Simulations & Immersive Virtual Environments

Principle · All principles
Evidence · 101 claims (38 for, 55 mixed, 8 against) · 71 studies (22 causal, 17 review, 11 quant-synthesis, 9 qualitative, 7 theoretical, 4 design, 1 associational), q1–q4 · 19 of 71 report an effect size · 96 claims rest on one study

Conditional relationship

This page owns practising inside a designed representation of a real task: a screen simulation, a virtual lab, an immersive VR or AR environment, a mannequin, an AI-driven virtual patient or a simulated conversation. For a learner preparing for a performance that is costly, risky, rare or invisible in the real setting (a clinical procedure, a hazardous lab, a safety drill, a difficult conversation, a system whose dynamics cannot be seen), repeated attempts in a simulation that represents the task's decisions and cues are expected to improve performance on that task (skill-performance, immediate-performance, near-transfer) more than no simulated practice or passive instruction (not, on present evidence, more than active practice of another kind), when three conditions hold: learners are guided while they explore (a goal for each run, prompts or a facilitator who responds to what they do); each run is followed by feedback or a debrief that connects what happened to the real task; and the simulation keeps the parts of the task the learner must attend to, not only its look. More immersion or visual realism is not expected to add learning by itself.

How far this is tested. Part of it is tested; the whole relationship is not. A meta-analysis of 609 studies in the health professions found technology-enhanced simulation training beat no intervention on knowledge, skills and behaviours, with large pooled effects, but it does not compare simulation with equally long non-simulated practice, and its subgroup analyses could not single out feedback, repetition or curricular integration as the active feature (Simulation Based Education Improves Outcomes [+M]). A meta-analysis of 65 studies of simulation games, reported second-hand with no effect size printed, found higher skill-based knowledge (the practical application of knowledge) in simulation conditions than in non-game conditions, which is a comparison with instruction rather than with nothing, though the claim does not say what that instruction was (The review reports that participants in simulation-game conditions had significantly higher skill-based knowledge outcomes than those in non-game conditions [+M]). The same meta-analysis, read directly, splits that comparison by what the comparison group did: simulation games beat no substitute instruction (d = .61) and passive instruction (d = .38), but learners taught with active methods learned slightly more than the simulation-game group (d = −.19) (The advantage of simulation games shrinks or reverses when the comparison group learns via active instructional methods [~M]); and its published studies averaged d = .52 against d = −.10 in unpublished ones (Published studies of simulation games report substantially larger effects than unpublished studies, indicating publication bias [~M]). So the expectation here is against no simulated practice or passive instruction, not against active practice of another kind, and pooled simulation effects are likely inflated. One experiment found that VR with guided exploration produced much larger gains than either the same VR explored freely or no VR, and that free exploration in VR was no better than no VR (Assimilator learners gained significantly more in the VR (guided exploration) mode [+M]; VR (non-guided exploration) did not differ significantly from Non VR [~M]). No claim in this wiki compares a simulation with and without a debrief, varies the number of runs, or compares levels of fidelity directly, so those parts of the design below are proposals.

Virtual, remote and online laboratories are one family of simulations with questions of their own (can one stand in for a hands-on lab, for which outcomes, and who guides a learner working alone); they are treated in a marked subsection at the end of the default design, with their own rows in the situation table and their own evidence entries.

The relationship is conditional on the learner's expertise on the task (novice learners need staged complexity and support during the run), on the kind of skill (gains are reported for communication confidence, structured knowledge and procedural skill; tactile psychomotor skills practised on a screen did worse than practice with a standardized patient), on the horizon (near transfer and immediate tests are where gains are reported; delayed and far transfer are weaker), and on access (devices, motion sickness, teacher preparation time).

How the converted siblings sit beside this one. Experiential Learning owns the general relationship (an experience becomes learning when reflection, a concept and a further attempt follow); a simulation is one kind of experience. Debriefing owns the facilitated conversation after the run and its four moves; this page uses it and does not repeat it. Deliberate Practice owns targeted refinement against a standard once a learner can perform roughly; simulation-based mastery training is one of its settings. Game-Based Learning owns learning through a game whose winning move is the target reasoning; simulation games sit between the two pages. Guided Practice and Scaffolding and Fading own support during attempts and its withdrawal. Digital Learning owns the argument that a medium is not a method, which applies here to VR and AI: what is expected to matter is the practice, guidance and feedback the simulation carries.

Default design

The earlier page's guidance, made concrete. Simulation against no simulation and guided against free exploration have claim support; every dose, sequence and threshold below is a proposal, labelled as such, to be set locally and recorded.

1. Choose a task worth simulating, and name the decisions and cues it must keep. Use a simulation when the real performance is risky, costly, rare, slow or invisible; write down the three to five decisions or cues a learner must attend to, and check the simulation represents those, whatever its look. Evidence: simulation practice beat no intervention in the health professions (Simulation Based Education Improves Outcomes [+M]); higher perceptual fidelity does not reliably improve learning and can add extraneous load (Increased perceptual fidelity fails to reliably improve serious game learning outcomes [~M]). Make those cues easy to notice (highlight them, slow events down, remove distractors) instead of adding visual detail: an editorial reading the higher-education simulation meta-analyses reports that how closely the simulation's functions match the task, together with how noticeable the relevant information is, explains more of the variation in outcomes than physical resemblance does (Functional correspondence predicts simulation learning outcomes better than physical resemblance or appeals to authenticity [+M]; a second-hand reading of Chernikova et al., not an independent test, no effect size printed). The list of decisions and cues is an untested proposal.
2. Prime the parts and terms before the first run (5–10 minutes). For learners new to the system or tool, give a short primer naming its key parts and what each does before entering the simulation. Evidence: in one randomized VR experiment, a pretraining video on the parts of a micropipette produced better knowledge scores and fewer errors on a real-world transfer task than the same VR lesson without it (Pretraining learners on a system's key parts and terms before the main lesson improved knowledge and real-world transfer in one VR experiment [+M]; its table counts 4 distinct studies). The length is an untested proposal.
3. Give each run a goal and guide the exploration. Before each run, state what the learner is trying to do and what to watch for; during it, use built-in prompts or a facilitator who responds to what the learner actually does (a cue, then a more specific hint, then a demonstration). Do not leave novices to explore freely. Evidence: guided VR exploration beat free VR exploration and no VR for both learning-style groups in one experiment (Accommodator learners gained significantly more in the VR (guided exploration) mode [+M]); contingent support beat fixed or absent support, mostly in one-to-one tutoring (Contingent scaffolding improves learning more than fixed or absent support. [+M], carried from tutoring; its table counts 20 distinct studies). In a head-mounted display, do not assume spoken prompts beat printed ones: in one meta-analysis of the modality effect, studies using VR headsets showed larger effects of narration over on-screen text (d = 1.34) than studies using computer screens (d = 0.55) (In one meta-analysis, modality-effect studies using virtual reality head-mounted displays show larger effects than studies using computer screens or audio-tapes [~M]; a moderator comparison across studies, not a test inside simulations), but a later VR experiment with 61 learners found short printed text beat narration on recall (In immersive VR, visual-only text yields higher recall than audio-visual text [-M]). Pilot both on the brief's own test, and keep any printed text short and anchored to the object it names (see Modality).
4. Stage the complexity for novices. Early runs use a simplified scenario (fewer variables, slower events, one decision at a time); add variables and time pressure as the learner succeeds, and end the unit with the whole scenario. Evidence: part-task practice reduced load for novices on high-interactivity tasks (Part-task practice reduces cognitive load for absolute novices during initial skill acquisition. [~M]); the case for ending on the whole task is a design argument, not a tested comparison (A training-design argument, not tested by any study recorded here, holds that practising whole complex tasks improves transfer of complex skills [~W]).
5. Debrief every run, and give it as much time as the run. Plan roughly as long for the debrief as for the run: what happened, the reasoning at two or three decision points, what the real task would have done differently, and one change to try next. Evidence: none here compares a run with and without a debrief (see Debriefing); trainees in tabletop exercises consistently wanted more debriefing time than scheduled (Trainees consistently wanted more debriefing time than scheduled [+W]). The time ratio is an untested proposal.
6. Run again straight after the debrief. A second attempt at the same or a varied scenario lets the change be tried. Evidence: simulation built around deliberate practice (repetition, mastery standards, immediate feedback) beat traditional clinical education on procedural skills in a meta-analysis of 14 studies, d = 0.71 (Simulation Based Education With Deliberate Practice Improves Clinical Outcomes [+M]); it tests the bundle, not the second run alone. Between sessions, leave the simulation open for learners to rerun as often as they like: across simulation-game studies, the advantage over comparison groups was larger where trainees had unlimited access to the game (d = .68) than limited access (d = .31) (Trainees learn more from simulation games when they have unlimited access to the game [+M]; a moderator comparison between groups of studies, not a test of access within one).
7. Build the bridge to the real task explicitly. End each unit with an activity that names how the simulated case maps onto the real setting, then a check in a non-simulated or real setting (a real instrument, a standardized patient, a paper case, a supervised workplace task). Evidence: games and simulations support near transfer more robustly than far or long-term transfer, and learners "struggle to generalize" without explicit connections (The review reports that simulation games support near transfer of knowledge but evidence for long-term or far transfer is less robust [~M]); bridging activities that connect game learning to classroom instruction are reported to increase learning (Students learn more from games when bridging activities connect implicit game learning with formal, explicit classroom instruction [+M], one report, no study detail); and in one study of 50 middle-school students, virtual labs did not help learners carry what a teacher had demonstrated into new inquiry contexts (Virtual labs showed a limitation in helping students transfer what they learned from teacher demonstration to new inquiry contexts [-W]; no statistic printed for that finding), so a demonstration in the simulation is not itself the bridge. In vocational simulator training, teachers were observed using "a real machine" as the benchmark when settling how a simulated task should be done (VET teachers use the real machine as the benchmark of legitimacy for simulator instruction and shift between teacher and future-colleague positions. [+W]; one ethnographic study, a practice observed, not an outcome), which is one way to make the mapping explicit during the run as well as after it.
8. Keep tactile skills on physical practice. Where the skill depends on touch or force (palpation, injection, assembly by hand), use a mannequin, model, standardized patient or real equipment for the skill itself, and use the screen or AI simulation for the decisions and communication around it. Evidence: AI-powered nursing simulations showed short-term gains in communication confidence and structured knowledge but limited or inferior effects for tactile psychomotor skills compared with standardized patients (AI-powered simulations show short-term gains in communication confidence and structured knowledge in RCTs, with limited or inferior effects for tactile psychomotor skills [~M]; One RCT found AI-assisted simulation inferior to standardized patient simulation for breast self-examination skills [-M]).
9. Plan across sessions, not one event (proposal). A default arc for a skill that matters weeks later: sessions 1–2, primer, simplified scenarios, prompts available throughout the run, facilitator-led debrief; sessions 3–4, full scenarios, prompts only on request, learners lead part of the debrief against the stated criteria; sessions 5 onward, unfamiliar variants and mixed cases with no prompts during the run, then a check in the real or non-simulated setting. A typical session of 45–60 minutes: 5–10 minutes briefing, 15–20 minutes run, 15–20 minutes debrief, 10 minutes second run. Move a learner on when they meet the stated criterion on two consecutive unfamiliar scenarios without prompts and pass the non-simulated check; add a refresher run after a gap, because gains reported at immediate tests have not always lasted (AI-generated and prebuilt simulations produced higher immediate conceptual-assessment scores than physical laboratory work, but the advantage did not persist in later examinations [~M]). All numbers here are untested proposals.

When a condition is not met: if no one can guide or debrief (self-paced use), build the run's goal, prompts and a written debrief (what happened, why, what next time) into the simulation itself, and add a short non-simulated check before the learner moves on; if the simulation cannot represent the decisions that matter, do not use it for that objective.

Virtual, remote and online laboratories (added 2026-10-17)

Scope. A virtual lab is simulated apparatus or a simulated phenomenon on a screen (an interactive science simulation, a virtual bench, a virtual patient's test results); a remote lab is real apparatus operated at a distance over a network; an online lab exercise is any lab activity run in an online course, including one done with a home kit while an instructor is online; a technology sandbox is a resettable software, network or system environment that learners configure and break. The wiki's claims test the first and the third. No claim in this wiki tests a remote lab, a home lab kit or a technology sandbox, so what follows about those is carried from virtual labs and labelled.

Stand in, supplement, or keep hands-on: decide by the outcome.

• Conceptual knowledge: a guided virtual lab can stand in. In one study of 234 undergraduates learning heat and temperature, physical, virtual and combined manipulative experimentation were equally effective and all beat instruction with no experimentation; the authors conclude that manipulation, not physicality, mattered (Physical Experience Enhances Science Learning [~M]; random assignment not stated, and a nonsignificant difference is not shown equivalence). In university physics, AI-generated and prebuilt simulations beat physical lab work on an immediate conceptual test but the difference was gone by the midterm and final (AI-generated and prebuilt simulations produced higher immediate conceptual-assessment scores than physical laboratory work, but the advantage did not persist [~M]). An older review found no clear advantage for pure discovery in a simulation over expository teaching (Studies comparing pure simulation-based discovery learning with expository teaching show no clear and univocal outcome in favor of simulations [~M]). Two later trials point the same way or further. In two randomised comparisons with secondary students on marine ecology, learning with an interactive simulation produced higher content knowledge than hands-on experiments, slightly in a 90-minute school lesson (443 students, d = 0.19) and by a wide margin in a 180-minute student-laboratory session on a more complex, multidimensional topic (367 students, d = 1.05) (Secondary school students learn more complex marine ecology content from interactive computer simulations than from hands-on experiments, with a larger advantage for the more complex multidimensional topic [+M]; post-test timing not stated on the claim page, so no delayed result). With fourth- and fifth-graders taught to design unconfounded experiments by the same instruction on physical ramps or a computer simulation, both gained equally, including in explicit verbal knowledge of the strategy (Both physical and virtual training conditions produced equal gains in children's explicit verbal knowledge of the control of variables strategy [~M]; Physical and virtual instructional materials are equally effective for teaching fourth- and fifth-graders to design unconfounded experiments when instructional method is controlled [~M]; no difference found, equivalence not tested; see Digital Learning). And manipulating a model beat watching one: undergraduates who explored nanoscale phenomena with agent-based models gained more explanatory knowledge than those who watched visualization videos of the same phenomena (27 students, η² = .15) (Undergraduate students learning nanoscience with agent-based models show greater gains in explanatory knowledge than students learning with visualization videos [+W]; a small randomised trial). Read together: substitution is defensible for concepts when the virtual lab is guided and a delayed check is planned, the simulation should be one learners operate rather than a recording of one, and where the content is complex and multivariate the simulation may do better than the bench. Hold the comparison honest: in immersive-VR comparisons where learning activities were not controlled, the VR arm came out ahead in 14 of 17 (When activities were not controlled, the confound favored the IVR condition in 14 of 17 comparisons [~M]). Where the concept is about a felt force, the same claim page records experiments in which a brief physical experience of angular momentum improved understanding, so keep a short physical experience there if one can be had.
• Procedural and manipulative skill: supplement, do not substitute. No claim in this wiki compares a virtual science lab with a bench lab on manipulative skill. The nearest evidence points one way: for a tactile clinical skill, a screen-based AI simulation did worse than a standardized patient (One RCT found AI-assisted simulation inferior to standardized patient simulation for breast self-examination skills [-M]); virtual patients showed small or negative effects against active comparators such as mannequin-based learning, against large ones over lectures and reading, in an exploratory reading of one meta-analysis (Virtual patients show larger advantages over passive traditional learning than over active learning [~M]); and across higher-education simulation studies, mannequin simulators (g = 1.07) and VR (g = 0.85) were associated with larger effects than screen-based simulations (g = 0.68), a comparison between groups of studies rather than within them (Higher levels of technology support during simulation are associated with greater effects on learning outcomes [~M]). Use the virtual lab to prepare (parts, procedure, predictions) and the hands-on session for the manipulation itself; a primer before a VR procedure improved a real-world transfer task in one experiment (Pretraining improves transfer [+M]).
• Inquiry skill (planning an experiment, testing a hypothesis, reading data): a virtual lab is a cheap place to practise it, but only with guidance. Guidance during inquiry learning improved learners' use of inquiry skills during the inquiry (d = 0.66) and learning outcomes afterwards (d = 0.50), with no difference found between types of guidance (Guidance during inquiry learning produces a medium-to-large improvement in learners' use of inquiry skills [+M]; Guidance during inquiry learning improves post-inquiry learning outcomes [+M]; inquiry learning generally, carried to virtual labs). Left alone in a simulation, learners design poor experiments (as few as 22% effective) and hold on to disconfirmed hypotheses (Learners design poor experiments [~M]; Learners have difficulty generating and adapting hypotheses [~M]); structuring the experiment into before, during and after steps helped up to two to four times as many students distinguish central concepts (Structuring the discovery process can outperform unstructured environments [+M]); successful discoverers plan across experiments and monitor progress (Successful discoverers regulate learning more systematically [~M]). Framing the same inquiry task in concrete, familiar terms lets learners use what they already know: in a randomised experiment with 57 first-year students, the concrete version of an isomorphic simulation task produced higher performance success than an intermediate version (r = .32), while the intermediate and abstract versions did not differ significantly (Concrete-task participants achieved higher performance success than intermediate-task participants in simulation-based inquiry, while intermediate and abstract conditions did not differ significantly [~W]; the pattern is not a simple concrete-to-abstract gradient). Opening variables in stages helped in complex domains and not in a simple one (Model progression benefits learning only in sufficiently complex simulations [~M]; Multiple-hypothesis strategies improve performance only when simulation complexity is low [~W]).
• Attitudes and engagement: expect favourable reports, and do not read them as learning. After a guided-inquiry lesson with a collision-carts simulation, most of 64 students rated it enjoyable and valuable and their self-rated knowledge rose (After a guided-inquiry lesson with the collision-carts simulation, students' self-reported knowledge increased and most rated the lesson enjoyable and valuable [+W]; Students reported that the virtual-lab lesson made learning fun and helped them figure out concepts themselves [+W]); some found the worksheet "just copy values" (Some students found the guided-inquiry worksheet activities insufficiently thought-provoking [~W]). VR and AR in K-12 STEM most often improved achievement and motivation in a review, but sometimes did no better than conventional technology (VR/AR most often improved academic achievement and motivation, but sometimes performed no better than conventional technologies [~M]). When courses moved online suddenly, undergraduates named the loss of hands-on experiences such as labs among their main challenges (Staying motivated was the most pervasive challenge of learning remotely, and students missed instructor feedback, peer collaboration, and hands-on experiences [~M]).

Guidance in an online lab: who does it. Students saw online lab exercises as more effective when the instructor was present online and used pre-lab, during-lab and end-of-lab questions to direct attention to the key steps (Students perceived online lab exercises as more effective when instructors were present online and used scaffolds and probes [+W]; student interviews only); teacher-led inquiry outperformed student-led inquiry by about .40 across studies (Teacher Guided Inquiry Outperforms Student Led [+M]). In higher-education simulations, examples helped learners with less prior knowledge and reflection phases helped those with more (Prior knowledge moderates scaffolding effectiveness [~M]), and prompts alone added nothing while examples combined with prompts did well in a few medical studies (Presence versus absence of individual scaffolding types did not support the added-value hypothesis, but scaffolding combinations partly did [~M]). An AI assistant can take part of this load, with limits: answering student questions about an undergraduate biology lab, ChatGPT was rated acceptable but inconsistent and less accurate than human teaching assistants, and weaker on voice, understanding and helpfulness, though students preferred its answer 40% of the time (SMEs rated ChatGPT's content accuracy acceptable but inconsistent and below human TAs [-M]; Human TAs demonstrated stronger instructor voice, understanding, and helpfulness than AI [-M]; Students rated human TA responses more favorably overall but preferred ChatGPT responses 40% of the time [~M]); in a 35-student trial, instructor feedback on lab assignments raised scores more than AI feedback (Students receiving instructor feedback showed significantly greater improvements in lab assignment scores than students receiving AI-generated feedback [-M]); and personalised ChatGPT feedback inside an augmented-reality physics lab improved outcomes in a trial reported by a review (Personalized ChatGPT-supported feedback in an augmented-reality quantum physics laboratory improved learning outcomes [+M]); and in a randomised trial with 81 undergraduates learning Internet-of-Things tasks in VR, ChatGPT-integrated feedback produced higher scores on memory, understanding and application tasks than the VR's ordinary feedback, and improved problem solving and critical thinking but not creativity (Undergraduates learning IoT tasks in VR with ChatGPT-integrated feedback reached significantly higher cognitive levels [+M]; GPT feedback in VR markedly improved problem-solving and critical thinking but had limited impact on creativity [~M]; no effect sizes printed, immediate tests only). The pattern across these: AI feedback inside the run, compared with a simulation's fixed feedback, has helped; AI in place of a person's feedback on written lab work has done worse. Logs can show who needs help: detectors trained on simulation logs inferred a student's inquiry skill after three runs (Machine-learned detectors of science inquiry skill made accurate inferences after a student ran a simulation three times [+W]). For dividing the work between an AI tutor and a person, see the AI Tutoring and Mentor Support in Self-Paced Online Learning patterns.

Safety and access. Virtual-first practice is how science teachers described handling risky or inaccessible phenomena: try the materials virtually, then use the real ones (Teachers report AI tools enable authentic, contextualized science experiences [+W], interviews only). Immersive versions bring cybersickness and cost (see the situation-table row on headsets); a browser or phone version keeps access open where devices and bandwidth are limited (proposal). No claim in this wiki measures lab-safety outcomes (incidents, safe-handling behaviour) after virtual pre-lab practice.

A default online lab (proposal, untested as a whole). (1) Pre-lab, 10–15 minutes: a primer on the apparatus and terms and two prediction questions answered in writing. (2) Guided run, 20–40 minutes: a goal for each run, a data table, a structured plan (hypothesis, variable to change, what would count against it) checked before the learner runs, prompts during the run, variables opened in stages if the system is complex. (3) Post-lab, 15–20 minutes: explain the result with the simulation off; a short write-up that a person (instructor or mentor) reads and answers, with an AI assistant allowed for procedure questions but not as the only reader. (4) Where a manipulative skill is an objective, a hands-on session (campus, local lab, residency or skills lab, or a home kit) assessed on real equipment before the skill is credited. (5) A delayed conceptual check two to four weeks later. Across a course: early labs fully scaffolded with worked examples of an experiment plan; middle labs with the plan template only; late labs with an open question, the learner's own plan, and the reviewer's feedback only after submission. Move a learner on when two consecutive lab write-ups meet the stated criterion without prompts.

Fitting the design to a situation

The three facts that most change the decision:

• What the learner must do in the real setting, and whether the simulation represents it. Decisions and communication travel well to a screen; tactile skill does not. If unstated, ask: what will the learner do in the real setting, with what hands-on component, and which decisions and cues must the simulation keep?
• The learner's starting point on this task. A novice needs a primer, simplified runs and guidance during the run; a learner who already performs the task needs varied, harder scenarios and less prompting. If unstated, ask: can they name the parts of the system, and can they carry out one run of the simplest scenario now?
• Who can guide and debrief, and for how long. Most of the support evidence comes from a facilitator or tutor responding to one learner. If unstated, ask: who will see each run, who leads the debrief, and how many minutes are there for both?

If the brief says…	Then change…	Basis
Learners are novice on the task or system	Add a 5–10 minute primer on parts and terms; start with a simplified scenario framed in concrete, familiar terms; give a goal and prompts during every run; never begin with free exploration	Pretraining learners on a system's key parts and terms before the main lesson improved knowledge and real-world transfer in one VR experiment [+M]; Assimilator learners gained significantly more in the VR (guided exploration) mode [+M] (one experiment; the learners' ages are not stated on the claim page); concrete framing from Concrete-task participants achieved higher performance success than intermediate-task participants in simulation-based inquiry [~W] (first-year students, one experiment)
Learners are advanced or practising professionals	Skip the primer and simplified runs; use unfamiliar or rare scenarios, time pressure and mixed cases; let learners lead the debrief against stated criteria	untested proposal; the part-task finding is for absolute novices only (Part-task practice reduces cognitive load for absolute novices [~M])
Learners are child or school-age in a classroom	Use a teacher-led screen simulation with a goal sheet for each run, pairs at a device, and a whole-class debrief that names the real-world phenomenon; keep runs short (10–15 minutes); where physical materials are scarce, a simulation taught with the same instruction can stand in for them	the format is an untested proposal; the substitution rests on one study of one science skill with fourth- and fifth-graders (Physical and virtual instructional materials are equally effective for teaching fourth- and fifth-graders to design unconfounded experiments when instructional method is controlled [~M], no difference found, equivalence not tested); the guided-VR experiment is the nearest evidence on guidance, and teachers report AI and virtual labs let students "try all materials virtually first, then use real ones" (Teachers report AI tools enable authentic, contextualized science experiences [+W], teacher interviews only)
Goal is a tactile or psychomotor skill	Practise the skill on a mannequin, model, standardized patient or real equipment; use screen or AI simulation only for the decisions and talk around it	One RCT found AI-assisted simulation inferior to standardized patient simulation for breast self-examination skills [-M] (nursing students)
Goal is communication, history-taking or a decision under uncertainty	An AI virtual patient or scripted role-play works as rehearsal; add a human check of communication quality, because automated scoring agreed less well for communication than for information gathering, and some learners rated human cases more authentic	Controlled GenAI virtual-patient and communication studies reported short-term skill gains, but positive findings were not universal [~M] (medical education)
Goal is a concept in science (a field, a force, a hidden process)	Use the simulation to make the invisible visible, with learners operating the model themselves rather than watching a video of it; then ask them to predict and explain with the simulation off; schedule a delayed check, not only an immediate one	AI-generated and prebuilt simulations produced higher immediate conceptual-assessment scores than physical laboratory work, but the advantage did not persist [~M] (university physics); Secondary school students learn more complex marine ecology content from interactive computer simulations than from hands-on experiments [+M] (two randomised studies, larger for the complex topic); Undergraduate students learning nanoscience with agent-based models show greater gains in explanatory knowledge than students learning with visualization videos [+W] (27 undergraduates)
Adults in work settings (vocational training, safety, customer handling)	Build scenarios from the learners' own workplace incidents, including the rare and risky ones, not only volume practice of routine tasks; make errors produce their real consequences in the simulation; refer settings and moves to the real machine during the run; end with a supervised task at work	Intelligent Extended Reality interventions in VET show consistent positive effects on procedural competence, practical skills, and learner motivation [+M] (review, no pooled effect); adult learners interviewed said they want simulations that closely resemble actual job scenarios (Adult learners expect AI to support transformative and experiential learning through assumption-challenging simulations, role play, and authentic virtual contexts [+W], expectations, not outcomes); one ethnography of forestry-machine simulator training found instruction centred on volume training of routine tasks rather than the risky situations that justify a simulator (Simulator-based VET instruction in this study centres on volume training of routine tasks rather than rehearsing specific high-stakes, risky situations. [~W]), with the simulator showing errors' consequences (Simulator feedback responds authentically to student error [+W]) and teachers using the real machine as the benchmark (VET teachers use the real machine as the benchmark of legitimacy for simulator instruction [+W]); observations of practice, no outcomes; workplace incidents untested
Setting is mixed: some learners remote, some in the room, or phone-only	Use a browser or phone simulation everyone can run, not a headset for some; let remote learners share their screen for the debrief; give everyone the same goal and debrief questions	untested proposal
No headsets, low bandwidth, or learners prone to motion sickness or with disabilities	Use a desktop or phone version, or a tabletop or paper scenario with the same decisions; offer a seated, non-immersive option; keep immersive sessions short	Cybersickness remains a persistent practical challenge, affecting approximately 42% of VR users [~M]; a K-12 review also reports dizziness, distraction, cost, infrastructure and a lack of private space for headset use (VR/AR implementation challenges fall on three fronts [~W]); the fidelity point supports a non-immersive version (Increased perceptual fidelity fails to reliably improve serious game learning outcomes [~M]); high-end equipment can cost over $72,000 a station, and the reported gain for one such system was in engagement, not learning (In one reviewed study, Anatomage virtual dissection increased medical student engagement by 34% over cadaver-based practice [~W])
The alternative to the simulation is already an active method (hands-on practice, role-play, worked problems with feedback)	Do not expect the simulation to beat it; use the simulation where it adds what the active method lacks (safety, access, rare cases, unlimited repetition) or as a supplement to it, and compare the two locally on the same aligned check before replacing one with the other	The advantage of simulation games shrinks or reverses when the comparison group learns via active instructional methods [~M] (d = −.19 against active methods); Virtual patients show larger advantages over passive traditional learning than over active learning [~M]; the supplement use is untested here
One facilitator for many learners, or scarce facilitator time (volunteers, a busy clinical shift)	Build the run's goal and prompts into the simulation; run in pairs or teams of three to four, one observer per team with a checklist; facilitator joins only the debrief, and debriefs the commonest error with the whole group	untested proposal; contingent-support evidence is mostly one-to-one (Contingent scaffolding improves learning more than fixed or absent support. [~M], carried to groups it did not test)
Teachers have little preparation time or are new to the technology	Choose one ready-made simulation aligned to one standard; give the teacher a one-page run sheet (goal, three prompts, debrief questions) and a plan for managing the room while devices are in use (who holds which device, a signal to stop and look up); start with one lesson before a unit	The review reports that insufficient prep time, curriculum-alignment difficulty, and low teacher confidence with new technologies are barriers [~M]; a K-12 STEM review adds learner distraction, loss of classroom control and difficulty designing activities (VR/AR implementation challenges fall on three fronts: learner distraction and discomfort, teacher classroom-management and design burdens, and school cost and infrastructure limits [~W]) (K-12; the run sheet and room plan are untested)
Time is one session only	One primer (5 min), one run (15 min), one debrief (15 min), one second run (10 min), one non-simulated check item (5 min); record the check and claim no retention	untested proposal
A fixed short time budget (for example 20 minutes)	Drop the second run before the debrief: primer 3 min, run 8 min, debrief 7 min, one-line plan for next time 2 min	untested proposal
Learners have low confidence, high anxiety or low literacy	Start with a private, repeatable run with no score shown; read prompts aloud or use icons; debrief one-to-one or in a small group first	Learners valued safe, repeatable, nonjudgmental AI practice environments that reduced anxiety [+W] (nursing, learners' reports); literacy untested
Recordings or logs of performance are private or sensitive (errors with a client, personal role-play)	The learner decides who sees a recording or log; debrief from the learner's own account if they choose; never show an individual's errors to the group without consent	untested proposal
The brief names its own debrief routine, rubric or checklist	Keep it as given and run it after every run; change only what the evidence says matters (a goal before the run, guidance during it, a second attempt after)	untested proposal; see Debriefing
The brief wants a whole performance kept whole (a full resuscitation, a whole negotiation)	Keep the whole scenario; simplify it by slowing events or removing distractors, not by splitting it into separate drills; debrief at its decision points	untested; A training-design argument … holds that practising whole complex tasks improves transfer [~W]
High stakes (certification, patient or public safety)	Set a mastery criterion for each scenario, require repeated runs until it is met, and confirm on a non-simulated assessment before real practice	Simulation Based Education With Deliberate Practice Improves Clinical Outcomes [+M] (procedural clinical skills)
The product or performance is for a real audience (a briefing to a real client, a demonstration for the public)	Use the simulation as rehearsal before the real event; debrief after the rehearsal and again after the real event	untested proposal
The brief reports an error learners actually made	Build that error into a scenario as an event the learner must catch, and use it as the compared case in the debrief	untested proposal
The simulation uses an AI agent or virtual patient	Brief learners that the agent may respond oddly, lag or lack non-verbal cues; ask them to note such moments, and treat them as a limit of the simulation in the debrief. Where the goal is communication, ask learners to talk to the agent as to a person (open questions, plain language) and debrief their questioning, because learners otherwise tend to query it like a database	Persistent technical limitations constitute an authenticity gap in AI-powered nursing simulations [~M]; Students treated the genAI avatar as a sophisticated question base rather than a virtual patient, using predominantly closed-ended, efficiency-focused questioning [~W] (two medical-school problem-based learning groups, descriptive; the instruction to learners is untested)
Learners study online-self-paced with no lab access, and the goal is conceptual knowledge or inquiry skill	Use a guided virtual lab in place of the bench: a goal and data table for each run, a required experiment plan before running, prompts during it, an explain-with-the-simulation-off task after; a mentor or instructor reads each write-up and answers within a stated time; schedule a delayed conceptual check	Physical Experience Enhances Science Learning [~M] (physical = virtual for concepts, one undergraduate study); Guidance during inquiry learning improves post-inquiry learning outcomes [+M]; Students perceived online lab exercises as more effective when instructors were present online and used scaffolds and probes [+W]; immediate-only gains (AI-generated and prebuilt simulations … advantage did not persist [~M]); mentor role from Mentor Support in Self-Paced Online Learning, untested for labs
Self-paced online, no lab access, and the competency includes a manipulative or bench skill	Do not credit the skill from the virtual lab; use the virtual lab to prepare (parts, steps, predictions) and add a hands-on component (a local or campus lab, a residency, a home kit) assessed on real equipment	One RCT found AI-assisted simulation inferior to standardized patient simulation for breast self-examination skills [-M] (carried from a tactile nursing skill); Pretraining improves transfer [+M]; no claim in this wiki compares a virtual and a bench science lab on manipulative skill
Health-professions skills lab (clinical procedures, examination, resuscitation)	Use screen-based virtual patients for the reasoning before the skills lab and blended with it, not in place of it; keep the procedure on mannequins, task trainers or standardized patients with a mastery criterion; rehearse the conversation with a simulated person	Virtual patients more effectively improve skills than traditional education [+M]; Virtual patients show larger advantages over passive traditional learning than over active learning [~M]; Higher levels of technology support during simulation are associated with greater effects [~M]; Simulation Based Education With Deliberate Practice Improves Clinical Outcomes [+M]; for the conversation see Practice with Simulated People
Technology labs: a sandbox, virtual machine or simulated network learners configure, break and repair	Give each lab a fault to diagnose or a configuration goal, a reset button, components opened in stages, and a short written account of what was changed and why that a person reviews; finish on the real or production-like system where the competency requires it	untested for technology sandboxes in this wiki; carried from vocational XR (Intelligent Extended Reality interventions in VET show consistent positive effects on procedural competence [+M], review, no pooled effect) and from simulation effects being largest on technical performance (Simulation effects vary by complex skill [~M])
Learners ask an AI assistant questions during an online lab, or AI gives the lab feedback	Let the AI answer procedure and terminology questions and give feedback during the run in place of the simulation's fixed messages; keep a person reading the write-ups and checking content the AI explains; tell learners the AI can be confidently wrong; do not expect in-run AI feedback to help creative or open design work	Undergraduates learning IoT tasks in VR with ChatGPT-integrated feedback reached significantly higher cognitive levels [+M] and GPT feedback in VR … limited impact on creativity [~M] (one undergraduate RCT, against the VR's ordinary feedback); SMEs rated ChatGPT's content accuracy acceptable but inconsistent and below human TAs [-M]; Students receiving instructor feedback showed significantly greater improvements in lab assignment scores than students receiving AI-generated feedback [-M]; Personalized ChatGPT-supported feedback in an augmented-reality quantum physics laboratory improved learning outcomes [+M]; see AI Tutoring
A mentor is available on demand but does not watch runs	Use the simulation's logs (experiments run, variables changed, plan written or not) as the signal for a mentor to reach out, and check the signal against write-up quality before relying on it	Machine-learned detectors of science inquiry skill made accurate inferences after a student ran a simulation three times [+W]; outreach on that signal untested
The procedure is hazardous, or the phenomenon cannot be reached (toxic reagents, high voltage, a rare event)	Rehearse it virtually first, with the safety decisions as events the learner must handle, then do it supervised on real equipment	Teachers report AI tools enable authentic, contextualized science experiences [+W] (interviews); safety outcomes untested in this wiki
Learners' prior knowledge of the domain differs (novice vs advanced)	Give novices worked examples of an experiment or case before they run; give advanced learners reflection phases after the run; do not rely on prompts alone	Prior knowledge moderates scaffolding effectiveness [~M]; Presence versus absence of individual scaffolding types did not support the added-value hypothesis, but scaffolding combinations partly did [~M] (higher education)
Learners are teachers, student teachers or teaching assistants rehearsing teaching	Choose the medium by the teaching situation: a VR or virtual classroom for situations that are hard to stage live (a lab session, a rare incident), live role-play with peers for whole-class interaction; AI-simulated student ideas can serve as material for practising noticing and planning a response	VR simulation better promoted lab-teaching knowledge development than live classroom simulation, while live simulation better fostered class-teaching knowledge development [~M] (one randomised comparison, 40 teaching assistants, one 4-hour session); Teachers judged LLM-simulated student ideas realistic and most useful when the ideas sparked sensemaking about student thinking [+W] (six teachers' judgements, no learning outcome)
Learners work in pairs or groups at one simulation, or role-play together in a shared VR scene	Give each person a role (operator, predictor, recorder, or a part in the role-play) and rotate them; stop the run at set points for the pair to say what they expected, what happened and what they now think; in multiuser VR, script who speaks and when, and check that everyone takes part, because shared presence does not produce joint engagement by itself	Computer-supported collaborative learning with simulation programs fostered conceptual change for students who were cognitively engaged and prepared to reflect on and reconstruct their conceptions [~W] (seven Grade 10 dyads, qualitative); VR participants lacked mutual engagement in collaborative role-playing during the teaching simulation [-W] (teaching assistants); roles and stop points untested
A simulation or VR "field trip" is used to build background knowledge before reading or a lesson	Keep it short (5–10 minutes), name the three or four things learners should notice, and go straight to the text or lesson that uses them; ask learners afterwards what they saw that the text mentions	untested proposal; the case rests on prior knowledge aiding comprehension (Prior Knowledge Impacts Text Understanding [+W], a practitioner article's synthesis with no new data), not on any test of VR as the source of that knowledge

Twenty-six of the thirty-three rows cite a claim, and most of those carry evidence from health-professions, physics or VR studies to the setting named; claims for remote learners, single sessions, private material and technology sandboxes are absent from this wiki, so those rows are proposals. For children the wiki holds one study (physical against virtual materials for one science skill), which supports the substitution clause in that row and not its format. The teacher-education and group rows rest on one small study each, and the work-settings row's additions rest on one ethnography with no outcome measure.

Observation, state and explanation

What can be observed is a run under stated conditions: the scenario and its settings, the prompts given, the actions taken and their order, the time taken, the errors, what the learner said in the debrief, and what changed on the next run. "Understands the system", "can transfer to the real setting" and "is overloaded" are inferred states. Success inside the simulation shows what the learner can do in that representation; it does not show what they can do in the real setting.

Response	Plausible explanations	Observation that could change the interpretation
Succeeds in the simulation, fails on the real equipment or a standardized patient	The skill is tactile and the simulation did not represent it; the real setting adds cues or pressure the simulation lacked; success relied on in-run prompts	Give a matched real-setting task with no prompts; ask what was different; compare with a prompt-free simulated run.
Explores freely, clicks through, cannot say what happened	No goal for the run; the interface is consuming attention; the learner lacks the terms to make sense of what they see	Repeat with a stated goal and a primer on parts; ask the learner to narrate one decision.
Rates the simulation highly, shows no gain on the check	Enjoyment or presence without the target practice; the check does not match what the simulation practised; novelty	Compare a check aligned to the simulated decisions with one aligned to the real task; ask which decision the learner would now make differently.
Stops early, reports nausea or discomfort	Cybersickness; anxiety; an inaccessible control	Offer a seated or desktop version and see whether the learner continues and performs.
Repeats the same error across runs	The debrief did not reach the reasoning behind it; a misconception the simulation does not expose; the scenario rewards the error	Ask the learner to explain the decision before the next run, and give a scenario where the error produces a visible consequence (as a forestry-machine simulator does when a load falls from a wrongly opened grapple: Simulator feedback responds authentically to student error [+W], an observation, not a test).
In a pair, group or multiuser VR run, one learner acts while the others watch or drift	No roles were assigned; the headset or interface isolates each person; the others lack the confidence or the controls to join; they are engaged but not thinking about the concept	Assign and rotate roles, then see whether the quieter learners act; ask each person separately to predict the next result and explain it afterwards. Joint activity alone did not ensure conceptual change in one study of pairs (CSCL with simulation programs fostered conceptual change for students who were cognitively engaged [~W]).

These rows are proposals, untested with learners as diagnostics. A probe is itself practice and changes the state it measures, so record each one.

Evidence and qualifications

• Simulation Based Education Improves Outcomes [+M]: a meta-analysis of 609 studies (35,226 health-professions trainees) found technology-enhanced simulation beat no intervention, with large pooled effects on knowledge (g = 1.20), skills (about 1.09–1.18) and behaviours, and a moderate effect on patient outcomes (g = 0.50). The comparison is with no intervention, not with matched non-simulated practice; heterogeneity was large; subgroup analyses found no consistent interaction with feedback, repetition, distributed practice or curricular integration. It supports using simulation over nothing; it does not say which design features matter.
• Simulation Based Education With Deliberate Practice Improves Clinical Outcomes [+M]: a meta-analysis of 14 studies found simulation-based medical education with deliberate practice beat traditional clinical education on clinical skill acquisition, d = 0.71, mostly procedural and psychomotor skills. It tests the bundle (repetition, mastery standards, immediate feedback) against ordinary training.
• Assimilator learners gained significantly more in the VR (guided exploration) mode than in VR (non-guided exploration) or Non VR modes [+M] and Accommodator learners gained significantly more in the VR (guided exploration) mode [+M]: one experiment (Chen et al. 2005) with pre-test as covariate; guided VR had the highest adjusted gains, with η² = 0.284 and 0.134 for the mode effect in the two groups. VR (non-guided exploration) did not differ significantly from Non VR [~M]: free exploration in VR was no better than no VR (a nonsignificant difference, not shown equivalent). In the VR (guided exploration) mode, assimilator and accommodator learners did not differ significantly in gain score [~M]: equivalence was not tested. Together they support guidance, not immersion, as what made the difference in this one study; the learners' ages and the domain are not stated on the claim pages.
• AI-powered simulations show short-term gains in communication confidence and structured knowledge in RCTs, with limited or inferior effects for tactile psychomotor skills [~M] and One RCT found AI-assisted simulation inferior to standardized patient simulation for breast self-examination skills [-M]: a mixed-methods systematic review of AI-powered nursing simulation (3 RCTs). This is the main limiting evidence: for a tactile skill, a screen-based AI simulation did worse than a standardized patient.
• Controlled GenAI virtual-patient and communication studies reported short-term skill gains, but positive findings were not universal [~M]: a scoping review in medical education; several controlled studies reported short-term gains in history taking, communication and OSCE practice; some found no objective change, and human cases were sometimes rated more educational or authentic.
• AI-generated and prebuilt simulations produced higher immediate conceptual-assessment scores than physical laboratory work, but the advantage did not persist in later examinations [~M] and The review reports that simulation games support near transfer of knowledge but evidence for long-term or far transfer is less robust [~M]: both limit the horizon. Plan delayed checks and refreshers, and do not read an immediate gain as lasting.
• Increased perceptual fidelity fails to reliably improve serious game learning outcomes (the fidelity fallacy) [~M]: a book-chapter review reporting, from earlier systematic reviews, that more realism does not reliably improve learning and can raise extraneous load and cost. Conventional training received significantly higher ratings for perceived realism and immersion than VR-based training (r=0.774) [~M]: in one within-participants comparison of 20 students, conventional radiopharmaceutical training was rated more realistic than the VR module; a VR environment is not automatically the more realistic one. Immersive technologies are rarely adopted in robotics GBL/gamification but studies using them report heightened engagement, improved spatial reasoning, and enhanced perceived realism [~W]: a robotics-education review whose immersive studies report positive ratings and some performance gains, with no comparison of immersion levels; it does not show that immersion, rather than the practice, made the difference.
• Pretraining learners on a system's key parts and terms before the main lesson improved knowledge and real-world transfer in one VR experiment [+M]: 93 participants randomized; the benefit appeared on transfer to the real task, not on performance inside VR. Its table counts 4 distinct studies.
• The review reports that participants in simulation-game conditions had significantly higher skill-based knowledge outcomes than those in non-game conditions [+M]: Sitzmann's meta-analysis of 65 studies, cited by a later review with no effect size printed; the comparator is non-game instruction of unstated form. It extends the comparison beyond no intervention, for applied knowledge, but does not isolate any design feature.
• In one meta-analysis, modality-effect studies using virtual reality head-mounted displays show larger effects than studies using computer screens or audio-tapes [~M]: a moderator in Ginns's meta-analysis (VR d = 1.34, screens d = 0.55); it bears on how a headset simulation presents explanations, not on whether simulation helps. Marked mixed since 2026-10-19 (wave 2), for consistency with Modality: a later VR experiment found the reverse (In immersive VR, visual-only text yields higher recall than audio-visual text [-M]; Albus & Seufert 2023, 61 learners), so step 3 no longer tells a designer to narrate in a headset.
• Learners' and teachers' reports, used for single rows and steps and not as tests: learners valued safe, repeatable practice (Learners valued safe, repeatable, nonjudgmental AI practice environments that reduced anxiety and bridged the theory-practice gap [+W]); learners reported robotic interactions, delays and missing non-verbal cues in AI simulations (Persistent technical limitations constitute an authenticity gap in AI-powered nursing simulations [~M]); teachers described virtual-first, then real, practice (Teachers report AI tools enable authentic, contextualized science experiences [+W]) and named preparation time and confidence as barriers (The review reports that insufficient prep time, curriculum-alignment difficulty, and low teacher confidence with new technologies are barriers to implementing simulations in K-12 [~M]); about 42% of VR users are reported to have cybersickness symptoms (Cybersickness remains a persistent practical challenge [~M]); trainees wanted more debriefing time than scheduled (Trainees consistently wanted more debriefing time than scheduled [+W]); bridging activities between game and instruction are reported, without study detail, to increase learning (Students learn more from games when bridging activities connect implicit game learning with formal, explicit classroom instruction [+M]); and vocational XR showed positive effects on procedural competence without a pooled effect (Intelligent Extended Reality interventions in VET show consistent positive effects on procedural competence, practical skills, and learner motivation [+M]). Contingent scaffolding is carried from tutoring (step 3).

Simulation-based learning in higher education (Chernikova et al. 2020), added 2026-10-17. One meta-analysis of 145 studies (10,532 participants, mostly medical and teacher education) supplies the moderators below; each is a comparison between groups of studies, not a test inside a study, and all come from one source.

• Simulation-based learning has a large positive overall effect on complex skills in higher education (g = 0.85) [+M]: the effect held against no intervention (g = 1.02), against differently instructed controls (g = 0.82) and against baseline (g = 0.88), so it extends the health-professions meta-analysis above beyond a no-intervention comparator.
• Simulation effects vary by complex skill: highest for technical performance, lowest for communication skills [~M]: g = 1.06 for technical performance, 0.82 diagnostic, 0.88 problem solving, 0.50 collaboration and 0.44 communication. It tempers the communication gains reported for AI virtual patients above.
• High authenticity simulations show higher effects than low authenticity simulations [~M] (g = 0.86 against 0.58) and Higher levels of technology support during simulation are associated with greater effects on learning outcomes [~M] (mannequin simulators 1.07, VR 0.85, screen-based 0.68): both qualify the fidelity statement in the conditional relationship. Authenticity here means how fully the situation represents the real one, which is close to "keeps the decisions and cues" in step 1, and the technology contrast is between categories of studies whose tasks also differ; neither isolates visual realism, so the fidelity-fallacy reading stands for appearance and the decisions-and-cues reading is strengthened.
• Live simulations with real patients showed the greatest effects among simulation types [~M]: document-based simulations (working on X-rays, student homework and similar documents) were the least effective type (g = 0.31).
• Longer simulation duration is associated with higher effects, but evidence for simulations over one month is inconclusive [~M]: under an hour g = 0.65; several hours to a day 0.81. It bears on step 9's arc; it does not set a dose.
• Prior knowledge moderates scaffolding effectiveness [~M] and Presence versus absence of individual scaffolding types did not support the added-value hypothesis, but scaffolding combinations partly did [~M]: examples helped learners at lower education levels, reflection phases those at higher levels, and prompts on their own were associated with smaller effects (g = 0.65 with prompts against 0.92 without). This complicates step 3's prompts: pair them with an example for novices, and replace them with a reflection phase for advanced learners.

Virtual, remote and online laboratories, added 2026-10-17. The subsection under the default design reads these claims; their limits are recorded here.

• Physical Experience Enhances Science Learning [~M]: two entries pulling in different directions. Kontra et al. (2015), read from the abstract only, found a brief physical experience of angular momentum improved quiz scores; Zacharia and Olympiou (2011) found physical, virtual and combined experimentation equally effective for 234 undergraduates on heat and temperature, all better than no experimentation (random assignment not stated; no effect sizes recorded; equivalence not tested). Neither measures manipulative skill or a delayed outcome.
• AI-generated and prebuilt simulations produced higher immediate conceptual-assessment scores than physical laboratory work, but the advantage did not persist in later examinations [~M]: one quasi-experiment reported second-hand by a review (187 enrolled, three lab conditions, η² = 0.359 immediately).
• Studies comparing pure simulation-based discovery learning with expository teaching show no clear and univocal outcome in favor of simulations [~M], Learners design poor experiments [~M], Learners have difficulty generating and adapting hypotheses [~M], Structuring the discovery process can outperform unstructured environments [+M], Successful discoverers regulate learning more systematically [~M] and Model progression benefits learning only in sufficiently complex simulations [~M]: all from one 1998 narrative review of discovery learning with computer simulations of conceptual domains, reporting earlier studies, mostly without effect sizes. They describe what goes wrong without guidance; they are old and second-hand.
• Guidance during inquiry learning improves post-inquiry learning outcomes (d = 0.50) [+M] and Guidance during inquiry learning produces a medium-to-large improvement in learners' use of inquiry skills [+M]: one meta-analysis of inquiry learning (72 studies, 20 measuring inquiry activities); the claim pages do not say how many used simulations, so applying them to virtual labs is a carry. Teacher Guided Inquiry Outperforms Student Led [+M] is a between-study moderator in school science inquiry.
• Students perceived online lab exercises as more effective when instructors were present online and used scaffolds and probes [+W]: student interviews in course case studies during the 2020 move online; perceptions, no outcome measure. Staying motivated was the most pervasive challenge of learning remotely [~M]: a national survey of undergraduates in emergency remote teaching, not designed online labs.
• SMEs rated ChatGPT's content accuracy acceptable but inconsistent [-M], Human TAs demonstrated stronger instructor voice, understanding, and helpfulness than AI [-M] and Students rated human TA responses more favorably overall but preferred ChatGPT responses 40% of the time [~M]: one study of eight questions about one undergraduate biology lab, rated by experts and students, with no learning outcome. Students receiving instructor feedback showed significantly greater improvements in lab assignment scores than students receiving AI-generated feedback [-M]: a 35-student randomised trial, rigour coded low. Personalized ChatGPT-supported feedback in an augmented-reality quantum physics laboratory improved learning outcomes [+M]: a randomised trial reported by a review (d = 0.495–1.892 across outcomes including eye tracking). The AI evidence is mixed and comes from different generations of models; it supports AI as a supplement with a person in the loop, not as the only reader of lab work.
• Self-reports and engagement (collision-carts survey [+W], fun and figuring concepts out [+W], worksheets too thin [~W], VR/AR review [~M]) are evidence of attitudes and valued experience, one design report and one review; detectors of inquiry skill from logs [+W] are reported by an overview article without accuracy figures on the claim page.
• Whole-medium comparisons (virtual against hands-on, online against campus) confound the medium with the method delivered through it (Media comparison studies produce uninterpretable "no significant difference" findings [~M]); compare guided virtual with guided hands-on, holding the plan, prompts and write-up equal, before deciding a substitution.

Read against the model on 2026-10-19. Claims that had been listed under Further evidence were read and placed in the model:

• Functional correspondence predicts simulation learning outcomes better than physical resemblance or appeals to authenticity [+M]: an editorial in Instructional Science (Sellberg & Lindwall 2025) summarising Chernikova et al.'s meta-analyses; no effect size printed. It agrees with the authenticity moderator above and with step 1's "decisions and cues, not looks", and adds salience of the relevant information. It is the same underlying evidence as the Chernikova entries, not a second confirmation.
• Undergraduates learning IoT tasks in VR with ChatGPT-integrated feedback reached significantly higher cognitive levels [+M] and GPT feedback in VR markedly improved problem-solving and critical thinking but had limited impact on creativity [~M]: one randomised trial (Wang et al. 2025, 81 undergraduates); the comparator is the same VR with its ordinary feedback, so it tests the kind of feedback inside a simulation, not simulation itself; no effect sizes in the available text, rigour not established on the first page and r2 on the second, immediate outcomes only.
• VR simulation better promoted lab-teaching knowledge development than live classroom simulation, while live simulation better fostered class-teaching knowledge development [~M] and VR participants lacked mutual engagement in collaborative role-playing during the teaching simulation [-W]: one randomised comparison (Ke et al. 2021) of 40 teaching assistants in a single four-hour session, no effect sizes printed. The crossover is the useful part: neither medium won overall, each was better for the teaching situation it represented more fully, which is the functional-correspondence point again. The engagement finding is from the study's qualitative data.
• Computer-supported collaborative learning with simulation programs fostered conceptual change for students who were cognitively engaged and prepared to reflect on and reconstruct their conceptions [~W]: seven Grade 10 dyads at a physics microworld over a ten-week unit, qualitative case analysis with pre, post and delayed tests; good collaboration was not enough without reflection on one's own conception.
• Teachers judged LLM-simulated student ideas realistic and most useful when the ideas sparked sensemaking about student thinking and suggested scaffolds [+W]: interviews with six teachers; judgements of usefulness, no learning outcome.
• Prior Knowledge Impacts Text Understanding [+W]: carried in from the merged VR-for-reading fragment; a practitioner article restating the prior-knowledge literature, with no DOI and no data of its own. It supports building background knowledge, not VR as the way to build it.

Read against the model on 2026-10-19 (wave 2). Claims linked by the orphan-claim linker after the morning's re-read, read from their Evidence entries:

• The advantage of simulation games shrinks or reverses when the comparison group learns via active instructional methods [~M], Published studies of simulation games report substantially larger effects than unpublished studies, indicating publication bias [~M] and Trainees learn more from simulation games when they have unlimited access to the game [+M]: three moderator analyses from one meta-analysis of computer-based simulation games in work-related training (Sitzmann 2011; 14 active-comparison samples, N = 832; 44 published against 16 unpublished samples; 10 unlimited-access against 45 limited-access samples). Between-study comparisons, so the moderators are confounded with whatever else differed among those studies. The first two narrow the conditional relationship to comparisons with no or passive instruction and warn that pooled effects are inflated; the third supports leaving runs open between sessions (step 6). The claim pages carry no DOI for this source.
• Secondary school students learn more complex marine ecology content from interactive computer simulations than from hands-on experiments [+M]: two randomised comparisons (Krüger et al. 2022; 443 and 367 students in grades 10–13), d = 0.19 in a 90-minute school lesson and d = 1.05 in a 180-minute student-lab session on a more complex topic. Content knowledge only; the lab and topic both differ between the two studies, so complexity is not isolated as the cause of the larger effect. It strengthens the virtual-labs rule that a guided simulation can stand in for concepts.
• Both physical and virtual training conditions produced equal gains in children's explicit verbal knowledge of the control of variables strategy [~M]: Triona & Klahr 2003, the same study Digital Learning reads (92 fourth- and fifth-graders, same instruction on physical or virtual materials; no condition effect, equivalence not tested). Brought in with its companion claim on experiment design (Physical and virtual instructional materials are equally effective [~M]) as the basis for the child row's substitution clause.
• Undergraduate students learning nanoscience with agent-based models show greater gains in explanatory knowledge than students learning with visualization videos [+W]: Lai 2026, 27 undergraduates randomised, η² = .15 for the interaction; small. It supports operating a model over watching one (concept row), not simulation over hands-on work.
• Concrete-task participants achieved higher performance success than intermediate-task participants in simulation-based inquiry, while intermediate and abstract conditions did not differ significantly [~W]: Lazonder et al. 2009, 57 first-year students, three isomorphic task framings; concrete beat intermediate (r = .32), abstract was not significantly different from intermediate (p = .06). Marked mixed because the effect is not a simple gradient; it supports concrete framing of novices' first tasks.
• Virtual labs showed a limitation in helping students transfer what they learned from teacher demonstration to new inquiry contexts [-W]: Liu et al. 2022, 50 middle-school students; the authors' statement, with no statistic for it in the available text. It adds to the near-transfer limit under step 7.
• Simulator-based VET instruction in this study centres on volume training of routine tasks rather than rehearsing specific high-stakes, risky situations. [~W], Simulator feedback responds authentically to student error [+W] and VET teachers use the real machine as the benchmark of legitimacy for simulator instruction and shift between teacher and future-colleague positions. [+W]: one ethnographic study of forestry-machine simulator training (Messina Dahlberg & Gustavsson 2025), observations and video, no learning outcome. Used in the work-settings row, step 7 and the observation table as descriptions of practice.
• Students treated the genAI avatar as a sophisticated question base rather than a virtual patient, using predominantly closed-ended, efficiency-focused questioning [~W]: one mixed-methods study (Mool et al. 2026) of two medical problem-based learning groups using a generative-AI patient; descriptive. It adds a learner behaviour to the AI-agent row.
• VR/AR implementation challenges fall on three fronts: learner distraction and discomfort, teacher classroom-management and design burdens, and school cost and infrastructure limits [~W]: a K-12 STEM review (Jiang et al. 2025) coding challenges reported across studies; counts of articles, no effects. Used in the teacher-preparation and headset rows.

Learners, tasks and outcomes differ across these claims (health-professions trainees, university physics students, nursing students, learners in a VR lab), so do not rank them by effect size. None measures delayed retention of a simulation gain against matched non-simulated practice.

The evidence by question

• Does practising in a simulation improve knowledge and skill more than no simulation or conventional training, and for which skills and horizons? Simulation Based Education Improves Outcomes [+M]
• Does a simulation beat other active instruction, or only no or passive instruction? The advantage of simulation games shrinks or reverses when the comparison group learns via active instructional methods [~M]; Published studies of simulation games report substantially larger effects than unpublished studies, indicating publication bias [~M]
• Does more perceptual fidelity or immersion improve learning from a simulation? Increased perceptual fidelity fails to reliably improve serious game learning outcomes (the fidelity fallacy) [~M]; qualified by High authenticity simulations show higher effects than low authenticity simulations [~M] and Functional correspondence predicts simulation learning outcomes better than physical resemblance or appeals to authenticity [+M]
• Does AI-generated feedback inside a simulation help more than the simulation's own feedback? Undergraduates learning IoT tasks in VR with ChatGPT-integrated feedback reached significantly higher cognitive levels [+M]
• Is VR or live role-play the better rehearsal for teaching? VR simulation better promoted lab-teaching knowledge development than live classroom simulation, while live simulation better fostered class-teaching knowledge development [~M]
• Which features of a simulation go with larger effects in higher education (skill, authenticity, technology, duration, scaffolding)? Simulation-based learning has a large positive overall effect on complex skills in higher education (g = 0.85) [+M]
• Can a virtual lab stand in for a hands-on lab, and for which outcomes? Physical Experience Enhances Science Learning [~M]; Secondary school students learn more complex marine ecology content from interactive computer simulations than from hands-on experiments [+M]
• Does access between sessions change what learners gain? Trainees learn more from simulation games when they have unlimited access to the game [+M]
• How much guidance do learners need in a virtual lab or inquiry simulation? Guidance during inquiry learning improves post-inquiry learning outcomes (d = 0.50) [+M]
• Can an AI assistant stand in for a lab instructor or teaching assistant? SMEs rated ChatGPT's content accuracy acceptable but inconsistent across questions and significantly below human TAs' accuracy [-M]

Further evidence, not yet read against this model

Read against this model on 2026-10-15 and again on 2026-10-19; the claims below do not change the relationship, a step or a row, and are kept with how far they bear. Nine claims once listed here are now cited in the model: functional correspondence (step 1), the two ChatGPT-feedback-in-VR results (virtual-labs AI paragraph and AI row), the VR-versus-live teaching simulation and its engagement finding, the LLM-simulated students and the collaborative-microworld study (the teacher-education and group rows, and the observation table), prior knowledge and reading (the VR field-trip row), and multiple-hypothesis strategies (virtual-labs inquiry skill); see "Read against the model on 2026-10-19" under Evidence and qualifications. Fifteen claims linked later the same day were read in a second pass; thirteen are now in the model (see "wave 2" under Evidence and qualifications) and two are listed below.

• Learners in immersive media environments tend to treat pedagogical agents as real people, relying primarily on biologically primary communication patterns [~W] — a conceptual framework's prediction, no data; bears on designing agents inside VR, not on learning from the simulation.
• Immersive media environments enable higher visual and behavioral realism of pedagogical agents than non-immersive formats [+W] — a conceptual argument from the same framework, no test; realism is not shown to matter for learning (see the fidelity claims above).
• Agent realism, particularly behavioral realism, should be a crucial predictor of social presence in immersive agent-centered learning environments [~W] — a conceptual prediction about social presence, not learning; no test.
• A student began interacting with the sim within 10 seconds and verbalized sense-making without explicit guidance [~W] — added 2026-10-17: one sixth-grader's think-aloud with a PhET simulation designed with implicit scaffolding; it shows a well-designed interface can invite exploration, not that the exploration teaches without guidance, so it does not change the guidance rule in the virtual-labs subsection. Re-read 2026-10-19: one interview, an observation about interface design, not an outcome.
• Families' geoscience sense-making during learning-on-the-move was shaped by the immersive qualities of a mobile AR app [+W] — read 2026-10-19: a special-issue introduction summarising one qualitative study of families with a mobile AR app in a garden; second-hand, no outcome measure. It describes informal, place-based use, which no row here covers; one second-hand account is too little to write one.
• In a rule-based simulation game with narrow scope and high distance, facilitator interventions were predominantly first- and second-order and participant learning was mainly first- and second-order [~W] — read 2026-10-19: coded transcripts from one simulation-game case in professional training; a game whose rules absorb attention drew the facilitator to procedural interventions. It bears on facilitating simulation games and their debriefs (Debriefing, Game-Based Learning) more than on this page's steps.
• Guided inquiry learning with virtual and video experiments during Covid-19 remote schooling significantly improves seventh-graders' conceptual understanding in physics [+M] — attached 2026-10-23 from Flegr et al. (2023), which proposed "Use virtual experiments, optionally combined with video experiments, for guided inquiry learning when hands-on experimentation is not possible".

Objective and learner-valued goal

The designer's objective is usually performance in the real setting: a procedure done safely, a decision made well, a conversation handled, a concept used to predict. Elicit separately what the learner wants from the simulation: to stop feeling anxious before the real thing, to pass an assessment, to try equipment they cannot otherwise reach, or to enjoy it. Agreement and divergence are both observations. A learner who values a safe place to make mistakes may want many runs with no score shown; a designer who needs certification needs a criterion and a non-simulated check. Learners' high ratings of a simulation are evidence of what they value, not of the capability.

For example, a nursing student may want to feel confident talking to a distressed patient, while the course objective is a correct breast examination. An AI virtual patient serves the first; the second needs a standardized patient or model. Record both, so that confidence gained in conversation is not read as the examination skill.

What would revise this model?

The expectation should weaken if simulation practice shows no advantage over equally long non-simulated practice with the same feedback (the one moderator in the wiki that comes close, active-method comparison groups in simulation-game studies, already shows none), on aligned measures in the real setting; if gains seen at immediate tests regularly disappear at a delay even with refreshers; or if free exploration in a well-designed simulation matches guided exploration in further experiments. It should strengthen, and become more specific, if experiments vary one feature at a time (debrief or none, guided or free, high or low fidelity, one run or several) and find that feature matters. If immersive VR proves better than a desktop version of the same scenario with guidance held equal, the fidelity statement above should change.

For laboratories, the substitution rule above (a guided virtual lab may stand in for concepts, not for manipulative skill) should change if a trial comparing a guided virtual lab with an equally guided hands-on lab finds no difference on a real-equipment skill test, or finds the conceptual equivalence fails at a delay; it should change for remote labs and technology sandboxes only once claims about them exist in this wiki.

Success inside the simulation, immediate performance outside it, delayed retention, transfer to the real setting and effects on patients or clients are separate claims. The present evidence sets no number of runs, no debrief length and no fidelity level.

Claims

• VR improved movement-based transfer test performance versus reading in security training, while learning in VR without the possibility to act was detrimental [+W]
• Immersion principle: people do not necessarily learn better in 3D immersive virtual reality than with a corresponding 2D desktop presentation (d = -0.10 across 9 tests) [-W]
• Performance during the virtual pipetting lesson predicts real-life transfer test performance [+W]
• In Desktop-VR, interaction format (object movement vs. camera movement vs. no interaction) showed no significant main effect on learning outcomes [~W]
• Immersive lessons show a small effect size advantage over less immersive technology, but increased presence does not necessarily lead to more learning [~W]
• Collaboration format (individual vs. dyad) was experimentally manipulated alongside feedback format in simulation-based teacher learning [~W]

Related Principles

• Experiential Learning — simulations provide structured experiences that later reflection can interpret.
• Problem-based Learning — many simulations organize learning around decisions in realistic scenarios.
• Immediate Feedback — simulations are more effective when consequences and coaching are visible.
• Guided Practice — novice learners often need supported runs before independent performance.
• Debriefing — owns the facilitated conversation after a run, which this page's model relies on.
• Deliberate Practice — simulation-based mastery training is one of its settings.
• Game-Based Learning — simulation games sit between the two pages.
• Digital Learning — a medium is not a method; applies to VR and AI simulations.

Examples

• Clinical decision simulator: Learners diagnose, prioritize, and respond to a changing patient scenario.
• Virtual workplace rehearsal: Learners practice safety, customer interaction, or troubleshooting in a modeled environment.
• Civic or historical simulation: Learners make decisions within a constrained system and analyze outcomes.
• Goal-Based Scenarios (GBS)
• Use AI-based simulations as low-risk environments for practising and making visible teacher noticing
• Simulation and Debrief — the elements this principle is built from.
• Simulation Based Learning
• Teachers and students create their own 360-degree VR content using free apps like Google Street View
• Vocational education programs should let trainees experience and reflect on authentic errors, e.g. via error management training and negative knowledge-diaries
• Practice with Simulated People — the pattern for rehearsing a conversation or interpersonal performance with a standardized or virtual patient, simulated client or AI role-play partner.
• AI Tutoring — when an AI assistant guides runs or answers lab questions, and when a person takes over.
• Mentor Support in Self-Paced Online Learning — who guides and reviews a self-paced learner's virtual-lab work, and on which signals.
• Buoyancy physics virtual lab with variable datasheet
• Use advanced immersive learning technologies as research and learning tools to deliver developmentally-appropriate metacognitive scaffolding
• Text-based heterogeneity simulation with simulated peer feedback

Key Sources

• Akcayir, M., & Akcayir, G. (2017). Advantages and challenges associated with augmented reality for education: A systematic review of the literature. Educational Research Review, 20, 1-11. https://doi.org/10.1016/j.edurev.2016.11.002
• Merchant, Z., Goetz, E. T., Keeney-Kennicutt, W., Cifuentes, L., Kwok, O., & Davis, T. J. (2013). Exploring 3-D virtual reality technology for spatial ability and chemistry achievement. Journal of Computer Assisted Learning, 29(6), 579-590. https://doi.org/10.1111/jcal.12018
• Moyer, R., Snodgrass, J., Klein, S., & Tebben, C. (2017). Simulated work-based learning: Instructional approaches and noteworthy practices. Office of Career, Technical, and Adult Education, U.S. Department of Education.
• Shadiev, R., Xueying, W., & Huang, Y. M. (2020). Promoting intercultural competence in a learning activity supported by virtual reality technology. International Review of Research in Open and Distributed Learning, 21(3), 157-174. https://doi.org/10.19173/irrodl.v21i3.4752
• Pilgrim, J. M., & Pilgrim, J. (2016). The use of virtual reality tools in the reading-language arts classroom. Texas Journal of Literacy Education, 4(2). https://portal.issn.org/resource/ISSN/2374-7404
• Chernikova, O., Heitzmann, N., Stadler, M., Holzberger, D., Seidel, T., & Fischer, F. (2020). Simulation-based learning in higher education: A meta-analysis. Review of Educational Research. https://doi.org/10.3102/0034654320933544
• Zacharia, Z. C., & Olympiou, G. (2011). Physical versus virtual manipulative experimentation in physics learning. Learning and Instruction, 21(3), 317-331. https://doi.org/10.1016/j.learninstruc.2010.03.001
• Lazonder, A. W., & Harmsen, R. (2016). Meta-analysis of inquiry-based learning: Effects of guidance. Review of Educational Research, 86(3), 681-718. https://doi.org/10.3102/0034654315627366
• de Jong, T., & van Joolingen, W. R. (1998). Scientific discovery learning with computer simulations of conceptual domains. Review of Educational Research. https://telearn.hal.science/hal-00190680v1
• Means, B., Peters, V., Neisler, J., Wiley, K., & Griffiths, R. (2021). Lessons from remote learning during COVID-19. Digital Promise. https://doi.org/10.51388/20.500.12265/116
• Doğru, M. S., & Faulconer, E. K. (2026). ChatGPT as a virtual laboratory teaching assistant in undergraduate biology. Research in Science Education. https://doi.org/10.1007/s11165-025-10271-z
• Kononowicz, A. A., Woodham, L. A., Edelbring, S., Stathakarou, N., Davies, D., Saxena, N., Tudor Car, L., Carlstedt-Duke, J., Car, J., & Zary, N. (2019). Virtual patient simulations in health professions education: Systematic review and meta-analysis by the Digital Health Education Collaboration. Journal of Medical Internet Research. https://doi.org/10.2196/14676
• Wang, W.-S., Lin, C.-J., Lee, H.-Y., Huang, Y.-M., & Wu, T.-T. (2025). Enhancing self-regulated learning and higher-order thinking skills in virtual reality: the impact of ChatGPT-integrated feedback aids. Education and Information Technologies, 30, 19419–19445. https://doi.org/10.1007/s10639-025-13557-x
• Sellberg, C., & Lindwall, O. (2025). Simulation-based training in professional education: learning, participation, and instructional design. Instructional Science. https://doi.org/10.1007/s11251-025-09763-2
• Ke, F., Dai, Z., Pachman, M., et al. (2021). Exploring multiuser virtual teaching simulation as an alternative learning environment for student instructors. Instructional Science, 49, 831–854. https://doi.org/10.1007/s11251-021-09555-4
• Tao, P.-K., & Gunstone, R. F. (1997). Conceptual change in science through collaborative learning at the computer. Paper presented at the Annual Meeting of the National Association for Research in Science Teaching. https://eric.ed.gov/?id=ED406151
• Nguyen, H., & Cao, J. (2026). Exploring the capacity of large language models to simulate students' scientific thinking: Insights for responsive teaching. Journal of Science Education and Technology. https://doi.org/10.1007/s10956-026-10333-5
• Sitzmann, T. (2011). A meta-analytic examination of the instructional effectiveness of computer-based simulation games. Personnel Psychology, 64, 489–528. doi:10.1111/j.1744-6570.2011.01190.x
• Krüger, J. T., Höffler, T. N., Wahl, M., Knickmeier, K., & Parchmann, I. (2022). Two comparative studies of computer simulations and experiments as learning tools in school and out-of-school education. Instructional Science. https://doi.org/10.1007/s11251-021-09566-1
• Triona, L. M., & Klahr, D. (2003). Point and click or grab and heft: Comparing the influence of physical and virtual instructional materials on elementary school students' ability to design experiments. Cognition and Instruction, 21(2), 149–173. https://doi.org/10.1207/s1532690xci2102_02
• Lai, P. K. (2026). Enhancing learning of nanoscale phenomena: the role of agent-based models and embodied reasoning. Instructional Science. https://doi.org/10.1007/s11251-025-09776-x
• Lazonder, A. W., Wilhelm, P., & van Lieburg, E. (2009). Unraveling the influence of domain knowledge during simulation-based inquiry learning. Instructional Science, 37, 437–451. https://doi.org/10.1007/s11251-008-9055-8
• Liu, C.-C., Wen, C.-T., Chang, H.-Y., Chang, M.-H., Lai, P.-H., Fan Chiang, S.-H., Yang, C.-W., & Hwang, F.-K. (2022). Augmenting the effect of virtual labs with "teacher demonstration" and "student critique" instructional designs to scaffold the development of scientific literacy. Instructional Science, 50, 303–333. https://doi.org/10.1007/s11251-021-09571-4
• Messina Dahlberg, G., & Gustavsson, S. (2025). Keeping worlds apart to put them back together: VET teachers' instructional patterns in simulation-based training. Instructional Science. https://doi.org/10.1007/s11251-025-09714-x
• Mool, A., Schmid, J., Johnston, T., et al. (2026). Using generative AI to simulate patient history-taking in a problem-based learning tutorial: A mixed-methods study. Technology, Knowledge and Learning. https://doi.org/10.1007/s10758-025-09929-4
• Jiang, H., Zhu, D., Chugh, R., Turnbull, D., & Jin, W. (2025). Virtual reality and augmented reality-supported K-12 STEM learning: trends, advantages and challenges. Education and Information Technologies. https://doi.org/10.1007/s10639-024-13210-z
• Albus, P., & Seufert, T. (2023). The modality effect reverses in a virtual reality learning environment and influences cognitive load. Instructional Science. https://doi.org/10.1007/s11251-022-09611-7
• Lawson, A. P., Martella, A. M., LaBonte, K., Delgado, C. Y., Zhao, F., Gluck, J. A., Munns, M. E., LeRoy, A. W., & Mayer, R. E. (2024). Confounded or controlled? A systematic review of media comparison studies involving immersive virtual reality for STEM education. Educational Psychology Review. https://doi.org/10.1007/s10648-024-09908-8

Page metadata

Field	Value
type	principle
id	simulations-immersive-virtual-environments
aliases	vr-visual-experiential-scaffolding-literacy
title	Simulations & Immersive Virtual Environments
description	For a learner preparing for a risky, costly, rare or invisible performance, repeated runs in a simulation that keeps the task's decisions and cues, guided during exploration and followed by feedback or a debrief, are expected to improve performance on the represented task more than no simulated practice, with more immersion adding nothing by itself; claims test simulation against no intervention, passive and active instruction (no advantage over active methods) and guided against free VR exploration, but not the debrief, dose or fidelity.
status	review
canonical	True
generated	by claude/unspecified on 2026-10-19
trust	unverified — no one has confirmed this page's content
sources	27 — akcayir-2017, merchant-2013, shadiev-2020, pilgrim-2016, chernikova-2020, zacharia-2011, lazonder-2016, de-jong-1998, means-2021, dogru-2026, kononowicz-2019, wang-2025 … and 15 more

Frontmatter is the page's machine-readable record — see the Schema & Guide.