Does perceived task meaninglessness drive worker desire for AI delegation?

The study surveyed 202 U.S.-based workers on 171 O*NET-derived tasks they rated as highly familiar (620 task ratings total, across 22 occupations and 12 sectors), using a five-item scale adapted from Graeber's theory of "bullshit jobs." Workers rated whether a task "feels pointless," whether "nothing important would change" if they stopped doing it, whether it is "disconnected from organizational goals," would be "embarrassing to explain... to someone outside my field," or is "performed only to satisfy bureaucracy or appearances." The authors report the central finding directly: "perceived bullshitness strongly predicts desire for AI delegation," and the same bullshit-rated tasks "are also seen as requiring less human oversight."

The measure itself was validated before being used this way: Exploratory Factor Analysis found a single-factor structure explaining 59.7% of variance, with Cronbach's alpha of .877, meaning the five items cohere as one "bullshitness" dimension rather than several. The authors' reasoning for why this predicts delegation preference is that tasks perceived as disconnected from meaningful outcomes become "natural candidates for AI delegation, aligning worker preferences with perceived feasibility" — where a worker already feels a task doesn't matter, they are both more willing to hand it to an agent and more comfortable letting the agent run with less checking.

This paper builds directly on the Human Agency Scale, adapting two of its items to measure automation desire and required agency, then adding perceived task meaning as a new predictor sitting upstream of What collaboration level do workers actually want with AI?. Where HAS maps which collaboration level (H1–H5) workers want for a given task, this paper asks why a worker lands on a given level — and offers perceived meaninglessness, not task type or technical capability, as one answer. It also sets up What agent traits do workers prefer for meaningless tasks?, which asks not whether workers want delegation but how they want it carried out. The paper's own framing — using bullshitness to "dynamically configure agent autonomy and interaction style" — gives human-centered weight to the same question What UX principles do workplace users want in AI agents? addresses from a design-principle angle.

The excerpt does not establish that bullshitness is a stable or universal property of a task: the authors are explicit that "perceived bullshitness is context-specific" and varies by culture, organization, or period, and is "not a synonym for a task's objective value" — a task can feel meaningless to the worker while still serving functions like compliance or accountability. The sample is US-only, self-reported, and measures perceived rather than technical automation feasibility, and the finding is correlational: bullshitness predicting delegation desire does not establish that removing the task, rather than automating it, would be workers' preference, nor that these findings generalize outside the 22 occupations sampled. The implication the authors draw is correspondingly modest — bullshitness as an "audit lens" to complement, not replace, capability-based automation criteria.