Next-Generation Exposure Therapy

The project takes a mechanism-first approach. Rather than treating VR and AI as solutions in themselves, it asks what should be personalised in exposure therapy, why it should be personalised, and whether technology can support this work without displacing therapist judgement or patient agency.

Research focus

Research areas: Virtual reality, exposure therapy, fear learning, artificial intelligence, personalisation, human-centred design.

The project examines personalisation at two connected levels.

Mechanisms of exposure and learning

The first level concerns the psychological processes that may shape exposure therapy, including:

• Expectations about the probability and cost of feared outcomes;

• Coping beliefs and self-efficacy;

• Intolerance of uncertainty;

• Safety behaviours and avoidance;

• Learning through expectancy violation and the development of new safety-related associations.

This work is informed by contemporary theories of exposure and extinction learning, particularly inhibitory learning theory. In this framework, exposure is not understood as erasing fear memories. Instead, it supports new learning that a feared cue or situation can be encountered without the anticipated harmful outcome.

Environments and technologies

The second level concerns the design of exposure environments, including:

• The contexts and cues presented during exposure;

• The degree of interactivity and user agency in VR;

• The variation of stimuli and situations across sessions;

• The role of AI in decision support, feedback and adaptive scenario design;

• The relationship between VR-based exposure and real-world, in vivo exposure.

VR can make it possible to systematically vary these features while retaining a high level of experimental and therapeutic control. However, the project also considers the limits of VR: a virtual environment cannot automatically reproduce the complexity, variability or social meaning of real-world exposure.

AI as decision support—not replacement

A core principle of the project is that AI should support therapists and patients rather than replace them.

AI systems may eventually assist with tasks such as organising data from therapy sessions, identifying patterns in responses, supporting scenario configuration and presenting understandable recommendations. Yet these systems must be developed with careful attention to safety, fairness, transparency and clinical responsibility.

The project therefore promotes a stakeholder-in-the-loop approach, in which patients, therapists, researchers, engineers, designers and ethics experts contribute to the development and evaluation of AI-enabled VRET.

Key requirements include:

• Clear and interpretable recommendations;

• Clinician oversight and the ability to override system suggestions;

• Attention to potential bias and unequal performance across groups;

• Strong data-protection and governance practices;

• Evaluation of clinical value, rather than technical performance alone;

• Co-design with people who may use or be affected by these systems.

Research programme

The project develops through three connected strands of work.

1. What should be personalised in VRET?

A planned systematic review will examine which elements of exposure therapy can meaningfully be personalised, and what evidence supports different approaches.

The review will distinguish between:

• Mechanism-level personalisation, such as tailoring work on expectancies, coping, uncertainty and safety behaviours; and

• Environment-level personalisation, such as adapting contexts, stimuli, interaction and scenario structure.

The review will follow preregistered and transparent methods, including PRISMA 2020 reporting and risk-of-bias assessment. Where evidence is sufficiently comparable, meta-analysis will be considered. It will also examine the relationship between variability in VR environments and variability in real-world exposure, including implications for generalisation beyond virtual settings.

2. Artificial intelligence for virtual reality exposure therapy

A systematic review by Kamilla Bergsnev and Ana Luisa Sánchez Laws maps current uses of AI in VRET.

The review identifies three broad categories of AI application:

• Machine learning, including state classification and outcome prediction;

• Conversational AI, which may support interaction, guidance and engagement;

• Knowledge-based and hybrid systems, which can support therapist decision-making and adaptive scenario configuration.

The review concludes that AI for VRET remains an emerging field. While there are promising technical and clinical examples, future research needs stronger integration with exposure theory, larger and more diverse samples, external validation, explainability and meaningful stakeholder involvement.

3. Personalised VR exposure and personalised in vivo exposure

A future empirical study will compare personalised VR exposure with personalised in vivo exposure.

The study is designed to address a central translational question: what does VR add when the therapeutic targets are held constant?

Both study arms will use the same mechanism-level personalisation targets and decision rules, including probability and cost expectancies, coping, uncertainty and safety behaviours. Therapist training will also be aligned across conditions. The exposure modality—VR or real-world exposure—will be the main systematic difference.

Planned outcomes include behavioural approach, disorder-specific symptoms and follow-up measures of durability. The study will also assess implementation factors, including therapist time, personalisation fidelity, credibility and expectancy, safety behaviours, presence in VR and cybersickness.

Role of EDGE Lab

For EDGE Lab, this project contributes to research at the intersection of immersive technology, psychology, ethics and human-centred design.

It explores VR not only as a platform for delivering digital interventions, but as a tool for examining how environments, interactions and technological systems shape learning and behaviour. The project also foregrounds the ethical and practical questions that arise when AI systems are introduced into sensitive areas of mental-health care.

The work is guided by the principle that technological innovation should remain accountable to the people it is intended to support.

Open science and responsible research

The project is committed to transparent and reproducible research practices. Planned work will use preregistration, clear reporting standards, open code and ethically appropriate data-sharing practices where possible.

Future empirical studies will be designed and reported using relevant guidance, including CONSORT-EHEALTH and TIDieR, to support clear description, replication and critical evaluation of digital interventions.

Publications

Bergsnev, K., & Sánchez Laws, A. L. (2022). Personalizing Virtual Reality for the Research and Treatment of Fear-Related Disorders: A Mini Review. Frontiers in Virtual Reality, 3, 834004.
https://doi.org/10.3389/frvir.2022.834004

Bergsnev, K., & Sánchez Laws, A. L. (2026). Artificial Intelligence (AI) for Virtual Reality Exposure Therapy (VRET): A Systematic Review. Translational Psychiatry. https://www.nature.com/articles/s41398-026-03936-4

UiT The Arctic University of Norway

EDGE Lab