Is AI making you a better or worse thinker?

Bangerl et al. (2025, CHI) studied creative collaboration with AI and found something uncomfortable: users who relied on AI generated fewer of their own ideas, and the effect carried over to unaided work afterward. The cognitive cost was not temporary. Noy & Zhang (2023, Science) showed a related pattern: AI compresses quality toward a mean, improving weaker work but flattening stronger work.

This does not mean AI makes everyone a worse thinker. Dell'Acqua et al. (2023) found that consultants who understood AI's limits produced 40% higher quality work on suitable tasks. Those who misjudged the boundary performed 23 percentage points worse. Gajos & Mamykina (2022) showed that AI assistance can improve short-term performance even when it impairs long-term learning. The question is not whether to use AI but how.

The distinction between augmentation and offloading is not fixed. It depends on how you engage. This tool helps you see where the balance sits today.