|
Field
Abstract
|
Before
This online randomized crossover experiment examines how alternative collaboration arrangements and team incentive structures affect effort and performance in a timed visual retail task. Each participant completes four incentivized rounds under four collaboration conditions presented in a randomized, counterbalanced order. Randomization occurs at the individual level through computer assignment to one of four balanced condition sequences. We plan to recruit up to 220 participants, targeting at least 180 analyzable participants.
|
After
This online randomized crossover experiment examines how alternative collaboration arrangements and team incentive structures affect effort and performance in a timed visual retail task. Each participant completes four incentivized rounds under four collaboration conditions presented in a randomized, counterbalanced order. We plan to recruit up to 325 participants, targeting at least 264 analyzable participants.
|
|
Field
Last Published
|
Before
August 19, 2026 09:31 AM
|
After
September 24, 2026 04:05 PM
|
|
Field
Primary Outcomes (End Points)
|
Before
Round-level exact-counting productivity, measured as the number of shelf images answered with the exact correct product count per second of elapsed round time.
|
After
1. Mean item response time within a formal round.
2. Variance of item response time within a formal round.
3. Exact accuracy across all items within a formal round.
|
|
Field
Primary Outcomes (Explanation)
|
Before
For participant i in round r, the primary outcome is correct product count per second.
An answer is correct only if the submitted integer exactly matches the prespecified answer key for the image. Elapsed time is measured from the beginning of the round until the participant completes all available images or the 120-second limit expires. The outcome is zero when no response is correct. The primary analysis compares the four within-participant condition-specific outcomes.
|
After
|
|
Field
Experimental Design (Public)
|
Before
The study is an online, within-participant randomized crossover experiment. Each participant completes four timed formal rounds and experiences each of four collaboration conditions once. Participants are assigned by computer to one of four counterbalanced treatment sequences so that each condition appears in each round position and the principal first-order transitions are balanced across participants.
Every formal round uses the same underlying visual product-counting task, the same time limit, and the same number of images. Formal image stimuli are allocated without replacement from prespecified image pools, with balanced representation of the image groups in every round. Images and their presentation order are randomized within participant.
Adult online participants will be recruited through an online research-participant platform. Participants must provide informed consent, meet the preregistered eligibility requirements, and use a supported desktop or laptop browser. We plan to recruit no more than 220 participants and target at least 180 analyzable participants.
|
After
A 2-by-2 within-participant crossover experiment with individual-level random assignment to one of four counterbalanced treatment orders.
|
|
Field
Randomization Method
|
Before
Participants will be assigned by a computerized randomization procedure, with approximately equal allocation, to one of four prespecified counterbalanced condition sequences. Allocation will occur before the first formal treatment round and will not involve investigator discretion. Within the task, shelf images will be randomized across rounds and presentation positions by a computerized shuffling algorithm, subject to the restriction that each round contains equal representation from each image group.
|
After
Done by a computer before the four Prolific studies are published. Eligible participants are partitioned into four mutually exclusive participant groups in approximately equal proportions. Each group is assigned to exactly one of the four treatment-order studies: A-B-D-C, B-C-A-D, C-D-B-A, or D-A-C-B. The four order-specific studies are then published concurrently. Because the participant groups are mutually exclusive, a participant can see and enter only the study corresponding to the participant's assigned order. Assignment is fixed before treatment exposure and is not chosen by the participant or altered by the researchers after observing outcomes.
|
|
Field
Randomization Unit
|
Before
Individual participant for assignment to condition sequence. Shelf image within participant for allocation to round and presentation position.
|
After
Individual participant.
|
|
Field
Planned Number of Clusters
|
Before
Not applicable. Randomization occurs at the individual-participant level.
|
After
Up to 325 individual subjects. The unit of randomization is the individual participant.
|
|
Field
Planned Number of Observations
|
Before
Up to 220 participants. Each participant is scheduled to provide four participant-round observations and up to 80 formal image-response observations, corresponding to up to 880 participant-round observations and 17,600 image-level observations.
|
After
Up to 325 individual subjects will be recruited. The target is at least 264 analyzable participants.
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
All participants experience all four experimental conditions. Participants are assigned approximately equally across four order sequences, with a maximum target of approximately 55 recruited participants per sequence. The target analyzable sample is approximately 45 participants per sequence, or 180 participants total.
|
After
Participants are assigned to four mutually exclusive treatment-order arms rather than to separate treatment cells, because every participant experiences all four treatment conditions.
The target analyzable sample is approximately 66 participants in each order group, for 264 analyzable participants in total. If the full recruitment cap of 325 is required, recruitment will be distributed as evenly as possible across the four mutually exclusive study/order groups.
|
|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
The study is powered for its smallest principal within-participant factorial contrast. Assuming a standardized within-participant contrast effect of d_z=0.25, a two-sided significance level of 0.05, and 90 percent statistical power, approximately 171 analyzable participants are required. We target at least 180 analyzable participants and plan to recruit up to 220 participants to allow for incomplete participation, technical failure, duplicate participation, and preregistered exclusions.
|
After
The study targets a standardized within-participant effect size of absolute dz = 0.25. A sample of 264 analyzable participants provides approximately 95% power. Recruitment may continue up to 325 participants to achieve the analyzable-sample target after incomplete sessions, technical failures, duplicates, and prespecified exclusions.
|
|
Field
Intervention (Hidden)
|
Before
The study uses a 2 × 2 within-participant design. The two factors are teammate identity and team-payoff architecture. The four conditions are:
A. Human teammate–Additive payoff
B. Human teammate–Multiplicative payoff
C. AI teammate–Additive payoff
D. AI teammate–Multiplicative payoff
In Human-teammate rounds, the interface displays an automatically generated anonymous username and neutral avatar representing the same purported human teammate throughout the experiment. In AI-teammate rounds, the interface displays an AI teammate identified as “the bot.” There is no synchronous live teammate. Teammate scores are predetermined and held fixed across participants apart from the prespecified encounter schedule. The first encounter with each teammate type displays a final teammate score of 49, and the second encounter with that teammate type displays a score of 51.
Before the formal rounds, participants are shown prior-pilot performance benchmarks indicating that the human benchmark and AI benchmark are both approximately 50 on the task’s 0–100 personal-score scale. The displayed underlying score-per-second benchmarks are 0.138 for the human benchmark and 0.139 for the AI benchmark.
Each participant first completes five teaching examples and receives the correct answer after each practice response. Participants then complete four formal rounds. Each formal round lasts up to 120 seconds and contains 20 images. For each image, the participant reports the number of visible product units. Every visible appearance counts as one, including partially visible or occluded products.
The participant’s round score per second is the number of exact correct responses divided by elapsed round time. The personal score is
a = min(max(1000 × (score per second − 0.09), 0), 100).
The teammate score is denoted by b.
Under the Additive treatment:
Team score = (a + b) / 2.
Under the Multiplicative treatment:
Team score = (a × b) / 100 + 25.
Each team-score point produces a $0.02 round bonus. Participants also receive a fixed base payment of $1.00.
The fixed score schedule and the +25 adjustment are intended to keep the expected payoff level and average marginal monetary return approximately comparable across architectures while varying whether the value of the participant’s performance is independent of or contingent on teammate performance.
The study uses deception regarding the presence of a live human teammate. This procedure will be conducted only with Institutional Review Board approval, and participants will receive an appropriate debriefing after completing the experiment.
|
After
We randomize the order in which participants experience four within-participant treatment conditions that vary teammate identity and payoff architecture.
Participants complete one round in each of the following conditions:
1. Human teammate / Additive payoff;
2. Human teammate / Multiplicative payoff;
3. AI teammate / Additive payoff; and
4. AI teammate / Multiplicative payoff.
Under the Additive payoff rule, team score equals (participant score + teammate score) / 2. Under the Multiplicative payoff rule, team score equals (participant score × teammate score) / 100 + 25. The teammate-score distribution is held constant across Human and AI teammate conditions.
|
|
Field
Secondary Outcomes (End Points)
|
Before
Total number of exact correct responses in each round.
Proportion of available images answered correctly.
Accuracy among submitted non-skipped responses.
Number of images answered, skipped, and not reached.
Item-level response time and round completion time.
The transformed 0–100 personal performance score.
Team score and performance-contingent round bonus.
Post-experiment questionnaire scale scores: perceived task interdependence, perceived indispensability, other-oriented performance pressure, accountability, and team identity/solidarity.
|
After
Score per second; proportion of skipped images; post-task comparative mechanism measures.
|
|
Field
Secondary Outcomes (Explanation)
|
Before
Total correct responses are measured out of the 20 images assigned to each round. Accuracy among available images equals correct responses divided by 20. Accuracy among answered images equals correct responses divided by non-skipped submitted answers. Item-level response times are constructed from the elapsed timestamps recorded at successive submissions. The personal performance score, team score, and monetary bonus are deterministic transformations of task performance and the assigned collaboration condition and will primarily be used for descriptive and robustness analyses.
Post-task questionnaire outcomes comprise three multi-item scales—perceived task interdependence, accountability, and team identity/solidarity—each calculated as the unweighted mean of three items, and two single-item outcomes—perceived indispensability and other-oriented performance pressure. Item-level responses will also be retained.
|
After
|
|
Field
Pi as first author
|
Before
No
|
After
Yes
|