Story perspectives

AI's Human-Like Error Sparks Educational Insights

9/19/2025

30 5

Summary

Story summary

Researchers from the University of Cambridge and Hebrew University tested ChatGPT-4 with Plato's "doubling the square" puzzle.
ChatGPT initially used an algebraic approach instead of the known geometric solution, indicating improvisation.
The AI made a human-like error, stating geometric solutions for rectangles were impossible, showing reasoning rather than data recall.
The study introduced "Chat's zone of proximal development," highlighting the role of prompting in AI responses.
Findings suggest AI can enhance education by fostering critical evaluation and reasoning skills.

1 of 1

Story summary

Researchers from the University of Cambridge and Hebrew University tested ChatGPT-4 with Plato's "doubling the square" puzzle.
ChatGPT initially used an algebraic approach instead of the known geometric solution, indicating improvisation.
The AI made a human-like error, stating geometric solutions for rectangles were impossible, showing reasoning rather than data recall.
The study introduced "Chat's zone of proximal development," highlighting the role of prompting in AI responses.
Findings suggest AI can enhance education by fostering critical evaluation and reasoning skills.