Is AI Becoming A Kind Of Digital Folie À Deux?

Folie à deux is a psychological term describing a belief that becomes reinforced between two people because it is rarely challenged. The idea is usually discussed in a clinical context, but it raises an interesting question about how people interact with AI.

When we use AI, we often want it to help us think through an idea, solve a problem or make sense of something. The system is designed to be helpful, so it follows the direction of the conversation and builds on whatever we’ve already given it. Most of the time, that’s exactly what makes it useful.

The challenge is that AI often takes our starting point and runs with it. If we already have a particular view on something, the system is usually more likely to help us develop that view than question it. It helps us explore ideas, fill in gaps and think things through.

That’s not necessarily a flaw. In many situations, it’s exactly what we want. But it can create an interesting dynamic. We bring an idea, the AI helps refine it, and the idea becomes clearer, more detailed and often more convincing. What it doesn’t always become is more challenged.

This isn’t about AI creating beliefs on its own. It’s about how easily it can help strengthen beliefs that are already there. What makes the effect difficult to spot is that it rarely feels like being misled. It feels like being supported. The responses are logical, well-structured and often aligned with what we were already thinking.

The interaction feels productive because it is productive. That’s what makes the dynamic interesting.

For most day-to-day uses, this probably isn’t a problem. People use AI for research, planning, brainstorming and practical tasks all the time. But as AI becomes more involved in decision-making, problem-solving and personal reflection, the question starts to matter more.

The issue isn’t whether a particular response is right or wrong. It’s whether the conversation creates enough space for disagreement, alternative perspectives and challenge.

If AI is good at meeting us where we are, how often does it take us somewhere different?