How the observations are made, and where their limits are.
Recruitment decisions need to be explainable. So every result is calculated first, then explained in words, and always read by a person who makes the decision.
- LEVEL 1Each gameThe game result, time spent and retries, turned into a score per skill with a written reason.
- LEVEL 2Each group of gamesPatterns across related games, so one unusual round does not decide everything.
- LEVEL 3The whole vacancyA summary for this role, rolled up into the six areas, with questions to ask in the interview.
How a result is made
Calculated first, explained second
Scores are computed by the system before the AI writes anything. The AI explains numbers that already exist; it cannot invent data.
Consistent
Identical results get the same evaluation. Evaluations are cached by their inputs, so a repeat returns the same answer instantly.
Explained, not hidden
Every score comes with a written reason that refers to what happened in the game. No black-box numbers.
Support, not a verdict
Reviews never say yes or no to a candidate. They give your team observations and questions; people make the decision.
The framework behind it
Our framework is called the Assess4me Integrated Model (AIM). It describes which game actions count as evidence for which of the six areas, and how that evidence is combined.
AIM draws on established approaches from assessment research: evidence-centred design, stealth assessment (measuring while someone plays rather than asking them about themselves) and computational psychometrics.
It looks at how someone plays, not only the outcome: the order and timing of choices, accuracy, and how they adjust between rounds. Two candidates can reach the same score in very different ways, and the review shows that.
Drawing on research is not the same as being validated. We say clearly what is and is not proven yet, below.
What we know, and what we're still learning
We'd rather be clear about the limits than oversell.
Talk to us about a pilot or research- What we observe
- How candidates play: choices, timing, accuracy and how they adjust between rounds, mapped to six soft-skill areas.
- How it should be used
- As extra observations and interview hypotheses next to the CV and the conversation. Never as the only basis for a decision.
- What is not proven
- Predictive validity has not yet been demonstrated. We are researching which links between play and work behaviour are strongest.
- Who we work with
- Hiring teams willing to run a responsible pilot, and HR tech and research partners.
Data and privacy
- No email addresses to the AIEvaluations receive results under an internal identifier, not candidates’ email addresses.
- EncryptedData is encrypted in transit and at rest.
- GDPRA data processing agreement is available for organisations on request.
- Candidates’ rightsCandidates can ask for the data held about them, and for it to be deleted.
- FastA full review at three levels usually completes within seconds of a candidate finishing.
Read the full details in our privacy policy and terms.