AI as Manager and Job Preferences: A Field Experiment

Participants are presented with a series of job postings and asked to evaluate them on a numerical scale. The experimental design varies key job attributes, such as work flexibility, salary, and promotion determination, including differences in managerial type. By observing participants’ ratings and trade-offs, the experiment measures preferences over these job and management features.

2025-09-08

2025-10-13

The primary outcomes of this experiment are: (1) participants’ ratings of each job posting on a 1–10 scale; (2) willingness to pay, measured through the salary participants associate with job attributes; (3) trade-offs between key job features, including work flexibility, promotion determination, and managerial type; and (4) participants’ elicited beliefs about managerial promotion patterns (male-favored vs. female-favored), collected either before or after the main evaluation task.

Job ratings are measured as the numerical score (1–10) assigned by participants to each job posting. Willingness to pay is estimated by analyzing how changes in salary influence participants’ evaluations of job attributes. Trade-offs between key job features are constructed by comparing ratings across job postings with different combinations of attributes, including work flexibility, promotion determination, and managerial type. Elicited beliefs are recorded as participants’ stated expectations about managerial promotion patterns, which can be used to analyze how prior beliefs influence job evaluations.

This field experiment recruits participants to evaluate a set of job postings, each varying in key attributes such as work flexibility, salary, promotion determination, and managerial type. Participants rate each job individually on a numerical scale. The combinations of job attributes are randomly assigned across participants to allow measurement of trade-offs and preferences. The design aims to understand how different job and managerial features influence perceived job desirability and willingness to pay.

Not available

Job postings are randomly assigned to participants using Qualtrics with block randomization to ensure that each combination of attributes is presented an equal number of times across participants.

All randomizations in the experiment are conducted at the level of the individual participant. This includes the random assignment of job posting attribute combinations, the information provision, and the timing of the elicitation of beliefs.

No

Not applicable; randomization occurs at the individual level

Approximately 2,000 participants (estimated)

Arm 1: Elicitation of beliefs before main task, Neutral information provision yes – ~500 participants

Arm 2: Elicitation of beliefs before main task, Neutral information provision no – ~500 participants

Arm 3: Elicitation of beliefs after main task, Neutral information provision yes – ~500 participants

Arm 4: Elicitation of beliefs after main task, Neutral information provision no – ~500 participants

For the main outcome of job rating (scale 1–10), the experiment is powered to detect a minimum effect size of approximately 0.22 points. The unit of observation is the individual participant, the standard deviation of the outcome is assumed to be 2 points, and this corresponds to about 2.2% of the scale. Randomization is at the individual level, and no clustering is present.