Before we start, a quick gut check:
Who do you lie to more: humans or AI?
You probably have an instinctive answer. Hold onto it — because it sits right at the centre of a surprisingly old behavioural quirk now colliding with a very new world.
The Interviewer Effect: We’ve Always Adjusted Ourselves
Back in the 1970s and 80s, social scientists noticed something a bit awkward.
Two interviewers could read the same script, in the same room, asking the same questions — and walk away with completely different answers.
Not tiny differences. Big, personality-sized ones.
In classic studies on racial attitudes, people offered polite, socially acceptable opinions to white interviewers. Swap in a Black interviewer — same script — and suddenly those answers shifted.
People reported higher incomes to female interviewers. Fewer unhealthy habits to stern interviewers. More emotional difficulties to warm ones.
It wasn’t deception. It was calibration.
Humans don’t just respond to questions; we respond to who’s asking them.
And that’s the Interviewer Effect in a nutshell.
**AI Was Supposed to Remove That.
Instead, It Started the Whole Thing Again.**
Fast-forward to now.
HR teams, universities, clinicians — anyone who deals with selection or assessment — loved the idea of using AI.
Finally, a neutral observer. No judgement. No awkward pauses. No raised eyebrows.
But the new PNAS study shows something they didn’t factor in:
People change their behaviour when they think an AI is evaluating them.
Across 12 studies and 13,000 participants, the pattern was consistent:
-
More structured
More logical
More clipped
Less emotional
Less spontaneous
Slightly more robotic (in the “I prepped too hard for this” way)
Not because the AI demanded it. Because people assumed that’s what the system wanted.
It’s not bias entering the model. It’s bias entering the interaction.
A very human instinct — just aimed at a non-human audience.
The Interesting Part Isn’t Human or AI — It’s What Happens Between Them
Most conversations about AI are about the technology: how smart it is, how accurate, how fair.
But the more revealing question is: what do humans do when it’s in the room?
Because the moment people think they’re being observed — by anyone or anything — their behaviour shifts.
That’s where things get interesting.
Not in the human alone. Not in the AI alone. But in the little adjustments we make in response to each other.
It’s less “AI will change society” and more “AI will change how we behave, which then shapes society.”
That quieter, in-between space — that’s where everything moves.
The Cliff-Hanger
So let’s circle back to the question we started with:
Who do you lie to more: humans or AI?
No judgement. No right answer. Just a small behavioural mirror held up to the way you shift depending on who’s watching.
If you’re up for it, drop your answer below:
Human or AI — who gets the more honest version of you?