Does synthesizing data from a small survey beat analyzing the sample directly? A test on a rating scale: dataset

This repository holds the data underlying the study "Does synthesizing data from a small survey beat analyzing the sample directly? A test on a rating scale." The work tests whether synthetic data generated from a small survey sample supports better analytic conclusions than analyzing the small sample directly, using a 22-item AI empathy scale as the test case. The data are 142 de-identified responses to the AI Cognitive Empathy Scale (AICES), each item scored 1 to 5, with reverse-worded items already recoded. Responses were collected under Michigan Technological University (MTU) project IRBNet 1722087 (PI S. T. Mueller), determined exempt by the MTU Institutional Review Board. Participants provided written informed consent online, which permitted the sharing and reuse of de-identified data. No identifying information is included. Coder ID is a sequential index, not the actual coder ID. The analysis and generator code are provided as appendices with the published article.