More Ways to Teach, Learn, and Collaborate: 82 New Features and Improvements in Clinical Mind AI
Clinical Mind AI keeps growing because our community keeps imagining what is possible. A question from an instructor, an idea from a student, or a conversation between institutions can become a new way to teach, learn, or conduct research together.
Today, we are introducing 82 new features and improvements shaped by those conversations. They give institutions more control, help instructors create richer learning experiences, and give students more ways to practice clinical reasoning and work together.
1. A clearer institutional dashboard. Institutional administrators can see their institution’s users, courses, and public library contributions in one place. Direct links to pending invitations, conversation alerts, and data requests make it easier to follow up on work that needs attention.
2. Annual contribution tracking. Institutions can follow their contributions of patients, rubrics, and clinical case activities to the public library across collaboration years. The dashboard shows progress toward the annual commitment for each resource type, based on the collaboration agreement’s signed date and resources that remain publicly shared.
3. Bulk account creation. Institutional administrators can create users from Excel files. File validation helps identify missing or incorrectly formatted information before accounts are created.
4. Invitation management in bulk. Administrators can resend pending invitations together in one click. This makes it easier to follow up when onboarding a class or a new teaching team.
5. Institutional control over courses. Institutions can decide whether instructors may create courses and change course settings. These controls support both centrally organized teaching and arrangements where instructors manage their own courses.
6. More control over institutional libraries. Institutional administrators can preview and edit their institution’s patients, rubrics, and clinical case activities, including their learning outcomes. Active and archived resources remain organized within the institution’s own libraries.
7. Release activities directly from the institutional library. Institutional administrators can publish clinical case activities from their institution’s library directly into courses, helping coordinate teaching across the institution.
8. Institutional learning outcomes. Administrators can organize learning outcomes by program and course, giving instructors a shared framework for their activities. These connections help bring teaching materials and assessment goals together.
9. A central research workspace. Institutional administrators can find their collaboration agreement, shared Drive folder, list of tasks, and participating research projects in the Research page. The page also explains the institutional information available to Stanford University through the collaboration.
10. Control over medical tests availability. Institutions can define which medical tests are available by default. Instructors can use that selection when building activities, helping simulations reflect the resources relevant to their teaching context.
11. Institutional medical tests costs. Administrators can enter or import medical tests costs and manage the institution’s currency. Instructors can then use those costs in activities that teach students to consider resource use alongside clinical decisions.
12. Institutional oversight of conversation alerts. Administrators can see possible concerns flagged during patient conversations and whether instructors have reviewed them. This view supports follow-up across courses; the instructor remains responsible for reviewing and resolving an alert, and an automated flag is not a finding of misconduct.
13. Fifteen types of institutional data reports. Administrators can request reports covering institutional users, courses, library resources, learning outcomes, released activities, attempts, student responses, patient conversations, and more. Each report is limited to the requesting administrator’s own institution. More information is available here.
14. Data for studying group learning. A dedicated report brings together group membership, participation, collaboration messages, and recorded control history. Group answers, patient conversations, and medical requests are also available through their corresponding reports, making shared work distinguishable from individual work.
15. More precise data requests. Report-specific filters let administrators narrow requests by relevant courses, activities, dates, students, groups, attempts, visits, and conversation channels. Institutions can request the information they need for a particular question or project.
16. Expanded report documentation and codebooks. The institutional data exports website explains each report, its organization, and its variables. More information is available here.
17. Controlled report delivery. Prepared reports arrive through an email download link that expires and can be used once. The request history lets administrators follow processing status and see when their files are ready.
18. Two-factor authentication for institutional administrators. An additional verification step at sign-in strengthens protection for institutional administrator accounts, helping safeguard institutional data and administrative tools.
19. Clinical Mind Atrium. Our community space brings together updates, research, events, and opportunities for instructors and institutional administrators. It creates another place to exchange ideas and discover ways to contribute to the project. CMind Atrium has a direct assess in the main menu.
20. The weekly Atrium newsletter. Instructors and institutional administrators receive a Monday email and platform notification when there is new content to share. The newsletter includes all the platform updates, new institutions and news from the previous week.
21. Favorite courses. Users can mark courses as favorites, making regularly used courses easier to find and return to. Just click the star and the course will be moved to Favorite courses.
22. An institutional and public library of clinical case activities. Instructors can share and adapt complete clinical case activities through the institutional and public library. This extends collaboration beyond individual patients and rubrics to the learning experiences built around them.
23. Public library resources across languages. Public library translation helps instructors use shared resources in their platform language. All assets are translated to the language of instruction of your institution, so you can use them in your activities. Teaching ideas can reach colleagues beyond the language in which they were originally created.
24. An archive for personal teaching resources. Instructors can archive patients, rubrics, and activities they are not currently using without deleting them. Resources can be restored when they are needed again.
25. Save unfinished work as a draft. Patients, rubrics, and clinical case activities can be saved before they are ready for use. Instructors can return later to complete them, while unfinished resources remain separate from materials ready for teaching.
26. Create a patient from a document. Instructors can upload a text, Word, or PDF document to help populate a simulated patient’s information, including relevant examination findings and medical tests. The instructor reviews and completes the proposed information before using the patient.
27. Patient image generation with GPT Image 2. Simulated patient image generation has been upgraded to OpenAI’s GPT Image 2. Instructors can generate an image from the patient’s characteristics and appearance, then choose the result that fits their case. With this model, the images can also represent medical conditions.
28. Dental examination and tests. Expanded options cover oral and dental examinations alongside tests such as dental radiographs and vitality tests. Instructors can build activities around findings relevant to dental care.
29. Mental status examination. Instructors can represent findings concerning appearance, behavior, speech, thought, perception, memory, and other aspects of mental status. These options support cases in mental health and related teaching contexts.
30. Functional and rehabilitation assessment. Expanded examination options cover mobility, transfers, balance, assistance needs, and other aspects of functional performance. Instructors can create encounters centered on how a patient moves and manages everyday activities.
31. Clinical scales and assessment tools. Cases can include results from clinical scales and assessment tools, including measures used in mental health and functional assessment. Students can interpret those results alongside the rest of the patient’s information.
32. Pathology images for teaching. The multimedia library now includes additional biopsy and pathology images. Instructors can use these resources to bring tissue findings into their cases and give students more opportunities to interpret visual evidence. We thank the Department of Pathology at Stanford University for providing these biopsy and pathology images to support teaching and learning in Clinical Mind AI.
33. Simpler activity design. A simplified design makes it easier for instructors to create different types of activities across the platform.
34. Return to medical history taking within an activity. Instructors can add more than one medical history section to the same activity. Students can gather initial information, work through another task, and then return to the patient with further questions.
35. Sections dedicated to the patient medical record. An activity can include an interaction focused on reviewing and completing the patient medical record. This gives documentation and record review their own place in the learning sequence.
36. Scratch-off questions. Instructors can add questions in which students reveal answer choices as they work toward the correct response. Scratch-off questions are specially helpful in group activities in which discussion is needed. This provides another format for engaging students in assessment and feedback.
37. Script concordance questions. Instructors can explore how students interpret new information when a clinical situation remains uncertain. This question format adds another way to examine reasoning beyond choosing a single conventional answer.
38. Images within assessment questions. Instructors can add images to several question formats and let students enlarge them for closer inspection. Together with click-on-image questions, these options support assessment based on visual clinical information.
39. Preview the student experience with feedback. Instructors can try an activity from the student’s point of view and receive AI feedback when it is enabled. The feedback data or performance of the instructor are not saved.
40. Preview both patient visits without waiting. Instructors can complete both visits of a delayed-results activity (activities in which the patient comes back with the medical tests results at another time) during preview without waiting for the scheduled delay. This lets them check the transition between visits and review the complete student experience before releasing the activity.
41. Rubric-based feedback with GPT-5.6 Terra. AI rubric evaluation has been updated to GPT-5.6 Terra. The model uses the instructor’s rubric to assess student work and generate feedback tied to its criteria.
42. Group clinical case activities. Instructors can now release clinical case activities for students to complete together as a group. Students collaborate on the same case, discuss clinical decisions, and submit shared work.
43. Instructor-assigned groups. Instructors can organize students into groups themselves, choosing membership to fit a teaching plan. The arrangement can then be used to release collaborative clinical case activities.
44. Randomized group formation. Students can be assigned automatically to balanced groups. This reduces the work involved in organizing collaborative activities for a large class.
45. Student-selected groups. Instructors can let students choose their own groups before a specified deadline. Students who remain unassigned can then be placed into available groups so that the activity can proceed.
46. Group assignments from a spreadsheet. Instructors can prepare group assignments in a spreadsheet and upload them to the course. Validation checks help identify membership and group-size issues before the arrangement is used.
47. Reusable group arrangements. Saved group arrangements can be used again within a course. Instructors can plan a series of collaborative activities without rebuilding the same groups each time.
48. Group learning from individual devices. Students can join the same group activity from their own devices and follow its progress together. This supports collaboration when students are working in different locations as well as in the same classroom.
49. Group learning from a shared device. A group can also complete an activity together on one device. Attendance identifies the participating students, giving instructors a practical option for in-person collaborative sessions.
50. Group chat across individual devices. When students join a group activity from their own devices, they can use the built-in chat to exchange messages and images, discuss findings, and coordinate clinical decisions while working through the same case.
51. Visibility into group participation. Instructors can review who joined an attempt, who did not participate, and who controlled different sections. Shared responses and individual participation are presented together to provide more context for reviewing group work.
52. Scheduled activity access. Instructors can set opening dates, deadlines, timers, and attempt limits when releasing activities. These settings support scheduled teaching sessions as well as activities completed independently over a longer period.
53. Grant extra attempt to one student. Instructors can grant an additional attempt to a student or group and provide a new deadline when needed. Students receive a notification explaining the additional opportunity.
54. Cost-aware medical decisions. Instructors can make medical tests costs visible during an activity if student must request medical tests using a list of options. Students can consider the cost of their requests, and the feedback report can show the costs associated with their decisions. This feature must be enabled by the institutional admin of the corresponding institution as they must upload the costs in the local currency.
55. Medical tests budgets. Instructors can set a maximum budget for medical tests if the cost visibility option is on. Students must make requests within that allowance, and spending summaries give instructors another way to discuss prioritization and resource use.
56. Two-visit clinical activities when medical tests results are provided with major delay. Delayed medical tests results can now create a return visit within the same activity. Students complete an initial encounter, wait for the configured interval, and continue when results become available.
57. Continuity at the return visit. In two-visit activities, the returning patient retains the context of the earlier conversation and requested medical tests. Students can use the returned results and patient record to continue their reasoning from the first encounter.
58. A notification for the return visit. Students receive a notification when delayed medical tests results become available. They can return to the activity and begin the second encounter when the waiting period has ended.
59. Flexible feedback across visits. Instructors can provide feedback after each visit or reserve it until both visits are complete. The final report brings the work from both encounters together.
60. Review work even when feedback is disabled. Instructors can inspect completed attempts without turning on grading or feedback. This view makes student work available for review while keeping comments and score changes unavailable for that activity.
61. Compare attempts over time. Instructors can move between a student’s attempts to see how their work develops. This supports conversations about whether students are applying feedback and changing their approach.
62. Conversation alerts for instructor review. Course Analytics provides a place to review possible inappropriate language, attempts to manipulate the patient, and other flagged conversation patterns. Instructors can inspect the messages, record a review note, and update the alert’s status; automated flags do not determine misconduct or change a student’s grade.
63. Personalized course dashboards. Instructors can customize their dashboard for each course they participate in by choosing the metrics they want to see and dragging the cards into their preferred order. Each course can have its own layout, keeping the information that matters most easy to find.
64. More ways to explore course analytics. The course analytics table connects directly to student analytics, activity analytics, and individual performance reports. Instructors can move from a course overview to the evidence behind a particular result. Just click the name of a student, the name of an activity or a specific result to go to the analytics of that item.
65. Individual student analytics. A dedicated student view brings together performance across activities, time spent, and attempts. Where assessment data are available, it also helps identify activities and rubric criteria that need attention.
66. Detailed activity analytics. Instructors can explore performance section by section, including each case within an OSCE. Response distributions, correct and incorrect totals, rubric criterion results, and patterns in physical examination and medical tests requests reveal strengths and areas needing attention. Spending summaries, when costs and budgets are enabled, and activity and conversation-time measures, including total OSCE time and averages per case, provide further insight into students’ clinical decisions and progress.
67. Course results in Excel. Instructors can download a formatted spreadsheet with course details and activity types identified by color. Scores are numeric values with maximum scores in the column headings, while activities without feedback use completion labels.
68. Analytics reports as PDFs. Student and activity analytics can be downloaded as PDFs. These reports provide a portable record for teaching discussions and follow-up.
69. A clearer route to research data. A research data button in Course Analytics explains how instructors can request course data through their institutional principal investigator. It connects them to the available reports, variable definitions, and information needed to describe a research request.
70. Friday teaching summaries (for institutions participating in a research study proposed by Stanford University). Instructors in eligible courses at institutions participating in the corresponding Stanford University study can receive weekly course summaries on Fridays. The summaries bring course information together for regular review; access and content follow the study’s assigned conditions.
71. A course analytics assistant (for institutions participating in a research study proposed by Stanford University). Eligible instructors can explore course information through questions, charts, and tables. The assistant can also help find relevant library resources and develop ideas for patients or activities within the scope of the course.
72. The patient medical record stays within reach. After the first patient interaction, students can open the patient medical record while working through later content sections. Their notes continue across sections, so information remains available when they need to answer questions or reflect on the case.
73. Notifications for new and updated feedback. Students receive email and platform notifications when feedback is released or updated. They can return to their report to review the latest assessment and comments.
74. Voice transcription with GPT-4o Transcribe. Patient conversations now use OpenAI’s GPT-4o Transcribe, with language settings informed by the simulated patient. This upgrade is intended to better capture spoken exchanges for later review and feedback.
75. More voices with GPT-Realtime 2.1. Spoken patient interactions have been upgraded to the GPT-Realtime 2.1 model family. The update supports more expressive conversations and gives simulated patients more ways to convey their emotions and behavior during an encounter.
76. Connection guidance during voice conversations. A warning lets students know when their internet connection may be unstable during a voice interaction. It suggests considering text chat so they can choose a more suitable way to continue.
77. Practice professional communication. Patients can ask students to slow down when too many questions arrive in one message, while possible inappropriate behavior can be flagged for instructor review. Repeated offensive or sexual remarks can end the conversation for that attempt, with the patient ending the conversation and saying they would make a complaint.
78. Personal performance reports. Students can revisit their attempts and download available performance reports as PDFs. These reports bring their work and available feedback together for reflection and continued practice.
79. More languages for our community. French, Swedish, and Traditional Chinese expand the platform’s language options alongside English, Spanish, Brazilian Portuguese, and Arabic. Multilingual interfaces and resources help more institutions teach and learn in their own context.
80. Mobile access for students. Students can complete activities, review feedback, and access the patient medical record from their phone’s browser. The interface adapts to smaller screens, keeping learning tools and notes within reach.
81. Direct support through “Need help?” Users can send help tickets directly to the support team from the platform, making it faster and easier to report issues and get assistance.
82. Experimental patient videos with Veo 3.1 (for institutions participating in a research study proposed by Stanford University). We are integrating Google’s Veo 3.1 into Clinical Mind AI to create short patient videos for physical examination findings or content within clinical case activities. Videos combine the simulated patient’s image with the instructor’s description of actions and dialogue. This experimental feature is being developed for institutions participating in a research study proposed by Stanford University, with instructors reviewing each video before attaching it.
These developments are part of an ongoing collaboration. As long as our community brings new questions, ideas, and possibilities, there will be more to build. Thank you to the instructors, students, institutions, and supporters who help Clinical Mind AI move forward. Your participation, shared resources, and support make it possible to keep improving how we teach, learn, and conduct research together.
