And why the rage depends more on your day than on the machine
Anyone who works with AI has felt it: a flash of rage that seems out of proportion to a chatbot getting something wrong. What’s interesting is that the intensity of that anger often tracks our own state more than the AI’s actual behavior.
A few mechanisms stack on top of each other. AI is fluent and confident, which sets an expectation of competence — so when it misreads us, ignores a constraint, or states something wrong with total assurance, the gap between how capable it sounded and how it performed is jarring. Expectation violation is a core anger trigger. And because the interface is language, the brain routes the interaction through its social systems: a failure feels less like a tool breaking and more like a person refusing to listen, which is far more infuriating. It’s also a black box. With an ordinary tool we can usually see why it failed and adjust; here we can’t form a clear fix, so repeated failures produce a helpless, looping frustration. The effort is asymmetric, too — you explain, it drops half of what you said, you re-explain, it breaks something else — and each cycle feels like wasted effort.
But the most telling part is how the intensity tracks our own state. Frustration tolerance isn’t fixed; it’s a resource that depends on physiology and mood. When we’re tired, hungry, stressed, or already emotionally activated, our capacity for top-down regulation drops, so the same trigger crosses the anger threshold that it wouldn’t have on a good day. It’s less that the AI changed and more that the buffer shrank. Two things compound this: working with AI is cognitively taxing — holding context, judging output, correcting it — and that load is most aversive exactly when reserves are low; and in a negative state we read ambiguous things more hostilely, so an imperfect-but-neutral response gets interpreted as the machine being obtuse. Then a feedback loop kicks in: anger is itself depleting, which lowers the threshold further, making the next glitch land harder.
The takeaway is almost reassuring. The rage isn’t really caused by the AI. The AI’s friction is fairly constant — what changes is how much of it we can absorb before tipping over. On a bad day, the machine didn’t get worse. Our tolerance just got thinner.