Bloomberg Businessweek The Intersection of Human Psychology with AI
6 snips
May 8, 2026 
Dr. Pattie Maes, MIT Media Lab professor leading the Fluid Interfaces group, studies AI, HCI and cognition. She discusses how large language models affect attention, memory and learning. She explores short-term boosts versus long-term skill erosion. She argues for designs that augment thinking, add useful friction, and benchmark AI’s human impact.
AI Snips
Chapters
Transcript
Episode notes
LLMs Improve Performance Now But Can Weaken Skills
- LLMs can boost immediate task performance but risk eroding the underlying cognitive skills if overused.
- Pattie Maes's Media Lab tests show short-term gains in tasks like writing while long-term reliance reduces users' problem-solving and critical-thinking abilities.
Add Friction So AI Augments Not Replaces
- Design AI to augment rather than replace human effort by adding friction so users keep practicing skills.
- Maes recommends systems that support skill maintenance instead of doing all the work automatically for users.
Keep Humans In The Loop To Catch Hallucinations
- Maintain human oversight because LLMs still hallucinate and lack deep world understanding.
- Maes emphasizes users must be able to detect incorrect outputs and continue solving problems independently when necessary.