When an AI and a person make the identical mistake, why does the human still get most of the blame?

This looks at how much an AI seems to act on its own, and how that changes the way outside observers hand out blame and credit to the AI compared with the humans around it. The short answer from the corpus: perceived autonomy does move blame toward AI, but not as far as you'd expect. Humans stay the default target. In a vignette study, participants gave AI more causal responsibility as its agency increased. Yet when a human and an AI did exactly the same thing, people still judged the human as more responsible Does AI autonomy change how we assign blame?. So autonomy turns the dial without taking away the human's place at the center of blame.

That leftover bias has real consequences, and an older idea from automation research explains them. Elish's 'moral crumple zones' describe what happens after automated systems fail. Investigators and the media blame the nearest human operator even when that person had little authority or information, and the technology's reputation comes through intact Do automated systems shift blame onto their human operators?. Put these two findings side by side and you see an uncomfortable pattern. Making an AI more autonomous may not protect its operators much, because observers still look for a person to absorb the failure.

What observers believe can also differ from what the people involved feel. Leonardi finds that when humans credit AI with independent judgment, they feel *less* in control and less responsible, even though their formal authority hasn't changed Does attributing agency to AI actually reduce human sense of control?. This opens a possible gap. The operator feels that responsibility has moved to the machine, while bystanders and investigators keep holding the operator accountable. Something similar shows up in authorship: users feel ownership of AI-written text in proportion to how much they actually shaped it Does user control over AI text shape feelings of ownership?. That makes control, more than the label on who produced the work, the thing people use to judge both credit and blame.

It also matters *why* an AI seems autonomous. One line of work separates anthropomimesis, where designers build in human-like features, from anthropomorphism, where users read human qualities into a system. Each one points responsibility at a different party: designers in the first case, users or user education in the second Who bears responsibility when AI seems human-like?. Judgments also depend on whether people know the source. People rated AI-written moral arguments highly until they learned an AI wrote them, and then their agreement dropped Do people prefer AI moral reasoning when they don't know the source?. Labeling a decision as AI-made can change how it's judged, separately from its content.

One limit: the corpus has only one direct study of third-party blame. The rest comes from nearby work on operators, authorship, and design. The surprising takeaway is that giving AI more autonomy doesn't cleanly hand responsibility over to it. It can widen the gap between who feels responsible and who gets blamed. That helps explain why some researchers argue for keeping humans meaningfully in the loop, so that accountability matches real control Should AI systems stay collaborative rather than fully autonomous?.