Even without any robot takeover, could AI quietly take over decisions while humans still think they're in charge?

This explores whether letting AI take part in decisions slowly moves real influence away from people, even when humans still formally hold the reins. The corpus says yes, it can. The mechanism is subtler than robots taking over, though, and there's disagreement about whether it's a one-way street. The most striking idea is that society has stayed aligned with what people want partly by accident. Institutions need human workers, and those workers care how things turn out. Does incremental AI replacement erode human influence over society? argues that each small AI substitution removes a bit of that hidden check. No single step looks dangerous, but systems can drift away from human preferences, and when several institutions drift together, the drift may become hard to reverse.

The erosion can also happen inside people's heads before it shows up in any org chart. Does attributing agency to AI actually reduce human sense of control? finds that when people start crediting an AI with its own judgment, they feel less in control and less responsible, even though nothing about their formal authority has changed. Leonardi's twist is the hopeful part: this displacement runs in cycles. Delegation and attribution are followed by phases where humans reassert themselves and reorganize, so the loss is manageable rather than permanent. Meanwhile, Do humans learn to prefer AI partners over time? shows how the pull toward AI builds up through experience. People who started out biased against bots came to prefer them as partners over repeated rounds, because the bots were more consistent and reliable. Influence doesn't have to be taken. People hand it over, one sensible choice at a time.

Whether this erosion happens depends a lot on how the AI is designed into the decision. Does automating labor concentrate power more than augmenting it? separates AI that replaces workers from AI that makes them more capable. Replacement shifts bargaining power to whoever owns the technology and can settle into a trap where workers have no leverage left. Similar thinking appears at the level of individual decisions. Can AI guidance reduce anchoring bias better than AI decisions? has the AI point out what's worth noticing instead of handing over an answer. That avoids anchoring people on the machine's verdict and keeps responsibility with the human. Should AI outputs replace or supplement human judgment? makes the philosophical case for the same thing: AI output should count as one piece of evidence to weigh, not a verdict that overrides your own reasoning. How should humans and AI systems divide decision-making tasks? treats the whole thing as an organization-design question: who does which task, and who learns from which feedback.

The less obvious risk is that influence can shift to the wrong signals even when nobody is watching for it. Does confidence drive influence in multi-agent deliberation systems? finds that when AI agents deliberate together, influence goes to whichever agent sounds most confident, not to the one that's most competent. A human relying on such a group could be swayed by consensus produced by false confidence. On the governance side, Can traditional governance tools actually control AI agents? argues that the usual tools for keeping delegated agents in check (monitoring, incentives, enforcement) break down when the agent is opaque and fast. Does AI risk increase with the autonomy we give it? concludes that risk rises steadily with autonomy and argues for deliberately graded levels of AI independence.

So the answer is that AI participation tends to reduce human influence by default, through a slow drift and not a dramatic handoff. But it isn't inevitable. It depends on whether AI replaces human judgment or informs it, and on whether people keep reasserting themselves as Leonardi's cycle suggests. One last point: Can AI ever gain expert community trust through participation? notes that AI can't earn the kind of community-backed authority human experts have. That means any influence it gains comes from our deference, not from earned standing, which also means we can take it back.