Medical Education, Год журнала: 2024, Номер unknown
Опубликована: Авг. 23, 2024
I have a confession to make. been slow generate an official policy statement for Medical Education about artificial intelligence (AI) because find the discussion terribly boring. Don't confuse that with lack of interest—I consider technology exhilarating, use it routinely, and marvel at its potential.1 either being dismissive—I recognise, appreciate, wish help guard against ethical harms could be done from, among other things, loss intellectual property reinforcement systemic bias.2 However, most AI in publishing (be writing, enabling better faster peer review, or need unscrupulous practices) boil down same basic sentiment: Responsible effective practice requires human oversight. With over 14 500 seemingly viable resources readily available,3 there is great risk overgeneralization will not profess having deep knowledge means through which each has generated. do, however, believe this class technologies, as whole, best conceived tools (that happen proliferating unprecedented speed little empirical testing).4, 5 Some panic rate development creates amounts worry we ourselves become tools, used by computers, but reality are dealing moment very good reasons futurists regard.6 As such, must focus on what all require responsible practice: Human, least biological,7 So let's role group involved journal publication play regard. We encourage authors if when helps strengthen their capacity improve awareness pre-existing literature,8 formulate stronger research questions bolster designs analyses (i.e. any time make scholarship better). going force disclosure every way influenced submissions would impossible craft sufficiently detailed guideline (especially given people often unaware how embedded common software packages). Further, dominant theme our International Editorial Advisory Board's debate issue was requiring such likely increasingly nonsensical, tantamount needing disclose Google, spell-check, keyboard, tool similarly omnipresent academic work. If using fundamental importance your project, then made so should disclosed body paper. That standard, always applied disclosing like nVivo SPSS primary countless databases literature searches: clearly describing aspects efforts readers know understand rigour replicability study. In doing so, course important keep mind, just do routinely can misused, caution investment learning tool's strengths limitations.1, 9, 10 Optimistically, hope position equity field reducing barriers those who publish pages despite first language English. created virtue privileging afford technology, remain hopeful lesser challenge truism computer gets cheaper time.11 There no doubt hallucinates,10 individuals double check claims world while recognising author, computer, accountable final text.12 For reasons, never myself dream submitting paper than one fluent without careful triangulated effort confirm translation said exactly intended to. Whether not, review remains disposal improving work collectively undertaking study.13 Given currently built corpus predominantly English,2 reviewers raise whether project responsibly represents state world. attempts judge adequacy paper's framing. Similarly, inappropriate submit manuscript received device permission aligns confidentiality standard existed decades. Asking question clarify one's thinking contemplate clearer (or more courteous) ways conveying concerns encouraged improves reviewer's offer feedback professional themselves.14 Out curiosity, once submitted some my own writing ChatGPT request 'write rejection letter' (to see predict objections might raise). After response largely parroted back had abstract, instructed try again, stressing wanted letter. Its informative: 'I am programmed critical appraisal.' Even 'knew' curators stewards journal, too pledge reader, reviewer experience. example, enabled implementation 'free format' submission system Education, longer go tedium formatting references specific way; also note making now uploading confirming accurately identified author names, title, abstract on. years detect unethical practices duplicate plagiarism. manage suggests thinks particularly well suited content under consideration. While these systems undoubtedly continue improve, far from perfect. Truly fraudulent behaviour cannot caught existing transgressions generally over-called. editors cognizant value hearing diverse voices inform reviews facilitate truly inclusive community.15 result, automate decision-making empower conduct it. Instead, taking advantage greater flag potential issues opportunities continuing investigate them care thoughtfulness required yield outcomes achieve. less perspective, editorial proof did (eventually) conclude necessary share views sake transparency, however boring non-reactive reinforcing status quo) they be. changes, policies evolve, everyone whatever available health education. That, defines Data sharing applicable article datasets were generated analysed during current
Язык: Английский