Learning Technology · XR & Simulation Research

The Wrong Question in Immersive Learning

Adrian Cowell  ·  December 2025  ·  12 min read

On fidelity, cognitive load, and why the most expensive environments are often built for the wrong learners.

Every institution building XR or simulation has asked it. It sounds rigorous. It gives procurement committees something to benchmark. It is, in almost every meaningful sense, the wrong place to start.

The question is: how realistic should this be?

The more useful question; the one the research has been quietly building toward for more than a decade; is this: what is the learning actually requiring from the student’s brain, and does the production investment serve that?

These questions produce very different decisions. And the gap between them is where most institutional XR investment quietly fails.

Fidelity Is Not a Single Dial

The first mistake in HE commissioning is treating fidelity as a single axis; low to high, cheap to expensive, unsophisticated to impressive. The simulation research is clear that fidelity has at least five distinct dimensions, and conflating them is where most institutional decisions go wrong.^[1,2]

The research finding that consistently surprises institutions: lapses in physical fidelity are better accepted by participants than lapses in conceptual fidelity.^[3] Learners need the scenario to make sense. If it does, they will accept a great deal of artificiality in the visual environment.

This is the mechanism behind what simulation researchers call the fiction contract; the implicit agreement that enables learners to suspend disbelief and engage authentically despite inherent limitations.^[4] The student accepts the representational conventions of the world and projects meaning onto it. A visually abstracted environment with rigorous conceptual and functional fidelity will almost always outperform a visually spectacular environment with logical inconsistencies or shallow decision consequences.

Two Schools, One Unresolved Evidence Base

The clinical XR field has split into two camps, and the honest position is that neither has comprehensively won the evidence argument.

The case for high fidelity is real. Healthcare professionals in cinematic VR studies report that visual and emotional immersion makes training meaningful and memorable; engaging affective learning in ways that create durable memories.^[5] Research on 360° video in health education finds a consistent positive effect on emotional response to the learning climate, which in turn affects motivation.^[6] For pre-clinical students preparing for OSCEs, high-fidelity VR environments demonstrably reduce anxiety and improve confidence.^[7]

But the evidence gap is equally real. As one well-designed mixed methods pilot found: “360° video creates an immersive experience with an associated high-value motivational position; however, this could not be translated to an increase in exam scores.”^[8] Motivation and learning are not the same outcome. Engagement is not retention. Presence is not transfer.

The case for low fidelity rests on older and arguably more robust foundations. The most-cited paper in this space; Norman et al., 2012; remains striking in its directness: high-fidelity simulators have enjoyed increasing popularity despite costs that may approach six figures, yet most commonly, learning from a simulator is compared with a ‘no-intervention’ control group, which fails to clarify the relationship between simulator fidelity and learning.^[1] Twelve years on, this methodological gap has not been fully closed.

The principle that has emerged from the standards literature; the INACSL formulation; is precise: the level of realism should be the minimum required to achieve the learning goals.^[3] Not maximum achievable. Not most impressive at the demo.

The cost reality reinforces this. As fidelity increases, so do constraints on accessibility: specialist developers, high-end hardware, significant QA overhead, and near-total rebuild costs when the curriculum or clinical guidance changes.^[9] High-fidelity production has a short half-life in rapidly evolving subject areas.

The Finding That Should Change Everything

There is one piece of research that should fundamentally alter how HE institutions think about fidelity investment. Most commissioners have not encountered it.

It is the expertise reversal effect.

A 2025 meta-analysis across 60 studies and nearly 6,000 participants found that low prior knowledge learners learn better from high-assistance, scaffolded instruction (effect size d = 0.505), while high prior knowledge learners learn better from low-assistance, more complex instruction (d = −0.428). These effects are robust across a wide variety of contexts.^[10]

The institutional implication is stark. A rich learning environment is effective for experts, but may be contra-productive for novices; demonstrated directly in a comparison of a high-fidelity serious game against a simpler instructional e-module: the game improved complex cognitive skills for residents with limited self-study time, but did not improve the cognitive skills of inexperienced medical students compared to what they learned from the e-module.^[12]

The first-year undergraduate; the learner most likely to be given the immersive experience because they have the least existing knowledge and it makes for the most compelling open-day demonstration; is precisely the learner for whom high physical fidelity is most likely to be counterproductive.

We are routinely building the most expensive environments for the learners who will benefit from them least.

The AAA vs Indie Parallel

The commercial games industry learned something that edtech has been slow to absorb: high production values do not produce superior engagement or retention on their own.

Games like Papers Please, Undertale, Return of the Obra Dinn, and Celeste; pixel art or deliberately abstract; routinely generate deeper emotional investment and sustained engagement than titles costing a hundred times more to produce. The mechanism is identical to the fiction contract in simulation research. Meaning is constructed by the player’s brain, not rendered by the GPU.

The edtech parallel is well-evidenced. Complex game designs can create cognitive overload for learners, and over-reliance on extrinsic motivators such as rewards may hinder deeper learning.^[13] Meanwhile, hands-on and experiential learning tasks, closer to a card game than an Unreal Engine showcase, foster deeper comprehension of subject matter, bringing significant performance improvements compared to traditional lecture-based instruction.^[14]

The game engine cost structure tells the same story. A high-fidelity Unreal Engine 5 environment built to showcase Nanite geometry requires specialist 3D artists, significant QA overhead, and near-total rebuild if curricular content changes significantly. A Unity or Godot lower-fidelity environment can be built by a single developer-educator, iterated rapidly, and deployed to web or Quest without high-end GPU. Unity is effectively free for small projects; Godot is entirely open source, MIT licensed, no royalties.

Faculty ownership of the toolchain is itself a pedagogical asset. An academic who can iterate their own simulation is more valuable than one who submits a change request to an external studio and waits six weeks.

The Institutional Problem Is Structural

The fidelity question is not primarily a pedagogical failure. It is a structural one. Fidelity decisions in HE are rarely made by pedagogists. They are made by innovation leads responding to institutional prestige pressures; by procurement committees impressed by demos; by external grant funders requiring “demonstrable technological ambition”; by industry partners whose commercial interest is in selling high-end tooling.

None of these actors are systematically incentivised to ask whether a simpler, cheaper, more pedagogically targeted approach would produce better learning outcomes.

The hidden failure mode is faculty adoption. Faculty and staff may resist new tools if training is limited or interfaces are complicated, with the result that technology remains underutilised.^[15] A complex, high-fidelity environment built to showcase capabilities that the average faculty member cannot operate, modify, or integrate into their teaching practice is not an educational asset. It is an expensive liability that will be demoed twice and then abandoned.

The EDUCAUSE data captures the accountability gap precisely: 71% of HE technology leaders rate ROI as at least moderately important when evaluating edtech tools; yet fewer than half report that their institution actually measures it.^[15]

Five Questions Before Commissioning

1. Is the visual environment the learning object, or the backdrop?
If students are learning a decision, concept, communication skill, or process; the environment can be abstract. Realism in the backdrop is noise. The closer you get to needing students to read the environment itself (anatomy, spatial navigation, material properties), the more fidelity earns its cost.

2. What is the learner’s stage?
Novices (early undergraduates) benefit from low to medium fidelity, scaffolded and low-complexity. Intermediates can handle medium fidelity with functional emphasis. High fidelity is genuinely justified for postgraduate, CPD, and professional upskilling; where the expertise reversal effect no longer applies.^[10,11]

3. How often will the content change?
High-fidelity production has a short half-life in rapidly evolving subject areas. Faculty ownership of the toolchain; the ability to iterate and maintain the environment without external support; is itself a pedagogical asset.^[2]

4. Can the learning outcome be achieved through abstraction?
Abstraction enables playability, supports generalisation across scenarios, and focuses attention on conceptual relationships unhindered by additional cognitive demands.^[16] Systems thinking, ethical reasoning, resource management, clinical decision-making; all of these can be taught in environments that look nothing like the real world.

5. Who maintains this after the project ends?
The hidden cost of high-fidelity production in HE is not the build. It is the sustainable operation, update cycle, and faculty adoption that never gets funded in the original grant.