What if AI agrees with you, but you’re wrong?

Two colleagues walk to their cars after a meeting.

“Well, that was a good meeting,” says one.

“Really?” replies the other. “We spent an hour and a half talking and got absolutely nowhere.”

They sat in the same room. Heard the same people. Saw the same presentation. Yet they experienced completely different meetings.

Who’s right?

Perhaps they both have a point. Perhaps they’re both missing something.

We all see reality through our own lens.

Our experiences, expectations and beliefs help shape what we see and the meaning we give it. That’s human. But here’s where it gets interesting.

And then came AI

We train AI on vast amounts of information, much of it created by people. Research papers, books, websites and all sorts of other texts.

There’s an enormous amount of knowledge in there. But also human assumptions, biases, mistakes and contradictions.

And then we ask that same AI to help us understand how the world works.

Quite a thought.

Because what happens when we bring our own beliefs into the questions we ask?

Suppose you ask an LLM tool:

“My colleague never listens to my ideas. How can I deal with that?”

Chances are you’ll get an understanding response, with useful tips for starting a conversation.

But hang on. Who says your colleague never listens?

Perhaps he had a good reason to reject your proposal. Perhaps you’ve missed something yourself.

AI only knows the story you’ve told it.

A mirror or an amplifier?

Research shows that AI can amplify human biases, especially when systems tend to agree with their users. Researchers call that tendency sycophancy.

But the good news is that AI can also help us think more clearly.

Try asking the same question differently:

“I feel that my colleague is blocking my ideas. What assumptions am I making, and what other explanations might there be?”

Suddenly, you’re using AI to examine your own thinking rather than to prove yourself right.

And perhaps that’s where its real value lies.

We’ve built machines that are getting better and better at answering questions. But the quality of our thinking still depends on our willingness to question our own beliefs.

The next time AI agrees with you, ask one more question:

What if I’m wrong?

Further reading