Ask any experienced mentor how they work and they will describe the same rhythm: watch the learner act, step in when something goes wrong, explain what they saw, and let the learner try again. Almost no digital learning works this way. Feedback is saved up and delivered as an end-of-module score, or deferred to a debrief after a branching storyline has run its course, long after the learner can remember why they chose what they chose.
Decision-point mentoring moves the coaching back to where a mentor would put it: the decision itself. This guide covers what the method is, why timing carries so much of its effect, how it differs from consequences-only simulation and end-of-module feedback, what good mentoring content contains, how mentoring events double as measurement, and how instructional designers write mentoring from expert rationales.
What happens at a mentored decision point?
In a scenario built around decision-point mentoring, the learner works through a realistic situation, a patient presenting with a complicated history, a customer pushing back on price, a direct report disputing feedback, until they reach a decision that matters. They commit to a choice from a set an expert has defined. The choices are not a quiz's right-or-wrong pair. They reflect how the situation actually presents: some options are optimal, some are suboptimal, some are inappropriate, and there is often more than one optimal way to proceed.
What happens next depends on the choice. If the learner selects an optimal option, the scenario acknowledges it, reinforces why it is optimal, and moves forward. If they select a suboptimal or inappropriate one, expert coaching arrives immediately, addressed to that specific choice in that specific situation. The learner takes in the rationale, reconsiders, and chooses again, continuing until they recognize the optimal approaches. Decide, receive coaching, revise: that loop is the core mechanic.
Three properties of the loop are worth making explicit. First, the coaching is specific. It is not a generic hint or a restatement of the content; it explains why this option falls short here, in this context. Second, the learner always finishes the decision on an optimal path. Selecting a weak option triggers correction, not punishment, and the learner leaves the decision having found what great looks like rather than merely having been told they missed it. Third, the storyline is not derailed. The scenario continues along a single expert-designed spine, and the correction happens in place, inside the situation where the reasoning applies.
Why does the timing of the feedback matter?
The case for timing starts with what actually develops expertise. Ericsson's research on expert performance found that the traditional markers, length of experience and reputation, bear only a weak relationship to observed performance. What does predict it is deliberate practice: training focused on particular tasks, often designed by teachers and coaches, with immediate feedback, time for problem solving and evaluation, and opportunities for repeated performance to refine behavior (Ericsson, 2008). Decision-point mentoring is that structure applied to judgment. The focused task is the decision, the immediate feedback is the mentoring, and the retry is the opportunity to refine.
Cognitive apprenticeship makes the same point from the teaching side. In Collins, Brown, and Holum's framework, coaching is not a synonym for feedback delivered at the end; its content is "immediately related to specific events or problems that arise as the student attempts to accomplish the target task" (Collins, Brown & Holum, 1991).
The mechanism behind both is that at the moment of choosing, the learner's reasoning is still inspectable. They know what they noticed, what they weighed, and what they assumed. Coaching that arrives then can attach directly to that reasoning and revise it: the learner can compare the expert's account of the situation against their own, see where the two diverge, and act on the difference immediately. Feedback deferred to the end of the module faces a reconstruction problem. The learner no longer holds the reasoning behind choice four of twelve, so the feedback can only land on the outcome, a score, rather than on the thinking that produced it. And without a retry, even a well-explained correction goes unrehearsed: the learner is told what to revise but never revises it.
How is this different from consequences-only simulation?
Traditional branching simulations follow a different logic: the learner chooses, the simulated world responds, and the learner rides the consequences down a path, often to a debrief at the end. Consequences are valuable, and guided scenarios use them too. The difference is what stands in for the teacher, and the branching approach has two problems, one pedagogical and one structural.
The pedagogical problem is that consequences alone reproduce experience, and experience alone is precisely what Ericsson found to be an unreliable developer of expertise. A learner can ride out a bad consequence without ever learning why the choice was weak or what an expert would have noticed instead. The lesson is left implicit at exactly the moment it should be explicit.
The structural problem is what all those paths do to the feedback. As branches multiply, the learner meanders through a tree of consequences, and by the end of a path it is genuinely hard to reconstruct where the run went wrong, at which decisions, and which choices were sound. The feedback arrives far from the decisions that earned it.
Decision-point mentoring resolves this with a non-branching structure. The scenario holds one expert-designed storyline, and correction happens in place: sometimes as an immediate coached exchange, sometimes after the choice plays out and a character working alongside the learner responds to what it produced. Either way, the feedback lands at the decision that earned it, and the story continues.
End-of-module feedback shares the pedagogical problem in a different form. A score summarizes what happened but cannot revisit the reasoning behind any single decision, and it arrives when nothing can be done about it. Repetition is technically available: a learner could replay a branching simulation until the optimal paths reveal themselves. But that takes more time than almost anyone invests, and most learners never run it twice.
What does good mentoring content contain?
Delivering feedback at the decision is necessary but not sufficient; the content has to earn the moment. Published work on corrective mentoring in continuing medical education proposes three criteria (Seifert, 2021). Feedback must be authentic and conversational, because most real decisions are nuanced, with caveats, controversy, and points of view, and bare text responses read as assessment: "wrong," followed by a rationale. Feedback should be fluid and in context, elucidating the rationale in the specific situation where the learner chose a less optimal option, not retrieved as generic advice. And it should be a one-on-one, personal experience, so the learner can build on their own existing mental model rather than absorbing a broadcast.
In practical terms, a good mentoring passage does three jobs:
- It explains why the choice falters here. Not why it is wrong in general, but which features of this situation make it suboptimal: the history that changes the calculus, the constraint the learner discounted, the risk that this option quietly accepts.
- It reveals what an expert would notice. The cues an experienced practitioner attends to first, and how those cues reorder the options. This is the "making thinking visible" work of cognitive apprenticeship, delivered in a moment small enough to absorb.
- It encourages without judgment. Choosing a suboptimal option inside a scenario is the learning working as designed, and the mentoring should treat it as a normal move in a professional conversation, not a failure event. Learners who feel graded start optimizing for the grader; learners who feel coached keep reasoning.
Delivery matters as much as substance. The coaching lands best through character dialogue rather than system messages: a peer or mentor character inside the scenario responds the way a colleague would, weighing the choice from within the situation. A system message judges from outside and breaks the frame. A character keeps the learner inside the situation while they reconsider.
How do mentoring events double as measurement?
Every mentoring event is also a record: this learner, in this situation, at this decision, needed coaching before recognizing an optimal option. Aggregated across learners, those records become a map of exactly where the gaps between current decision-making and expert decision-making sit, which no completion rate or quiz average can draw.
Published program data shows what the map looks like. In a continuing medical education activity reported in the ACEhp Almanac, initial competence climbed from 36% to 89% across three comparable clinical situations while the need for corrective mentoring fell sharply (Seifert, 2021). The declining need for mentoring is itself the evidence that skill is developing: the mentoring that helps the learner also records exactly where help was needed.
That measurement dimension, where decision-level data sits in outcomes frameworks and how to report it, is a subject of its own, treated in our guide to measuring behavior change. The point here is the design implication: an instructional designer who builds mentoring into every decision gets the analytics without adding a single assessment item.
How do instructional designers write mentoring content?
Mentoring content is written from expert rationales, and the central difficulty is that those rationales are invisible by default. Experienced practitioners carry mental models that guide their decisions, but the models do not announce themselves; they have to be deliberately extracted (Seifert, 2021). At the level of principles, the work runs in four steps.
- Start from the gap. Identify the situations where practitioners' decisions diverge from what top performers do, the places where knowing has not become doing. These situations, not the content outline, are the raw material for scenarios.
- Define the decision and its options. Ask experts which options they would actually consider in the situation, and which are optimal here versus in neighboring situations. Include the plausible suboptimal and inappropriate choices learners genuinely make; a decision point with only defensible options teaches nothing.
- Capture the rationale for every non-optimal option. For each one, ask what makes it tempting, what it misses in this specific situation, and what an expert notices that reorders the choice. The most productive interview question is some version of: what do the best people do differently here, and how do they explain it?
- Voice the rationale as conversation. Transform each rationale into the dialogue a mentor character would actually say, applying the three content criteria above: specific to the situation, revealing of expert perception, encouraging in register.
Then vary the situations. Designers are routinely asked, "Why do another scenario, we covered that?" The published data above is the answer: initial competence climbed from 36% to 89% across three situations, not one. Recognizing an expert's decision pattern in a single context is a start; recognizing it across varied contexts is the skill.
This is the mechanism AliveSim's Guided Scenarios are built around: each decision point holds the learner's own choice open long enough for a mentor character to respond to the specific reasoning behind it, then returns the learner to the same decision to try again while that reasoning is still inspectable (AliveSim). But the method predates any platform. It is what a good mentor has always done, made repeatable: correction at the decision, in the voice of a colleague, with the chance to try again.
References
- Collins, A., Brown, J. S., & Holum, A. (1991). Cognitive apprenticeship: Making thinking visible. American Educator, 15(3), 6–11, 38–46.
- Ericsson, K. A. (2008). Deliberate practice and acquisition of expert performance: A general overview. Academic Emergency Medicine, 15(11), 988–994.
- Seifert, D. (2021). Incorporating skill development in CME via corrective mentoring. Alliance for Continuing Education in the Health Professions Almanac.
Related questions
What is just-in-time feedback in learning?
Just-in-time feedback is guidance delivered at the moment it is relevant to what the learner is doing, rather than batched into a later review. Decision-point mentoring is a specific, structured form of it: the learner commits to a decision inside a realistic scenario, and expert coaching addresses that particular choice in that particular situation before the learner moves on. The rationale comes from cognitive apprenticeship, where coaching is defined as feedback immediately related to the specific events or problems that arise as the learner attempts the task, and from deliberate practice research, which pairs immediate feedback with the opportunity to refine performance.
What is the difference between immediate and delayed feedback in training?
Immediate feedback arrives while the learner is still inside the task; delayed feedback arrives after the task, often as a score or debrief. For complex decisions, the practical difference is what the feedback can attach to. At the moment of choice, the learner still knows what they noticed, weighed, and assumed, so coaching can address the reasoning itself and the learner can revise it on the spot. Ericsson's account of deliberate practice pairs immediate feedback with opportunities for repeated performance to refine behavior, which is exactly the pairing a retry at the decision point provides and an end-of-module summary does not.
What is corrective mentoring?
Corrective mentoring is the term used in published continuing medical education work (Seifert, 2021) for the expert feedback a learner receives after selecting a suboptimal or inappropriate option at a decision point, before recognizing an optimal one. It distinguishes decisions that were initially competent, meaning the learner chose an optimal option on the first attempt, from decisions that became competent only after mentoring. That distinction gives the same event two functions: the learner gets the expert rationale in context and revises their choice, and the program records exactly where coaching was needed.
Why should scenario feedback come from a character instead of a system message?
Because the register changes what the learner does with it. A system message grades from outside the situation, which reads as assessment: wrong, followed by a rationale. A peer or mentor character responds from inside the situation, the way a colleague would, with the nuance, caveats, and points of view that real decisions carry. Published CME work on corrective mentoring argues feedback should be authentic, conversational, and delivered in context for exactly this reason. Dialogue also preserves immersion, so the learner stays in the situation while reconsidering rather than stepping out of it to read a verdict.
Published March 4, 2025 · Updated July 15, 2026 · 10 min read