Do people hand decisions to machines because they're too busy to think, or because there's no reward for thinking harder?

This explores two different explanations for why people hand decisions to machines: because their attention is limited and has to be rationed, or because they lack the motivation to do the work themselves. The collection makes the attention case directly and strongly. The incentive case shows up only indirectly, so the honest picture is lopsided.

The attention argument is the surprising one, because it makes delegation rational even when the human would decide better. In Do managers delegate to machines even when human judgment is better?, judgment is treated as a costly resource. Careful thinking takes deliberate effort, and a manager only has so much of it. So the right question isn't "who decides this one case better?" It's "where should my limited judgment go across everything I'm responsible for?" Handing a decision to a machine that does slightly worse can still be the smart choice overall, because it frees attention for decisions where human judgment matters more. Here delegation isn't laziness or a broken incentive. It's how you budget a resource you're short on.

The nearest thing to an incentive story in the collection is about motivation rather than pay. Does perceived task meaninglessness drive worker desire for AI delegation? finds that workers most want to hand off tasks they see as pointless or disconnected from their organization's goals. They also think those tasks need less human oversight. That is a different mechanism. The attention story says "I'd do this well, but my focus is better spent elsewhere." The meaninglessness story says "I don't care enough about this to do it well, or to check what the machine does." The difference matters in practice. Attention-driven delegation keeps the human watching where it counts. Motivation-driven delegation can quietly remove oversight from tasks nobody has examined closely, which may be exactly the ones that hide problems.

Neither explanation says when people take control back, and the trust research adds a third factor. What makes people distrust AI agents they delegate to? shows that people stop delegating when an action can't be undone and other people will see it, like sending an email. How much is at stake matters less. So delegation isn't governed only by attention or motivation. People also ask "can I fix this if it goes wrong, and who will see it?"

One lateral twist: the attention argument has a close parallel inside AI systems. Can delegation teach models to manage context more actively? trains models to hand subtasks to helper agents and take back only summaries. That is a model rationing its own limited context window, much as a manager rations attention. The skill carries over to single-agent work, which suggests that learning what to delegate is really learning what deserves your focus. If you want the classic economic version of incentive problems, such as misaligned rewards between the person delegating and the one doing the work, this collection doesn't cover it yet.