How does self-evaluations influence others' evaluations? An online experiment

How to cite

Gao, Qianyun. 2026. "How does self-evaluations influence others' evaluations? An online experiment." AEA RCT Registry. AEARCTR-0017261. https://doi.org/10.1257/rct.17261-1.2

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Qianyun Gao*

* Primary investigator

Registry ID
AEARCTR-0017261
DOI
https://doi.org/10.1257/rct.17261-1.2
First registered
Last updated
Status
In development
Trial dates
2026-07-16 – 2027-06-01
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/17261

Abstract

Differences in workplace outcomes can depend on factors beyond actual task performance. This project studies whether the way workers describe their own performance affects how others evaluate that performance. In an online experiment, participants are shown profiles of workers doing either a Raven-style logical reasoning test or a writing test and are asked to rate the workers on their test performance. Each profile contains either 1) the pieces of work with suggested solutions alongside self-evaluations of the worker describing their task performance on this task; or 2) the pieces of work with suggested solutions alongside some descriptive texts that match in length and positivity of language (control group), randomly shown to participants. I hypothesise that exposure to worker self-evaluations influences evaluators' assessments of worker performance. As a mechanism, I explore how the effect varies with aspects of self-evaluations, particularly its positivity.

Intervention

I show participants randomized profiles of workers doing certain tasks. While all profiles contain the worker's actual responses, some profiles also contain workers' self-evaluations.

Experimental design

I plan to recruit participants from Prolific (UK-based working-age adults) as evaluators. Participants who serve as evaluators will be introduced to the tasks with suggested solutions and shown pieces of work, and rate the performances as well as how much they like the workers (on a scale of 0-10). Randomisation occurs at the evaluation level, where evaluators see slightly different versions of task responses. Collection of worker data was separately pre-registered.

Design details

Randomization method
Randomization will be implemented electronically through variables embedded in the survey software. Each evaluator will complete eight evaluations. Assignment will be balanced such that each evaluator sees four workers in the treatment condition and four workers in the control condition. Assignment will also be balanced across task types, with each evaluator evaluating four workers who completed the Raven-style task and four workers who completed the writing task.
Randomization unit
Unit of randomisation is each evaluation (evaluator-worker pair)
Clusters
300 evaluators
Observations
2400 evaluations
Arms
Each evaluator will be exposed to both the conditions. 1200 evaluations will be in control group, 1200 in treatment group.
Primary outcomes
Rating on task performance
Secondary outcomes
The main mediator is positivity of self-evaluations. Additional exploratory variables include perception towards the worker (ie: good to work with); mechanism questions include perceived informativeness of the self-evaluations, and the cognitive load cognitive difficulty of making an assessment. I will also look at free-text responses on the evaluators' thinking-process, and other linguistic features of the self-evaluation text.

Cited by

No citation data available for this registration yet.

Keywords