In two studies with Democrats and Republicans in the United States (N=951), researchers find that universal moral concerns related to welfare, rights, and fairness predict perceptions of political stances as sacred values, and that LLM-generated arguments appealing to individual rights and fairness increase people’s willingness to fight and die and justify violence to defend political stances.

Full summary → Read paper ↗

🎓 2025

Rethinking Educational Assessment in the Age of Artificial Intelligence: Insights From Recent Training Workshops

Zuheir Khlaif

AI Summary

As generative AI makes traditional assessments increasingly vulnerable to automation, educators need new frameworks for evaluating student learning. This paper draws on workshops with 300+ educators across the Global South to propose practical strategies for AI-era assessment design.

Full summary → Read paper ↗

🎓 2025

Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task

Nataliya Kosmyna, Eugene Hauptmann, Ye Tong Yuan, Jessica Situ, Xian-Hao Liao, Ashly Vivian Beresnitzky, Iris Braunstein, Pattie Maes

Abstract

This study explores the neural and behavioral consequences of LLM-assisted essay writing using EEG to assess cognitive load during essay writing across four months, revealing that LLM users consistently underperformed at neural, linguistic, and behavioral levels.

Full summary → Read paper ↗

🎓 2025

The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers

Hao-Ping (Hank) Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, Nicholas Wilson

AI Summary

From Carnegie Mellon University and Microsoft Research, this is one of the first large-scale studies examining how generative AI affects critical thinking in real workplace settings. The central finding — that trusting AI more leads to thinking less — has profound implications for knowledge work and education alike.

Full summary → Read paper ↗

🎓 2024

Generative AI Can Harm Learning

Hamsa Bastani, Osbert Bastani, Alp Sungu, Haosen Ge, Özge Kabakcı, Rei Mariman

AI Summary

This is one of the first rigorous field experiments demonstrating that generative AI access can actively harm learning outcomes. The finding that students perform worse after losing AI access than students who never had it suggests a dependency effect with serious implications for education policy.

Full summary → Read paper ↗

🎓 2024

Are We Inventing Ourselves Out of Our Own Usefulness? Striking a Balance Between Creativity and AI

Noel Carroll

AI Summary

Published in AI & Society’s “Curmudgeon Corner,” this essay cuts through the hype cycle around generative AI to ask a fundamental question: when AI can produce text, images, and music in seconds, what happens to the human creative journey that has driven civilization?

Full summary → Read paper ↗

🎓 2024

AI Meets the Classroom: When Do Large Language Models Harm Learning?

Matthias Lehmann, Philipp B. Cornelius, Fabian J. Sting

AI Summary

While the debate around AI in education often focuses on whether AI helps or harms, this study reveals a more nuanced picture: how students use AI matters more than whether they use it. The distinction between substitution and complementary use has direct implications for how educators should structure AI-assisted learning.

Full summary → Read paper ↗

🎓 2022

Language Models Exhibit Emergent Abilities in Complex Reasoning Tasks

Jason Wei, Yi Tay, Rishi Bommasani, et al.

AI Summary

This paper helped change how people think about scaling in large language models. Instead of assuming that abilities improve smoothly as models get bigger, the authors show that some capabilities appear suddenly and unevenly at particular scales. That matters because it suggests that model behavior can shift faster than our intuition, evaluation habits, or governance frameworks expect.

Full summary → Read paper ↗