I've always loved psychology, always been fascinated by how irrational we can be and what unpredictable turns that irrationality takes. The more I learn about how we perceive and work with AI and robots, the more curious I get about this space. Here's your first clue: we're still irrational. We assume a mistake gets judged on its own terms, same error, same consequence, whether a person made it or a machine did. That assumption doesn't survive contact with the research.
Start with trust, because in HRI almost everything comes back to it, and this is no exception.
A 2024 study by Aikaterina Manoli, Janet Pauketat, and Jacy Reese Anthis looked at what happens after a single bad act: one AI assistant caught doing something clearly wrong at work. People didn't just distrust that one AI. Their opinion of AI as a whole category dropped, as if one system's misconduct implicated every other AI on the market. Run the same test with a human assistant behaving just as badly, and people's opinion of humans in general barely moves.
One bad robot ruins robots. One bad employee doesn't ruin employees. The researchers' explanation is almost sad. People see AI as one big undifferentiated group, an outgroup, in a way they don't see other people.
So the public standard is brutal. A robot or an AI gets one shot, and gets punished for its entire category when it blows it. You'd expect that to make companies careful about handing AI any responsibility a human could be blamed for instead.
But that's not what's happening inside the workplace, and this is the part that should actually worry you. Once a company starts framing its AI as a teammate, something flips. Managers catch fewer errors. Individual accountability for mistakes drops. Blame attributed to the AI itself rises to cover the gap. People aren't just avoiding blame themselves, they're quietly routing it toward the one team member that can't object, defend itself, or drag anyone else's reputation down when it takes the fall.
Put those two findings side by side and you get a genuinely strange picture. The public holds AI to an impossible standard, one mistake and the whole category is guilty. And the workplace has already figured out how to use that exact asymmetry as a pressure valve, because everyone quietly understands the AI will absorb blame without consequence and without anyone generalizing it back onto the humans who deployed it.
If you're the one deciding how to introduce AI or robots into your team, the framing you choose isn't cosmetic. Calling it a teammate doesn't just change how people talk about it. The research says it changes who gets caught making mistakes, and who doesn't. The fairness question here was never really about the robot. It's about whether your best people are still checking their own work, or quietly letting something that can't push back check it for them.
Manoli, A., Pauketat, J. V. T., & Anthis, J. R. (2024). The AI Double Standard: Humans Judge All AIs for the Actions of One. arXiv:2412.06040. Read the study →