Leibniz PhD Network - Power Abuse Survey 2025 - Anonymised Consented Dataset

Survey purpose and structure The purpose of this survey is to address the prevalence of Power Abuse (PA) in German academia, as reported by doctoral researchers (DRs) working at Institutes affiliated to the Leibniz Association (German non-university research association). Additionally, this survey aims at identifying mechanisms of PA manifestations and educating the DRs about support structures assisting with prevention of PA. To ensure unbiased answers, the survey was designed as follows: first, the free-text blocks were provided allowing participants to describe their own experiences without previous priming. The following questions were defined in more detail, providing multiple-choice answers. The first page of this survey is a consent of using sensitive data, followed by a common definition for PA ensuring all responses align with the same baseline. Dataset anonymization, randomization and structure To safeguard participant confidentiality, two steps were implemented prior to downstream analysis: Anonymization of the raw free-text answers: Names of people, institutes, machines, analytical techniques and funding programs were omitted. Personal pronouns were replaced with gender neutral pronouns unless the testimonial was describing gendered issues. Descriptive characteristics or rare attributes such as specific personal roles in the workplace were generalized. Randomization of the complete dataset The complete dataset was divided in 12 sub-datasets to avoid identification of the participants by combining information from the demographics section and personal testimonials. The “Leibniz_PhD_Network-Power_Abuse_Survey_2025-000-Multiple_Choice_and_Demographics.csv” sub-dataset mirrors the structure of the complete dataset. The order of the answers was randomized separately for each sub-dataset and merged to a final master table. Briefly a python script ("RndTableGenerator.py") was used to assign a random number to the rows of each sub-dataset. The “randomized ID were merged based on the “original ID” of the complete dataset.