Dataset from the Scenario Experiment on the Effect of Generative AI Emotional Support on Employees' Organizational Citizenship Behavior

This experimental dataset was collected through a single-factor, two-level between-subjects scenario experiment. The experiment was conducted on an online research platform without specific geographical restrictions, and participants were recruited from full-time employees with work experience across the country. Data were collected during the implementation phase of the present study, with no detailed timestamps or spatial resolution. Initially, 200 questionnaires were distributed, and all initial responses were retrieved. Data cleaning was performed based on manipulation check results, attention test pass rates, and valid response duration criteria. A total of 20 invalid samples were excluded due to failed manipulation checks, unqualified attention tests, and excessively short response times, resulting in a final valid sample of 180 participants and 180 valid data records.The dataset was collected using questionnaire scales. Devices used included participants’ personal computers, mobile phones, and other terminals; data collection and entry were completed via a general online questionnaire platform; statistical analyses were conducted using SPSS 26.0 with the PROCESS macro. The dataset contains five categories of indicators: control variables, independent variable, mediating variables, moderating variable, and dependent variable. Control variables include gender, age, and work tenure; the independent variable is the source of emotional support (AI group / human coworker group); mediating variables are psychological resources and reciprocity motivation; the moderating variable is agreeableness; and the dependent variable is organizational citizenship behavior. All variables were measured using a 5-point Likert scale, with no physical units of measurement.After cleaning, the dataset has no missing values, and the variable integrity rate is 100%. Data errors mainly stem from participants’ subjective self-rating bias and differences in scenario immersion, and the error range falls within the acceptable interval for psychological experimental research. Data files are in standard Excel or SPSS format and can be directly opened and analyzed by common statistical software such as SPSS, Excel, and R. All relevant software can be obtained free or commercially through official channels, with no rare format limitations.