Generative artificial intelligence (AI) has entered doctoral research, yet academic libraries lack signals on which AI competencies doctoral students most need training in. This study reports a needs assessment of doctoral students' self-assessed knowledge and training needs for using generative AI in research, designed to inform library AI-literacy services. An experimental bipolar response scale asked respondents to position themselves between needing training and possessing sufficient knowledge across 31 competencies in five areas: basic AI knowledge, AI use in the research workflow, ethics and law, verification, and prompt engineering. Each response was recoded into mutually exclusive need and knowledge components, which were aggregated across competencies to provide separate summaries of perceived training need and self-assessed sufficient knowledge. The analytic sample comprised 169 complete cases in a pooled multi-site sample with unequal representation across five Polish universities. Prompt engineering was the most stable training priority, with a probability of 0.99 of ranking first; verification, ethics and citation, and research-workflow use formed a closely spaced second tier, and basic AI knowledge was lowest. The highest item-level needs spanned large language model settings, programming beyond chatbots, citing AI-generated content, and protecting personal data, indicating that prompt engineering should be taught within a broader research-information practice. A latent class analysis identified high-need, mixed, and high-knowledge response profiles useful for differentiated training. Self-assessed need and knowledge did not predict intention to develop AI skills, so the instrument is best read as a prioritization tool rather than an adoption model. The findings give academic libraries an empirical foundation for sequencing and differentiating doctoral AI-literacy instruction within their information-literacy initiatives.
For librarians with faculty status, issues of burnout may be compounded by the additional pressure of balancing librarianship with the service and scholarship expectations required for tenure and promotion, without the flexible schedules afforded to teaching faculty. Workload policies may be one tool for addressing the structural causes of burnout, but little guidance exists on the creation of such policies. This case study explores these issues by tracing the development of a library faculty workload policy at a public R2 university. Following unionization of tenure-line library faculty, library administrators and faculty were tasked with creating a workload policy, in accordance with the requirements of the new collective bargaining agreement. Development of the policy involved frank and open conversation with library faculty about issues of overwork and burnout, review of workload policies from other libraries, and critical examination of librarian workloads, as well as conversion of workload measurement metrics used by teaching faculty to units more meaningful for librarian schedules. This study suggests that, while workload policies can be an important structural tool for combatting overwork and burnout, their success depends on good faith collaboration and open communication with stakeholders throughout their development, and thoughtful enforcement following their implementation.
As generative AI (GenAI) tools have proliferated, coursework writing in higher education has changed profoundly. This study explores the behavioral patterns, user experiences, information evaluation behaviors, and awareness of academic integrity among university students who use GenAI to write course papers. The study used qualitative methods and conducted semi-structured interviews with 15 students from different disciplines. Interviewees reported that GenAI is integrated into the entire coursework-writing process. They use it most often for task disassembly and conceptualization, but less often for final review. They recognize GenAI's value in lowering the cognitive threshold, providing customized interpretation, and stimulating new perspectives, while worrying that over-reliance weakens independent thinking. Regarding information evaluation, they doubt the references generated by AI, but relatively trust the interpretation of the basic concepts, and deal with the unreliability of information through multi-tool cross-verification, reverse traceability, and other strategies. Thus, universities should shift assessment focus from text fluency to research innovation and argumentation. Academic libraries should also strengthen targeted AI literacy education and tailor AI usage norms to disciplinary characteristics to guide students in using GenAI responsibly and efficiently while upholding academic integrity.
