Behavioural measurement for spatial computing.
PeakXR sits where cognitive science, psychometrics and machine learning meet. We build the scenes, the models and the evidence behind them, and we treat each assessment as an instrument that has to earn its validity.
What we work on.
Six lines of work run in parallel. Each one feeds directly into the scenarios and scoring models used in client assessments.
Behavioural signal modelling
Turning raw interaction streams (movement, timing, object use, dialogue) into stable features that describe how someone approaches a task, not just whether they finish it.
Conversational virtual characters
Characters that listen and respond, so socio-emotional scenarios behave like real conversations: a colleague who pushes back, a client who changes their mind.
Psychometric validity of VR tasks
Reliability, construct validity and population norms for every metric. A score only goes into a report once we can show what it measures and how stable it is.
Adaptive scenario design
Scenes that adjust difficulty and pacing to the participant, so the assessment spends its time where the measurement is most informative.
Grounded report generation
Expert-review text generated from the scores themselves and checked against them, so every sentence in a report is tied to a measured result.
Privacy by design
Collecting only the signals a metric needs, separating identity from session data, and keeping individual results visible only to the participant unless agreed otherwise.
What we are still figuring out.
- How much of a socio-emotional competency can be inferred from a single conversation with a virtual character?
- Which behavioural features stay stable when the same person repeats a scenario weeks later?
- How short can an assessment get before measurement precision drops for leadership-level roles?
- How do scores from a spatial scene compare with outcomes on the job a year later?
- Where is the line between helpful adaptivity and a test that is no longer the same for everyone?
Four disciplines, one instrument.
Assessment design in a headset needs people who rarely share a room. Ours do.
Cognitive science
Constructs, task design and the link between behaviour and competency.
Psychometrics
Scoring models, norms, reliability and validation studies.
Machine learning
Feature pipelines, scoring models and report generation.
Spatial engineering
visionOS development, 3D scenes and real-time interaction.
Research partnerships.
We work with universities, HR research groups and companies who want to run validation studies with us.