A coworker's AI-assisted success prompted learning in shared work—and entitlement in separate work
A Chinese scenario experiment and a three-wave workplace survey found opposite reactions to visible AI-assisted performance depending on task interdependence. The field evidence is correlational, and the experiment had no non-AI comparison.
The short version
- What happened
- In a randomized scenario experiment with 240 full-time Chinese technology employees, high AI-assisted coworker performance sharply increased intended knowledge seeking when tasks were linked, but increased entitlement when tasks were separate.
- Who it affects
- For workers, visible AI success may feel more learnable when colleagues share a workflow and can exchange methods, prompts and verification practices.
- What remains unproven
- The scenario experiment manipulated high versus low AI-assisted performance but included no non-AI condition, so it cannot isolate an effect unique to AI.
Keep in mind
The scenario experiment manipulated high versus low AI-assisted performance but included no non-AI condition, so it cannot isolate an effect unique to AI.
The three-wave field study is observational; employees selected the focal coworker, most constructs were self-reported and unmeasured workplace relationships could confound the paths.
The samples were Chinese technology workers and Chinese employee-supervisor dyads; cross-cultural and cross-industry generality is untested.
The authors reported no funding or commercial conflicts and said raw data would be available on request.
Source analysis · 1 linked source. Primary analysed source date: 30 October 2026.
Read the full report · 4 min The evidence, methods, implications and limitations
Shared outcomes turned comparison into learning
Employees reacted very differently to a successful AI-assisted coworker when their own work depended on that colleague. In a randomized online scenario experiment, 240 full-time technology workers in China saw high or low AI-assisted coworker performance in either linked or separate tasks. With linked tasks, high performance raised knowledge-seeking scores from 3.60 to 5.23 on a seven-point scale; with separate tasks, the difference was small and not statistically significant in the direct group comparison.
The reverse appeared for entitlement. When tasks were separate, high coworker performance raised the average score from 3.14 to 4.52; when work was linked, the difference was not significant. The scenario held effort, review and responsible AI use constant, isolating performance quality and task structure—but every condition involved AI, so it did not test whether the same pattern follows ordinary coworker success.[1]
The field study supports the pattern, not causation
Researchers then surveyed employee-supervisor pairs in November, December and January. Of 380 recruited dyads, 360 completed all three waves. Employees rated one self-selected coworker's AI-assisted performance and task interdependence, then reported knowledge seeking and entitlement; supervisors later rated the employee's work. Under high interdependence, the performance–knowledge-seeking slope was positive (b=0.50). Under low interdependence, the entitlement slope was positive (b=0.33).
Because coworker performance, task structure and reactions were perceived rather than assigned in the field, unmeasured culture, leadership or relationship quality could explain the paths. Performance was a three-item supervisor rating, not objective output, and results come from one country. The practical hypothesis is testable: teams could make AI methods shareable and rewards visibly collective, then measure whether learning rises without resentment. Randomized workplace trials would be needed before treating that as established advice.[1]
The distinction that matters
The same success cue, two work structures
The experiment varied whether a colleague's AI-assisted output was needed for the participant's own task.
Linked tasks
Coworker output feeds your work
High performance prompted more knowledge seeking
Separate tasks
Coworker output does not affect your work
High performance prompted more entitlement
Field survey
360 dyads over three waves
Similar pattern, but causal direction remains unproven