Agentic education · Learning science

Actionable companion · Assessment & reasoning

Learning with agency: Virtual Patient Simulations Using Social Robotics Combined With Large Language Models for Clinical Reasoning Training in Medical Education: Mixed Methods Study

AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not claims or recommendations attributed to the source authors. No independent peer review of this guide is asserted.

Read original paperSearch paper agentCC BY 4.0

Evidence and context

Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Higher education. Health-professions learning must be distinguished from K–12 instruction.

Original authors’ abstract, reproduced under the linked source license:

Background Virtual patients (VPs) are computer-based simulations of clinical scenarios used in health professions education to address various learning outcomes, including clinical reasoning (CR). CR is a crucial skill for health care practitioners, and its inadequacy can compromise patient safety. Recent advancements in large language models (LLMs) and social robots have introduced new possibilities for enhancing VP interactivity and realism. However, their application in VP simulations has been limited, and no studies have investigated the effectiveness of combining LLMs with social robots for CR training. Objective The aim of the study is to explore the potential added value of a social robotic VP platform combined with an LLM compared to a conventional computer-based VP modality for CR training of medical students. Methods A Swedish explorative proof-of-concept study was conducted between May and July 2023, combining quantitative and qualitative methodology. In total, 15 medical students from Karolinska Institutet and an international exchange program completed a VP case in a social robotic platform and a computer-based semilinear platform. Students’ self-perceived VP experience focusing on CR training was assessed using a previously developed index, and paired 2-tailed t test was used to compare mean scores (scales from 1 to 5) between the platforms. Moreover, in-depth interviews were conducted with 8 medical students. Results The social robotic platform was perceived as more authentic (mean 4.5, SD 0.7 vs mean 3.9, SD 0.5; odds ratio [OR] 2.9, 95% CI 0.0-1.0; P=.04) and provided a beneficial overall learning effect (mean 4.4, SD 0.6 versus mean 4.1, SD 0.6; OR 3.7, 95% CI 0.1-0.5; P=.01) compared with the computer-based platform. Qualitative analysis revealed 4 themes, wherein students experienced the social robot as superior to the computer-based platform in training CR, communication, and emotional skills. Limitations related to technical and user-related aspects were identified, and suggestions for improvements included enhanced facial expressions and VP cases simulating multiple personalities. Conclusions A social robotic platform enhanced by an LLM may provide an authentic and engaging learning experience for medical students in the context of VP simulations for training CR. Beyond its limitations, several aspects of potential improvement were identified for the social robotic platform, lending promise for this technology as a means toward the attainment of learning outcomes within medical education curricula.

What kind of evidence is this?

Embodied learning agents · Qualitative / mixed methods

Stage: Higher education. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.

Study methods · source passage

An interventional, explorative, proof-of-concept study was conducted between May and July 2023. We combined quantitative and qualitative methodology to assess medical students’ self-perceived CR training using a previously developed index for VP design evaluation [26] as well as in-depth interviews.

Limitations and uncertainty · source passage

The theme “procedural limitations” consisted of two subthemes: (1) technical limitations and (2) user-related challenges. Within the subtheme “technical limitations,” students expressed technical problems such as being interrupted by the robot during conversational pauses that were conceived as the end of a spoken sentence. This resulted in the perception of a time constraint while responding to the platform, which sometimes hindered the flow of the interaction (Textbox 3). The subtheme “user-related challenges” sheds light on challenges relating to language barriers as well as a need for adapting to the conversational style of the social robot (Textbox 3).

Design the system before the pilot

Compare the embodied interaction with an equivalent human-led or screen-based activity. Separate the effects of novelty and social presence from learning. Review physical access, sensory needs and adult supervision.

  • Use common rubrics, blind scoring where feasible and a fresh unaided reasoning task. Document permissible assistance and avoid treating AI-detector scores as proof of misconduct.

A 90-day district pilot

  1. Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
  2. Days 16–30 · Design. Choose one disciplinary task. Require a first attempt, a documented AI consultation, a critique of the output and an independent explanation. Align the comparison group and scoring rubric with the same objective. Name an accountable owner and set the baseline, success threshold and budget before implementation.
  3. Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
  4. Days 61–90 · Decide. Blind-score reasoning and accuracy on a new task without AI. Distinguish model examination scores, student perceptions and actual learner gains; only the last directly tests educational benefit. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.

The board decision brief

Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.

Attribution and reuse

Alexander Borg; Carina Georg; Benjamin Jobs; Viking Huss; Kristin Waldenlind; Mini Ruiz; Samuel Edelbring; Gabriel Skantze; Ioannis Parodis. Virtual Patient Simulations Using Social Robotics Combined With Large Language Models for Clinical Reasoning Training in Medical Education: Mixed Methods Study. Journal of Medical Internet Research (2025-03-03). https://doi.org/10.2196/63312

Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.