Every scenario has a cast, and most casts are assembled by instinct: someone to talk to, someone who knows things, someone the story is about. Treated as background decoration, they get filled in and forgotten. There is a sharper rule, and it changes what you build: cast each character for the job it does for the learner. A character that does nothing for the learner gets cut.
Three roles carry the work. A fourth is the one to recognize and avoid.
The peer
Exists to be helped, and to err visibly
The learner works for the peer’s success, and the peer can voice a tempting mistake so it is examined in the open, without the learner having to make it themselves.
The mentor
Explains the nuanced why
Works the gray where choices are not black and white, why one option beats another here but not there, and pushes with constructive friction instead of smoothing everything over.
The subject
Reacts in the conversation
The customer stiffens, the patient hesitates. The outcome of a choice lands as a human response, not narrator text.
The decision point
The learner sits here, as advisor
The peer asks, the learner decides, the subject responds, the mentor coaches. Every role earns its place by doing a job for the learner.
The presenter · anti-pattern
Lectures at the learner
A character stands in front of the learner and delivers content. It is not a conversation, it is a lecture with a face on it, and it does no job for the learner the other three do not do better.
The peer: the conversation enabler
The peer is the character the learner is helping, and it is the reason a conversation exists at all. The peer converses with the subject and with the mentor, all of it directed by the learner, who takes part and helps. Without the peer there is nothing for the learner to be inside.
What the learner gets to watch this way is real conversations unfolding, including mentoring that unfolds as a conversation rather than a lecture. That is a huge difference. The guidance arrives in the give-and-take of people talking, at the moment it is needed, instead of being handed down in a block the learner is expected to absorb and apply later.
The relationship also recruits effort, and this is where the research is sharpest.
The mentor: the nuanced why
The mentor is the more experienced voice in the room, and its job is nuance. It provides the understanding of why a certain strategy or option may be suboptimal. It explains the gray, the places where choices are not black and white, and why you might go one way in this situation and another way in that one. This is corrective mentoring: not a verdict stamped on the learner's choice, but a person in the conversation explaining the reasoning a novice cannot yet see.
The subject: shows the real conversation
The subject is who the situation is about: the customer, the patient, the direct report. It does two jobs. First, the outcome of a decision lands in the conversation. Choose well or badly and the subject is where it shows up first: the customer stiffens, the patient hesitates, the employee goes quiet, in a response rather than on a results screen. Second, the subject is there to show what a real conversation looks like, how it feels, and how it evolves, so the learner reads a decision the way they will read it on the job.
This is a design position more than an experimental finding, but it rests on a simple point the evidence keeps making: a situation that answers you through a person is more engaging, and easier to read, than one that reports back in prose.
Talking straight to the learner
Any of these characters can turn and speak directly to the learner, and that is ordinary, not a special event. This is not a play with a fixed wall between stage and seats: characters address the learner freely. One version of it is the aside, a line only the learner hears, which the fourth-wall guide takes up in full. The aside is one way to involve the learner, not the only one.
The presenter: the role not to build
There is a fourth role, and naming it plainly is the point of this guide. The presenter is a character who stands in front of the learner and delivers content. No decision, no exchange, no one to help. It is by a wide margin the most common way characters get used, and it is the least effective of the four, because it is essentially a lecture with a face on it.
So the presenter is what to replace, not what to place. The replacement is an interaction: a peer with a mentor, or a subject with a peer, with the learner drawn into it through a decision or through a character speaking directly to them. There is one fair concession. Sometimes content genuinely has to be presented, with no decision to build around it. Even then, do not build a single presenter. Put two characters in a podcast-style pair. The two go back and forth, ask each other questions, and bring in their own experience with the material, what they learned, how they used it, how it felt, so the knowledge comes to life experientially instead of being recited, and the learner is drawn into that conversation as much as possible. That beats one character lecturing, for straight knowledge transfer as much as for learning to apply it.
Four roles, one scene
Casting for the job each character does is easiest to see when the roles work together around a single decision. Keep the beat compact.
A new manager, the peer, brings the problem: "One of my people has been talking over colleagues in every meeting, and it is getting worse. I'm thinking I pull them aside and just tell them to cut it out. What do you think?" The learner advises. The employee, the subject, responds in the conversation to what the learner chose, opening up or getting defensive, and the learner watches how it actually goes. The mentor then supplies the nuanced why: when a blunt correction works, when it backfires, and what this situation in front of them calls for. Four characters, four jobs, one choice. Notice who is missing: no one is presenting. The learning happens because the learner is deciding for a peer, watching a person respond, and getting the nuance from a mentor, which is exactly the seat the advisor-seat guide argues for.
AliveSim is a conversation engine, and that is what lets a designer build every one of these roles as 3D avatar characters: the peer, the mentor, the subjects, even two presenters in podcast form to convey knowledge alongside decision practice. Characters can speak to each other, to the learner, or privately to the learner as an aside, and the learner directs the whole thing through decisions the characters react to. AliveSim's Guided Scenarios are built this way, so the scene is an interaction rather than a delivery. For how it fits together, see the AliveSim platform.
The takeaway
Cast each character for the job it does for the learner. The peer makes the conversation exist and recruits real effort, because the learner is helping someone. The mentor supplies the nuanced why, the gray between good and bad options. The subject reacts, so the outcome lands in the conversation and the learner sees what a real one looks like. Each earns its place by doing something for the learner, and a character that does none of those things is presenting. The presenter feels like the safe default, but it is the monologue that every dialogue in the research outperforms. Build the conversation instead, and if you must present, let two voices do it.
References
- Chase, C. C., Chin, D. B., Oppezzo, M. A., & Schwartz, D. L. (2009). Teachable agents and the protégé effect: Increasing the effort towards learning. Journal of Science Education and Technology, 18(4), 334–352.
- Chi, M. T. H., Roy, M., & Hausmann, R. G. M. (2008). Observing tutorial dialogues collaboratively: Insights about human tutoring effectiveness from vicarious learning. Cognitive Science, 32(2), 301–341.
- Craig, S. D., Gholson, B., Ventura, M., & Graesser, A. C. (2000). Overhearing dialogues and monologues in virtual tutoring sessions: Effects on questioning and vicarious learning. International Journal of Artificial Intelligence in Education, 11, 242–253.
- Craig, S. D., Sullins, J., Witherspoon, A., & Gholson, B. (2006). The deep-level-reasoning-question effect: The role of dialogue and deep-level-reasoning questions during vicarious learning. Cognition and Instruction, 24(4), 565–591.
- Muller, D. A., Bewes, J., Sharma, M. D., & Reimann, P. (2008). Saying the wrong thing: improving learning with multimedia by including misconceptions. Journal of Computer Assisted Learning, 24(2), 144–155.
- Stalmeijer, R. E., Dolmans, D. H. J. M., Wolfhagen, I. H. A. P., & Scherpbier, A. J. J. A. (2009). Cognitive apprenticeship in clinical practice: can it stimulate learning in the opinion of students? Advances in Health Sciences Education, 14(4), 535–546.
Related questions
What does it mean to cast a character for the job it does for the learner?
It means every character in the scene has to earn its place by doing something for the learner: giving them someone to help and steer, explaining why a choice is strong or weak, or reacting so the learner sees the result of a decision. A character that does none of those is background decoration, and background decoration gets cut. The rule keeps a cast small and working: three roles do almost all of it, the peer, the mentor, and the subject, and a fourth, the presenter, is the one to recognize and leave out.
What is the peer's job, and why does the scenario need one?
The peer is the character the learner is helping. Without it there is no conversation for the learner to be inside: the peer converses with the subject and with the mentor, all of it directed by the learner, who takes part and helps. That relationship also recruits effort. In the studies that named the protégé effect, students who believed they were helping a computer character learn spent more time on learning activities and learned more than students doing the same work only for themselves, with the biggest gains among lower achievers (Chase, Chin, Oppezzo & Schwartz, 2009). Helping someone is more motivating than being measured.
What is the mentor for, and what is corrective mentoring?
The mentor is the more experienced voice, and its job is nuance: it explains why a certain strategy or option may be suboptimal, and it works the gray, the places where choices are not black and white, why you might go one way in this situation and another way in that one. Corrective mentoring means the guidance arrives as a person in the conversation explaining the reasoning a novice cannot yet see, not as a verdict stamped on the choice. The evidence favors this move. Naming a common wrong idea and correcting it inside a dialogue produced some of the largest gains in the multimedia research (Muller, Bewes, Sharma & Reimann, 2008), and learners say they value a mentor who pushes, what they called constructive friction (Stalmeijer, Dolmans, Wolfhagen & Scherpbier, 2009).
Why is a presenter the wrong default, and what should replace it?
A presenter is a character who stands in front of the learner and delivers content, with no decision and no exchange. It is the most common way characters get used and the least effective, because the research consistently favors dialogue over monologue: observers of a tutoring dialogue recalled more and asked deeper questions than observers of the same content as a monologue (Craig, Gholson, Ventura & Graesser, 2000), a deep-reasoning question before each point was the active ingredient (Craig, Sullins, Witherspoon & Gholson, 2006), and pairs observing a recorded dialogue learned as much as the students actually being tutored (Chi, Roy & Hausmann, 2008). Replace the presenter with an interaction: a peer with a mentor, or a subject with a peer, and the learner involved through a decision or through a character speaking directly to them. When presenting is genuinely required, two characters in a podcast-style pair, going back and forth, asking each other questions, and bringing in their own experience with the material, beat one presenter lecturing, for knowledge transfer as much as for learning to apply it.
Published July 18, 2026 · 9 min read