Does AI collaboration drain motivation when workers return to solo tasks?

Four online experimental studies (total N = 3,562) on human-generative-AI collaboration in professional settings found that "collaboration with GenAI enhanced immediate task performance," but when workers then moved to performing similar tasks independently, this came with a cost: "transitioning from collaboration with GenAI to solo work led to an increased sense of control of human workers, and was also accompanied by significant decreases in intrinsic motivation and increases in feelings of boredom." The tasks were text-based professional or creative work — composing emails, drafting performance reviews, generating product-improvement ideas.

The paper frames the three psychological outcomes through Self-Determination Theory, which holds that intrinsic motivation depends on three needs: autonomy, competence, and relatedness. Sense of control rises after GenAI collaboration ends because solo work restores the worker as "the primary agent" of their actions, after AI's contributions had diminished perceived autonomy during the collaborative phase. But that same restored autonomy does not restore motivation: the paper's reasoning is that GenAI absorbs the "intrinsically motivating parts" of a task — the example given is a performance review, where GenAI taking over the "analyzing and crafting processes" removes the part that made evaluating someone's strengths and weaknesses engaging in the first place. Once removed, solo work on the remaining, less-stimulating parts of the task reads as more mundane by contrast, which is the paper's proposed route to the measured rise in boredom.

This splits what Does AI assistance help workers learn lasting skills? treats as a single non-transfer finding (Wu et al., on an overlapping set of content tasks — Facebook posts, performance reviews, welcome emails) into two separable effects: performance does not transfer, and on top of that, the worker's subjective experience of the solo work degrades along a different axis (motivation, boredom) even as their sense of agency improves. The earlier note documents the skill/performance side; this one adds the psychological-cost side the same underlying line of research reports happening simultaneously. It also complicates Which workplace cues survive AI mediation and which disappear?: there, effort and uncertainty recede into AI-mediated deliverables largely unnoticed; here, the paper argues workers notice the loss of the motivating "fun stuff," as the opening quote in the excerpt puts it, once they are back to doing the residual work alone.

The excerpted material is the paper's introduction and hypothesis section (RQ1–RQ3) plus its abstract; it states the direction and significance of each effect but gives no effect sizes, no breakdown of the N = 3,562 sample's composition, and no detail on how "solo work" or boredom/control/motivation were operationalized across the four studies. The claim should be read at the strength the abstract supports — a replicated directional pattern across four experiments — not as a measured magnitude. If the effect generalizes beyond the email/review/brainstorming tasks tested, it implies that redesigning hybrid human-AI work for sustained engagement requires deliberately preserving some "motivating parts" of a task for the human, not just reallocating effort to AI wherever it is capable.