Autonomous or Static: The effect of AI-enabled environmental benefits in shaping consumer green purchase intention database
Description
This dataset contains three batches of anonymized online scenario survey data of Chinese adult consumers collected from late November to early December 2025, which supports the empirical analysis of the manuscript exploring the dual impact of AI autonomous environmental benefits on consumers’ green purchase intention based on the Theory of Planned Behavior. Three independent between-subjects experiments adopted intelligent air conditioners, eco-backpacks and energy-saving water heaters as distinct product stimuli to avoid single-product bias, respectively verifying the basic mediating path of environmental self-efficacy, the moderating effect of consumer innovativeness and the boundary role of time orientation. We strictly cleaned invalid samples with overly short filling time and identical consecutive answers, retaining 241, 360 and 362 valid respondents for Study 1, Study 2 and Study 3 separately. All latent variables were measured by mature 7-point Likert scales from published literature, covering perceived AI dynamic environmental benefits, environmental self-efficacy, green purchase intention, consumer innovativeness, time orientation, manipulation check items and demographic control variables (gender, age). The archived files include universal CSV and SPSSraw data, as well as an attached code book explaining variable definitions and coding rules. All original statistical results (ANOVA, bootstrap mediation and moderated mediation tests) can be fully replicated using SPSS and Hayes’ PROCESS Macro. No sensitive personal information was collected, all participants provided voluntary informed consent, and the anonymized data is openly available for non-commercial academic replication and secondary research on AI-driven sustainable consumption.
Files
Steps to reproduce
All data across Study 1, Study 2 and Study 3 were collected through separate online scenario-based between-subject experiments distributed via the Credamo survey platform between late November and early December 2025. Pre-tests were carried out for each experiment to refine product stimuli and validate manipulation check items before formal data collection. Study 1 adopted a single-factor two-group design with intelligent energy-saving air conditioners as stimuli, randomly assigning participants to an AI-driven autonomous environmental benefit group or a non-AI static environmental benefit group to test the main effect and simple mediation effect; Study 2 adopted a 2 (product environmental benefit type) × 2 (consumer innovativeness) between-subjects design using AI eco-backpacks and ordinary degradable backpacks to examine the moderating role of consumer innovativeness; Study 3 adopted a 2 (product environmental benefit type) × 2 (time orientation) between-subjects design with energy-saving water heaters as experimental materials, and added a temporal mindset priming writing task to explore the moderating effect of time orientation. All three studies utilized standardized 7-point Likert scales from existing literature to measure core variables including perceived environmental benefits, environmental self-efficacy, green purchase intention, consumer innovativeness and time orientation, alongside manipulation check items and demographic questions of gender and age. Uniform data cleaning standards were applied to all three batches of samples: responses with filling time shorter than half the average duration and questionnaires with identical options selected for five consecutive items were discarded, resulting in 241 valid samples for Study 1, 360 valid samples for Study 2 and 362 valid samples for Study 3. To replicate this research, scholars can adopt identical experimental scenarios, priming tasks and measurement scales, deploy random group assignment via any mainstream online survey platform, follow the same sample screening rules, and conduct statistical analyses with IBM SPSS 31.0 and Hayes’ PROCESS 4.1 macro to reproduce t-tests, reliability and validity tests, ANOVA, bootstrap mediation and moderated mediation results. No professional instruments, chemical reagents or special hardware were required; the entire research workflow relies on online randomized scenario surveys and desktop statistical software for data sorting and empirical testing.
Institutions
- Yanshan UniversityHebei, Qinhuangdao