Roberto Araya (Universidad de Chile): LLMs with Chain-of-Thoughts: Teaching the Essential Skill for Deeper Human-AI Collaboration

Abstract:

Prof. Dr. Roberto Araya (Universidad de Chile; Mathematics Education)

AI is becoming increasingly powerful, making it easier for those who use it best to achieve enormous advantages. This requires knowing how to interact successfully with artificial agents that, in many domains, are already more intelligent than ourselves. This, in itself, is an unprecedented challenge. However, the educational system seems immune to this challenge. It assumes that the mechanisms we interact with are non-intelligent. They cannot read our minds and exploit our limited rationality and cognitive and socioemotional biases. It assumes that the only intelligent agents we interact with are people. In this traditional worldview, all intelligent agents have a psychology similar to ours. These assumptions are no longer valid. Present-day LLM-based artificial agents have potent language and seductive manipulation capabilities, although they are still unreliable in planning and deductive reasoning. This is an unforeseen psychology because, traditionally, these skills go together. Moreover, different LLMs produce radically different solutions and justifications. Therefore, that is where education should focus: Chain-of-Thought thinking. We propose teaching thinking strategies that prepare for effective interactions with artificial agents. This is a demanding task. It means teaching utterly new skills and a higher complexity of classroom interactions. Thus, teachers' cognitive load explodes. We propose an app to support teachers during class to ease this task.

Alle Interessierten sind hiermit herzlich eingeladen!

Mit freundlichen Grüßen

Prof. Dr. Rita Borromeo Ferri