AI Role Play Simulations for Higher Education

AI role play simulations give students repeated practice in professional conversations and decision making, with faculty defined scenarios and rubric based review. This guide is for faculty and curriculum designers deciding whether a simulation fits their course. It explains what a simulation does, shows the shape of scenarios in several disciplines, walks through designing one from a learning goal, and ends with a small pilot plan.

What AI role play simulations do in a course

An AI role play simulation puts a student in a conversation with an AI character that you define. The character might be a patient, a client, a stakeholder, a witness or a parent. The student speaks or types, the character responds in role, and the exchange is recorded as a transcript. In Rocketproof, the student can respond by recorded audio, live voice or text chat, depending on how you set up the assignment.

Three things make this useful in a course.

• Repetition. Professional conversations are hard to rehearse. A lab, a standardized patient or a mock trial gives each student a few turns a term. A simulation can be attempted again, so students can try an approach, see what happened and try a different one.
• Decisions under realistic constraints. The character holds information and a disposition that you set. A student who asks only generic questions gets generic answers. A student who asks precise questions gets what a practitioner would need.
• Evidence you can review. Each attempt leaves a transcript. You score it against a rubric you wrote, and you decide what is released.

A simulation is a complement to existing practice, not a replacement for clinical placements, field work or live role play with peers. It suits the conversational and reasoning part of a skill.

Rocketproof is an authentic assessment, simulation and course intelligence platform. It is not an AI detector, it does not proctor students, and the AI does not issue final grades. Faculty judgment stays central: you design the scenario, you write the rubric and you release the result.

Example scenarios across disciplines

The examples below show the shape of a scenario. They are illustrations to adapt, not a tested scenario bank. Each names who the student talks to, what the character withholds or pushes back on, and what the student must do.

• Nursing. The student takes a history from an anxious patient who describes chest discomfort as indigestion. The character does not volunteer family history unless asked directly. The student is practicing symptom exploration and urgency. See nursing simulations.
• Business. The student negotiates contract terms with a skeptical procurement officer who objects to price and timeline. The student is practicing preparation, trade-offs and holding a position under pushback. See business negotiation simulations.
• Law and criminal justice. The student questions a witness who answers only what is asked and stays within a fixed set of facts. The student is practicing precise questions and follow-up. See witness examination drills.
• Teacher education. The student holds a conference with a parent who disagrees with an assessment decision. The student is practicing explaining a judgment clearly and respectfully. See the education solutions page.
• Emergency management. The student coordinates a response as a situation changes and information arrives in pieces. The student is practicing prioritization and communication. See the emergency management solutions page.
• Counseling. The student conducts an intake with a client who is reluctant to share. The student is practicing rapport and open questions. Check your program's rules on simulated client work before you use this in a graded setting.

More discipline detail is on the AI simulation platform page.

Designing a scenario tied to learning goals

Start from the outcome, not the character. A vivid persona with no assessable purpose produces an entertaining conversation and weak evidence.

1. Write the learning goal as something observable. "Student identifies the red flag in the history and recommends escalation" can be scored. "Student understands patient communication" cannot.
2. Choose the decision point. Name the moment where a student's choice matters, such as which question to ask, whether to concede a term, or when to escalate.
3. Set what the character knows and withholds. Decide which facts the character reveals freely and which only appear when the student asks well. This is what separates a prepared student from an unprepared one.
4. Set the disposition. Anxious, skeptical, rushed or evasive. Keep it realistic for the setting and kind to the student. The goal is practice, not a trap.
5. Define the boundaries. State what the character will not do, such as invent facts outside the case or break role to coach the student.
6. Write the rubric before you publish. Decide what strong, adequate and weak look like for each criterion, then check that the scenario gives students a chance to show each one.

What makes a strong scenario prompt

A strong prompt is specific about the role, the setting, the hidden information and the goal. Compare these two.

• Weak: "Pretend to be a patient with chest pain."
• Stronger: "You are Arthur, a 62-year-old retired teacher. You have had chest tightness for two days and believe it is indigestion. You are nervous about hospitals and answer briefly. Share your brother's heart attack only if the student asks about family history. Stay in role and do not offer a diagnosis."

The stronger prompt gives the character a reason to hold back, ties disclosure to the student's skill, and sets a boundary. Rocketproof's Assignment Creator can draft the scenario context, persona instructions and rubric criteria from a plain-language description. Treat the draft as a starting point, then edit it against your learning goal before students see it.

Using adaptive questioning during simulation

Adaptive questioning means the character's next turn depends on what the student actually said. If the student asks an open, well-aimed question, the character gives more. If the student skips an important area, the character does not rescue them. Follow-up questions come from the student's own answers rather than from a fixed script.

There are two ways this shows up in Rocketproof.

• In a role play scenario, the character responds in role to each student turn, within the knowledge and boundaries you set.
• In an artifact-based assignment, Rocketproof reads a submitted paper, case analysis or similar document and prepares a questioning plan tied to its specific arguments, methods and sources. You can review the plan and add your own questions before students begin. The student's answers then drive the follow-ups.

Keep the anchor decision points the same for every student so attempts are comparable, and let the follow-ups branch. For more on grounding questions in student work, see AI oral exams grounded in student work.

Reviewing performance with instructor rubrics

The rubric is where faculty judgment stays central. You write the criteria and level descriptors, and students can see them in advance.

A starting set of criteria for a conversation-based scenario might be:

1. Information gathering. Did the student ask questions that surfaced the key facts?
2. Reasoning and decision. Did the student reach a defensible decision and explain why?
3. Response to pushback. Did the student adapt when the character objected or withheld?
4. Communication. Was the student clear, respectful and appropriate for the setting?

The review workflow is as follows.

• Rocketproof prepares draft rubric scores with evidence cited from the transcript.
• You read the transcript, listen to the recording where one exists, and check each draft score against the evidence.
• You change any score you disagree with and add your own feedback.
• Nothing is final until you release it. Where LTI grade passback is configured, a released grade can pass to the LMS gradebook.

Keep delivery out of the score unless speaking is a stated outcome. Accent, pace and hesitation are not evidence of reasoning, and a student who types gives you no delivery to observe. Practice attempts can be set up separately from graded ones, so students can rehearse without gradebook impact. A concern raised by a transcript follows your institution's normal process. It is one piece of evidence, not a finding. For the wider picture on assessment design, see alternatives to AI detection for college assessments.

Planning a small pilot

Start with one course and one scenario.

1. Pick one learning goal that is hard to assess with a written test.
2. Draft the scenario from the design steps above, and have a colleague read the prompt for realism.
3. Write and publish the rubric before students begin.
4. Choose the response format. Recorded audio, live voice or text chat, depending on whether real-time speech is part of the outcome.
5. Run it yourself first, then with a few volunteers. Note where the character confuses students for the wrong reasons.
6. Offer a practice round that does not count toward the grade.
7. Set the accommodation path, including typed responses where speaking is not an outcome, and involve your disability services office.
8. Time your review on a handful of sessions so you can plan workload before you scale.
9. Collect student feedback and revise the scenario and rubric before adding more scenarios.

Check your institution's policies with the relevant offices before a graded pilot. Canvas is the LMS platform verified end to end. Other LMS platforms use the same LTI 1.3 standard, but a launch has not yet been observed on a vendor test system, so check with us before planning a term around them. See the LMS integration page for current status.

Frequently Asked Questions

What is an AI role play simulation for higher education?

It is a structured conversation between a student and an AI character that a faculty member has defined. The character plays a role such as a patient, client or stakeholder and responds to what the student says. The student practices a professional conversation or decision, and the instructor reviews the transcript against a rubric they wrote.

Which courses benefit most from scenario simulations?

Courses where the skill is a conversation or a judgment made under uncertainty tend to benefit most, such as nursing and other health professions, business and negotiation, law, teacher education, counseling and emergency management. Courses where students must explain and defend decisions can also use them. Where the outcome is purely procedural or physical, a simulation is a poor fit.

How are simulations evaluated?

You write the rubric and publish it in advance. Rocketproof prepares draft rubric scores with evidence from the transcript. You review them, change any you disagree with, and release the result. The AI does not issue final grades.

How long should a simulation session be?

There is no single right length, and we do not publish a recommended figure. Set it by how much evidence you need for the learning goal, then keep it short enough that your review time stays manageable. Try a short session in a pilot, read the transcripts, and extend only if the rubric criteria are not getting enough evidence.

What makes a strong scenario prompt?

A strong prompt names the role, the setting and the goal. It states what the character knows, what they withhold until asked well, and how they behave. It also sets boundaries, such as staying in role and not inventing facts outside the case. Tie it to an observable learning goal, then edit any AI-drafted version against that goal before students see it.

Explore scenario ideas for your discipline

Explore scenario ideas for your discipline, try the public demo as a student or instructor, or request a demo to talk through a pilot in your course.