Groupthink and Group Behaviour in the AI Era: Are We Still Thinking for Ourselves?

By Dr Faheem Naqvi  
Consultant Psychiatrist and Director   
18 September 2026

Dr Faheem Naqvi on Groupthink, group behaviour and artifical intelligence

AI helps us do more, could it leave us thinking less independently?

Artificial intelligence has slipped into our workplaces, our mobile phones and our everyday conversations and has become part of everyday life and is making possible, what once seemed like science fiction. I find that exciting, but also equally unsettling, a feeling I tend to supress.

Group behaviour is about how we act together and influence one another. Groupthink is more specific: it happens when the desire for agreement gets in the way of questioning assumptions and considering alternatives. Think of a meeting where everyone nods along, although several people privately have doubts. The apparent consensus may say more about what went unsaid than what everyone actually believed.

Now imagine everyone arriving at that meeting having consulted an AI assistant. Their arguments are polished and their recommendations reassuringly similar. Have several people reached the same conclusion independently, or have they brought versions of the same answer into the room?

Are we agreeing with each other, or just bringing the same AI answer to the meeting?

As with social media and advertising, I think we should ask who is shaping the conversation and whose interests it serves. If Artifical Intelligence can influence what we buy, believe or decide in the masses, there is an incentive to monetise that influence. There is also a risk that someone could exploit it, making carefully directed messages feel like independent advice.

Perhaps more diverse AI could help: systems designed to challenge assumptions, consider different perspectives and reach beyond the most familiar answer. Or perhaps each of us could have an assistant whose approach we can shape, with instructions to show its sources, acknowledge uncertainty and disagree when necessary. But personalisation has a catch. If I tune my assistant until it reflects all my preferences, have I taken back control, or simply built myself a more comfortable echo chamber?

An assistant that always agrees with me might be excellent for my ego, but less useful for my judgement.

For now, I would start with something simpler. Form a view before asking AI for one. Use it to test your thinking, not just tidy it up. In meetings, hear people’s ideas before circulating the machine’s summary, and leave room for the person who sees things differently.

Most importantly, keep talking to people whose experiences are not your own. I would rather AI helped us ask better questions than made us too comfortable with the answers.

Perhaps better AI will solve some of the problems AI creates. But the cure for something is not always more of the same. Before we appoint another chatbot to rescue us from the first one, we might try the slightly old-fashioned approach of thinking for ourselves.

And if your AI insists that all your ideas are brilliant, ask someone who knows you well. They will usually restore balance, free of charge.

Please sign up for useful updates - Sign Up