For policing programme leads and force trainers comparing simulation approaches
Picture a Tuesday afternoon at a regional training centre. Eight student officers are queuing outside a converted classroom, waiting their turn in a stop-and-search roleplay. Two actors are running through the same scenario for the fourth time that day, adjusting tone and resistance for each learner, while a facilitator keeps one eye on the clock and the next cohort. Down the corridor, a Hydra suite is booked solid for a two-day command exercise: six syndicate rooms, a control desk, and a facilitator team who spent the morning resetting scenario data between groups.
Both sessions matter. Both are exactly the kind of practice the College of Policing wants officers to have. And both are, by design, rationed — by room bookings, by actor availability, by how many learners can realistically get a turn before the training day ends.
That's the quiet tension behind scenario-based police training: almost everyone agrees it works. Almost nobody has enough of it to go round.
Scenario-based practice isn't a nice-to-have bolted onto firearms and restraint training — for England and Wales, it's mandatory. Since April 2024, every force has had to deliver at least 12 hours a year of public and personal safety training (PPST) to every officer, built explicitly around scenarios like stop and search, dealing with a vulnerable person, and working in custody, with what the College of Policing describes as "a strong focus on de-escalation and communication tactics."
The research behind that mandate is some of the strongest causal evidence policing has produced. A randomised controlled trial run by King's College London with the College of Policing, covering 1,843 officers in Avon and Somerset and published in the Justice Evaluation Journal in 2024, found that scenario-based safety training led to a 10.9% decrease in the likelihood of an officer using force in a given week. Hands-on force fell by 14%. Injuries to members of the public dropped by a third. Ninety-two per cent of officers rated the training "very satisfied" or "satisfied."
That's a rare thing in policing research: a large, controlled study showing that a specific training method changes what happens on the street. It's also why the debate among programme leads and force trainers isn't whether scenario-based training works. It's whether the two methods most forces currently lean on to deliver it — live actor-led roleplay and Hydra-style immersive syndicate suites — can be delivered often enough, to enough people, for that effect to hold across a whole force or cohort.
It's worth sitting with those numbers for a moment, because they're doing more work than a single "training improves outcomes" headline suggests. A double-digit fall in the weekly likelihood of using force, across a sample of nearly 1,900 officers, points to a training method with a genuine behavioural effect — not just a satisfaction score. The steepest reduction was in hands-on force specifically, the category most directly shaped by how an officer opens and manages a conversation before physical contact becomes necessary.
Findings from a randomised controlled trial of 1,843 officers, Avon and Somerset Constabulary.
Live actor-led roleplay and Hydra suites earned their place in the curriculum for good reason. They're high-fidelity, they're well understood by trainers, and — in Hydra's case — they've been purpose-built for exactly the kind of multi-agency, high-pressure decision-making that policing needs to rehearse. Neither is going anywhere, nor should it. But each comes with a resourcing profile that caps how much practice any one officer actually gets.
Live roleplay relies on trained actors, a facilitator, a booked space and a script that can flex to each learner. A 2023 peer-reviewed comparison of VR and live-action de-escalation training, published in Policing: A Journal of Policy and Practice, notes plainly that live-action scenario training "can be labour intensive and logistically challenging to stage consistently at scale."
That's not a criticism of the method — it's a description of what it costs to run well. Every extra rep for a learner means another slot with an actor and a room, which is why most forces can only offer a handful of live roleplay sessions per officer per year rather than the frequent, low-stakes repetition that skill-building actually needs.
Hydra and Hydra-Minerva suites — the immersive syndicate methodology originally developed within the Metropolitan Police and now stewarded by the Hydra Foundation — sit at the more intensive end of the spectrum. They're a permanent installation across four to six rooms, with a control centre and trained facilitators (the College of Policing runs its own dedicated facilitators' course).
One UK university disclosed capital investment of more than £350,000 for its Hydra/Minerva facility; its own materials state that "between one and thirty students can use the facility at any given time," and College of Policing exercises can run for one to five days. That's an outstanding environment for command-level and critical-incident training. It's also, inevitably, a facility you have to book, staff and travel to — which limits how often any individual learner cycles back through it.
The pattern across both methods is the same: quality is high, and access is the constraint. HMICFRS has separately flagged the broader consequence of that constraint — in its 2020 roads policing inspection, it found that because College training modules "aren't mandatory," forces "developed their own approaches," resulting in inconsistency in how, when and to what level officers were actually trained. Capacity limits don't just mean less practice. Left unmanaged, they mean uneven practice across a force.
None of this is an argument against roleplay or Hydra. It's an argument for taking seriously what learning science already tells us about repetition. Memory research going back to Ebbinghaus, and confirmed in more recent work, shows that skills decay rapidly without spaced review — distributed practice across multiple sessions, rather than one intensive block, is what makes retention hold. A 2022 study in Memory & Cognition found that at least a day needs to pass between repetitions to maximise long-term retention, and that spacing sessions out substantially slows the rate of decay.
That's the gap AI-powered VR roleplay is best placed to fill — not as a replacement for Hydra or live actors, but as the high-frequency layer underneath them. Comparable research bears this out.
A 2023 randomised study from Wilfrid Laurier University in Canada compared VR and live-action mental health crisis de-escalation training against a no-training control group, and found "no reliable difference between the LA and VR modalities" in de-escalation competency gains — both outperformed the control, and VR training produced a greater reduction in bias toward mental illness, without adding cognitive load. In other words: comparable skill gains, at a fraction of the resourcing cost per repetition.
A healthcare training study, shown here as an adjacent-field illustration of the repeat-practice effect — not a policing-specific finding.
Cost and throughput evidence tells a similar story, albeit from an adjacent field rather than policing itself. A 2023 study in BMC Nursing, covering 675 pre-registration nursing students, found that 95% of students actively participated when training used VR simulation, compared with an average of 15% in traditional simulation-based education — alongside lower per-student cost.
It's not a policing study, so it shouldn't be read as a direct substitute for policing-specific evidence. But it illustrates the underlying mechanism: when a learner can pick up a headset and run a scenario again without booking a room, a facilitator or an actor, they simply get more reps.
UK forces are already treating VR as a genuine complement rather than a novelty. Gwent Police was the first UK force to build immersive VR domestic abuse training, back in 2018. Greater Manchester Police has built VR hate-crime empathy scenarios.
The Metropolitan Police piloted VR domestic abuse training with 48 frontline officers, more than three-quarters of whom found it more engaging than other training formats. North Wales Police funded VR coercive-control training for student officers in 2023. The College of Policing itself is running active research projects on immersive learning and VR frameworks for UK police training — a signal that this is being treated as a serious pedagogical question, not a gimmick.
It would be overselling the case to say VR and AI-driven roleplay are proven equivalents to Hydra or live roleplay across the board. A 2024 critical review of extended-reality police training, authored by researchers across UCL, Glasgow and Lancaster, is direct about this: "empirical evidence of benefits compared to real-life training and of skill transfer to real life are still scarce."
That's a fair challenge, and one worth repeating to any vendor who claims otherwise. The honest, evidence-based position is that VR and AI roleplay currently do two things well and provable: they match live-action training on measured skill gains in the studies that exist, and they multiply how often a learner can practise. What they haven't yet been proven to do, at scale, is fully replicate the fidelity of a multi-agency Hydra exercise or the improvisational pressure of a skilled live actor.
That's precisely why "complement, not replacement" is the right frame — not a marketing softener, but the conclusion the evidence currently supports.
For programme leads mapping methods against budget and timetable, the practical shape of this is reasonably clear:
None of this requires choosing a side. It requires being honest about what each method is resourced to deliver, and building a training calendar that uses each one where its strengths actually apply.
Scenario-based training works — the evidence on that point is about as strong as policing research gets. The open question for most forces and PEQF programmes isn't whether to keep investing in Hydra suites and roleplay; it's how to make sure officers get enough repeated practice between those sessions to actually hold onto the skills. That's a capacity problem as much as a pedagogical one, and it's the reason more forces and universities are starting to treat AI-powered VR roleplay as a practical addition to the toolkit, rather than a competitor to it.
Bodyswaps' Policing & Justice Studies ILRP supports programmes with exactly this layer: repeatable, AI-driven roleplay scenarios that give students structured practice between live sessions, with the same behavioural feedback and evidence trail that quality assurance processes look for.
Explore our Policing & Justice Studies ILRP →
It's training where officers practise real situations, like a stop and search or calming down an angry member of the public, instead of just learning rules. The College of Policing says every officer in England and Wales must get at least 12 hours of this training each year.
A Hydra suite is a set of rooms built for group exercises. Officers work through a fast-moving incident together while trainers change the situation in real time. It's mostly used for command training and big, multi-team incidents.
Research so far says yes, on the skills that have been measured. A 2023 Canadian study found no real difference in de-escalation skill gains between VR and live-actor training. But experts still say more research is needed on how well VR skills carry over into real situations.
Every officer in England and Wales must get at least 12 hours of public and personal safety training a year. This is scenario-based and focuses heavily on de-escalation and communication.