Count the real attempts a student gets at the conversation that will define their career.
A nursing student might deliver bad news to a standardized patient twice before graduating. A counseling student sees their first real client with a handful of supervised practicum sessions behind them. A business student walks into their first salary negotiation having negotiated exactly zero times. A pre-service teacher faces their first furious parent with no prior reps at all, because the teaching program had no way to manufacture a furious parent on demand.
Meanwhile, the same student has read hundreds of pages about active listening, difficult conversations, and negotiation theory. The content is abundant. The rehearsal is scarce. That asymmetry is the rehearsal gap, and it is the single most fixable problem in how universities teach interpersonal skill.
Skill lives in repetitions
The pedagogy literature has been clear about this for decades. Bonwell and Eison’s classic work on active learning described role play as a way to put students inside situations that are “stressful, unfamiliar, complex, or controversial,” precisely because examining a situation from the inside builds something reading cannot. Harvard’s teaching resources list the same benefits: perspective-taking, applying knowledge under pressure, confronting views you disagree with. Duke Kunshan’s teaching center adds safe experimentation and mistake-based learning to the list.
None of this is controversial. Every center for teaching and learning recommends role play. The problem has never been whether role play works. The problem is dosage.
A well-run role play in a 70-person class gives each student one turn, maybe two, per semester. The students not currently performing are observing, which has value, but observation is not the thing that builds skill. Being in the seat is. Responding in real time, feeling the conversation turn on you, recovering from a bad opening: these are motor skills of speech, and motor skills are built through volume.
Think about any domain where performance under pressure is taken seriously. A pianist does not perform a piece after reading about it twice. A litigator moots the argument dozens of times before standing up in court. Medical education figured this out earliest: the entire standardized patient industry exists because medical schools accepted that you cannot learn patient communication from a textbook, and you should not learn it on real patients.
The rest of the university has largely not made this leap. We assess essays and exams, which scale, and we treat conversational competence as something students will pick up “in the real world.” The real world, as any career counselor will tell you, is an expensive place to learn. The student who fumbles their first five real interviews does not get those companies back. The new teacher who mishandles their first parent confrontation starts the year with an adversary. Real stakes make real situations the worst possible practice environment, exactly when the learner is worst equipped.
Why the gap persists
If everyone agrees rehearsal matters, why do students get so little of it? Because until recently, every unit of rehearsal required a human counterpart, and human counterparts are the most constrained resource in a university.
Consider the options a professor has had:
Peer role play. Free, but the counterpart is another novice. Two students who both do not know what a hard negotiation feels like cannot give each other one. Peers are also socially entangled: students soften the pushback because they will sit next to each other on Thursday.
The professor as counterpart. High quality, catastrophically unscalable. A 20-minute role play per student in a class of 70 is more than 23 hours of the professor’s time, per exercise, per semester.
Trained actors and standardized patients. The gold standard for fidelity, and priced like one. Programs that use them reserve them for a small number of high-stakes assessments, which means they serve as exams, not practice. Nobody develops a skill in the exam.
Avatar simulations with a human in the loop. Platforms like Mursion and TeachLivE put a trained simulation specialist behind a 3D avatar. The realism is excellent and the research base in teacher education is real. But each session still consumes an hour of a trained human’s time; published estimates put a single session at $49 to $164 per learner. That is a field trip, not a practice habit.
Every one of these options couples rehearsal to synchronous human labor. That coupling is why the rehearsal gap has survived fifty years of teaching centers recommending role play. The recommendation was right; the economics were impossible.
What changes now
AI role play decouples rehearsal from human labor. A conversational agent can play the skeptical interviewer, the grieving family member, the defiant student, or the examiner, at any hour, for any number of students simultaneously, at a marginal cost measured in cents. When NYU’s Stern School of Business ran AI-conducted oral exams for an entire class, 36 sessions of roughly 25 minutes each, the total bill was about $15.
This does not make actors, avatars, or professors obsolete; Chapter 3 is an honest map of where each modality still wins. What it changes is the default. Rehearsal used to be the scarce thing you rationed. Now it can be the abundant thing you assign, and the scarce human expertise moves to where it actually compounds: designing the scenario, and debriefing what happened.
The rest of this playbook is about doing that well: choosing the right modality for each activity, mapping the use cases across a university, designing counterparts that push back instead of flattering, grading with rubrics rather than vibes, and rolling the whole thing out without triggering the failure modes that kill new teaching technology. Each chapter ends with an exercise you can complete in under 30 minutes, because a playbook about practice should make you practice.
Exercise
Take the course you teach (or support) that has the highest-stakes conversation embedded in it: the counseling session, the client pitch, the patient interview, the parent meeting, the thesis defense. Write down two numbers. First: how many live, in-the-seat attempts at that conversation does a typical student get before they face it for real? Second: how many attempts do you believe would make them genuinely competent? The distance between those numbers is your rehearsal gap. Keep it; every later exercise builds on this course.