Public willingness to use a described humanoid robot for emergency department history taking: a cross-sectional US online vignette survey

Preprint of a manuscript currently under consideration at a peer-reviewed journal. This version has not been peer reviewed. Automated medical history collection before clinician evaluation is feasible. Whether the public would accept a humanoid robot for this in an emergency department, or prefer it to a tablet running the same software, is unknown. We measured willingness to choose a described humanoid robot for emergency department history taking, and compared physician, tablet, and robot vignettes. This anonymous cross-sectional survey recruited 254 US adults from the SurveyMonkey Audience panel in July 2026; a synthetic robot image appeared on the landing page. Participants rated willingness before and after five reassurances, then rated physician, tablet, and humanoid-robot vignettes on ease of use, diagnostic confidence, and perceived empathy; modality order was randomized. Willingness before reassurances was modeled on recent consumer artificial-intelligence use and age decade with cumulative-logit regression; ratings used Bonferroni-corrected pairwise Wilcoxon tests. Opinion was split: 43.3% were likely or very likely to choose it, 40.6% unlikely or very unlikely, and 16.1% neutral. After the reassurances the willing proportion rose to 50.8% (P < .001), but the 70 (27.6%) who said no reassurance would matter barely moved. Greater artificial-intelligence use was associated with willingness (odds ratio 1.78 per category, 95% confidence interval 1.51 to 2.09). Contrary to the pre-specified hypothesis, age showed no trend (0.93 per decade, 0.82 to 1.05); adjusted for artificial-intelligence use it reversed direction (1.20, 1.04 to 1.38). A physician was preferred by 58.7%. The robot was rated below the tablet on ease of use (P = .004) and diagnostic confidence (P = .009); empathy was near the floor for both (P > .99). Opinion is divided, not ambivalent; reluctance is not concentrated among older adults; and embodiment bought no advantage over a screen. These are reactions to scenarios and an image, not behavior with a working system. Aggregate data, analysis code, the fielded questionnaire, completed STROBE and CHERRIES checklists, and two verification scripts are deposited separately at https://doi.org/10.5281/zenodo.23027272. The deposited tables are constructed so that every statistic reported here can be independently re-derived from them without the participant-level file. Participant-level responses and free-text answers are not deposited and are not available on request, because the consent page shown to every participant stated that results would be reported only in aggregate. All three authors were associated with Emergency Robotics, a concept-stage project exploring robot-assisted medical history collection, which motivated the survey. At the time of the study there was no company or corporate entity, no outside investment, no author compensation, and no stock or options. The study produced findings that do not favor a humanoid interface over a tablet in this vignette. Ethics: determined exempt by Rapid IRB (IRB00014585; Federalwide Assurance FWA00035509) on 29 July 2026 under 45 CFR 46.104(d)(2).