Mapping trust in nurses with dimensions of trustworthy artificial intelligence: A scoping review

OBJECTIVES The overarching goal of this study is to map and synthesize the evidence on dimensions of trust that are perceived by patients to be important in their relationships and/or interactions with nurses in order to inform and envision novel approaches to developing trustworthy artificial intelligence (AI) applications . The aim is to leverage the longstanding public trust that nurses are perceived to hold and develop insight into what features of trustworthiness, if any, could potentially be built into or demonstrated by artificial intelligence models. Achieving the goal of the study will require completion of the study objectives, to: i) scope and synthesize current knowledge from the nursing trust literature to generate new insights into important human dimensions of trust in relation to nurses’ practice; ii) map and compare understandings of human and computational AI understandings of trustworthiness; iii) identify and prioritize knowledge and research gaps and recommend future research directions for development of trustworthy AI; and iv) synthesize results and develop research outputs in formats that will be readily accessible and contextually relevant for target audiences. BACKGROUND The use of artificial intelligence (AI) technologies in healthcare presents unique challenges and opportunities for transforming health and social services. AI has been described as “… the science and engineering of making intelligent machines, especially intelligent computer programs” (McCarthy, 2007). The implementation of AI is often envisioned to produce transformative outcomes, increasing efficiencies and supporting personalized and data-driven decision making (Brennan & Bakken, 2015; Ronquillo et al., 2021). Unfortunately, examples of AI algorithms that have produced harmful outcomes erode what growing public trust there may be. One example is a risk score algorithm that used health care costs as a proxy for illness severity; because less money is spent caring for Black patients compared to white patients, the algorithm treated Black patients as less sick (Obermeyer et al., 2019). While AI offers a lot of potential, building trustworthy AI is a substantial current challenge that must be overcome if AI is to achieve what it is envisioned to do. Focus is rightly being paid to developing computational approaches to develop trustworthy AI that target dimensions of trust such as safety, robustness, non-discrimination, fairness, explainability, privacy, etc., (Liu et al., 2021). the tendency for AI trustworthiness work to consider largely technical perspectives is arguably an important barrier; the human aspect of trust, is largely not an area of focus. there is a pressing need to understand the human aspects of trust and integrate these with computational understandings of trustworthiness, if we are to build AI systems that are deemed truly trustworthy. Integrating understandings of trustworthiness from clinical and computational perspectives require interdisciplinary perspectives and approaches. Year upon year, public opinion polls find nurses to be the most trustworthy profession (Stutzer & Rodriguez, 2020). Besides, the literature on nursing trust highlights the importance of human dimensions that facilitate trust such as nurses’ level of knowledge and being guided by evidence, and acting ethically and with empathy (Dinç & Gastmans, 2013; Rørtveit et al., 2015). The expertise of nurses in fostering trusting relationships and the breadth of literature on nursing trust is a valuable resource that has, so far, remained underexplored for the purpose of developing trustworthy AI. METHODS The updated Joanna Briggs Institute (JBI) methodology for scoping reviews (Peters et al., 2020) that incorporates enhancements by Levac et al. (2010) and provides methodological clarity in response to criticism to Arksey and O’Malley’s approach will be used. In adherence to this methodology and to increase the transparency in the process, the protocol for the scoping review will be developed and registered through the Open Science Framework. A draft CINAHL search strategy was developed which combined the concepts nursing and trust. The search strategy was developed and conducted by a nursing specialist librarian and information specialist. Final refinement of the search strategy was conducted through the through the Peer Review of Electronic Search Strategies guidelines (McGowan et al., 2016) where a librarian external to the team will reviewed the search strategy. Feedback from the PRESS process were incorporated into the search strategy. Databases searched include Cumulative Index to Nursing and Allied Health Literature (CINAHL), EMBASE, MEDLINE, PsycINFO, and forward and backward citation searching. The searches were conducted in October 2022. Results from the completed searches will be uploaded into the Covidence literature synthesis software to remove duplicates and screen articles. Screening and selection will adhere to the pre-specified inclusion and exclusion criteria outlined in the study protocol (e.g., exclusion of grey literature). A minimum of two researchers will review all titles and abstracts for relevance and inclusion. Included articles will proceed to full-article review and further evaluation of inclusion. Any disagreements will be reviewed with the PI and resolved through consensus. Data extraction will be completed by two researchers using a piloted data extraction form including standard categories (e.g., author(s), year of publication, study aims/purpose, population and sample size (if applicable), methodology and methods used, etc.) and key findings relating to the scoping review question (Peters et al., 2020). A thematic analysis approach and iterative synthesis of findings by the team will be used to guide the collation and summarization of findings in order to identify themes - an adaptation of the approach by Crampton et al. (2016). First, a minimum of two researchers will read and annotate each article using computational dimensions of trustworthy AI as identified by Liu et al. (2021) (safety & robustness, non-discrimination & fairness, explainability, privacy, accountability & auditability, and environmental well-being). Where there is uncertainty in whether a dimension of trustworthiness identified in the articles are captured within existing categories and/or whether a dimension fits under more than one category, the PI will discuss with the student researchers and the team will make a decision on how the dimension will be categorized and whether any new dimensions need to be named. Next, the team will review all of the articles, compare and finalize their annotations of trustworthiness dimensions as they correspond with the articles. Points of disagreement will be discussed as a team until a consensus is reached. Results: The scoping review and the journal article will be completed in 18 months (est April 2024). All-team meetings will be held at regular intervals, which will bring together the student researchers and interdisciplinary collaborators at key points during the study. . The results will describe where research has been done, what AI technologies have been developed and studied, how these technologies have been evaluated and how nurses have participated and how ethical issues have been addressed in the research. Conclusion: Research findings will contribute to the larger vision of developing trustworthy AI in health and social systems and enrich understandings of dimensions of trust that may have computational parallels for trustworthy AI systems. References Arksey, H., & O'Malley, L. (2005). Scoping studies: towards a methodological framework. International Journal of Social Research Methodology, 8(1), 19-32. Brennan, P. F., & Bakken, S. (2015). Nursing Needs Big Data and Big Data Needs Nursing. J Nurs Scholarsh, 47(5), 477-484. https://doi.org/10.1111/jnu.12159 Crampton, N. H., Reis, S., & Shachak, A. (2016). Computers in the clinical encounter: a scoping review and thematic analysis. Journal of the American Medical Informatics Association, 23(3), 654-665. https://doi.org/10.1093/jamia/ocv178 Dinç, L., & Gastmans, C. (2013). Trust in nurse–patient relationships: A literature review. Nursing Ethics, 20(5), 501-516. https://journals.sagepub.com/doi/10.1177/0969733012468463?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%3dpubmed Levac, D., Colquhoun, H., & O’Brien, K. K. (2010). Scoping studies: advancing the methodology. Implementation Science, 5(1), 1-9. Liu, H., Wang, Y., Fan, W., Liu, X., Li, Y., Jain, S., Liu, Y., Jain, A. K., & Tang, J. (2021). Trustworthy AI A computational perspective. arXiv preprint arXiv:2107.06641. https://arxiv.org/abs/2107.06641v3 McCarthy, J. (2007). What is artificial intelligence? McGowan, J., Sampson, M., Salzwedel, D. M., Cogo, E., Foerster, V., & Lefebvre, C. (2016). PRESS Peer Review of Electronic Search Strategies: 2015 Guideline Statement. J Clin Epidemiol, 75, 40-46. https://doi.org/10.1016/j.jclinepi.2016.01.021 Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447-453. https://doi.org/10.1126/science.aax2342 Peters, M. D., Marnie, C., Tricco, A. C., Pollock, D., Munn, Z., Alexander, L., McInerney, P., Godfrey, C. M., & Khalil, H. (2020). Updated methodological guidance for the conduct of scoping reviews. JBI evidence synthesis, 18(10), 2119-2126. Ronquillo, C. E., Peltonen, L. M., Pruinelli, L., Chu, C. H., Bakken, S., Beduschi, A., Cato, K., Hardiker, N., Junger, A., Michalowski, M., Nyrup, R., Rahimi, S., Reed, D. N., Salakoski, T., Salanterä, S., Walton, N., Weber, P., Wiegand, T., & Topaz, M. (2021). Artificial intelligence in nursing: Priorities and opportunities from an international invitational think-tank of

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.