A useful AI system can make a task feel easier. The harder question is whether it leaves the person more capable when the system is gone.
The experiment
A recent mixed-methods study asked young adults to identify logical fallacies. One group worked with ChatGPT; the other worked manually. The AI group produced more correct responses and reported substantially less mental effort.
That looks like a clean productivity win. But the qualitative results revealed two very different ways of using the same tool. Some people delegated the reasoning almost completely. Others questioned the output, verified claims, and monitored their own understanding.
When we hand thinking entirely to AI, it replaces cognition. When we continue to question, verify, and reflect, AI augments it.
Delegation or augmentation
The important dividing line is not whether a person used AI. It is how they used it. Complete delegation can improve the immediate result while removing the practice that develops judgment. Strategic engagement can lower routine effort while keeping the human responsible for the deeper work.
This is the signal behind AHI: performance is not enough. An intelligent system should be evaluated by what it helps a person produce now and by what it helps that person become over time.
Read the study in Cognitive Processing