We do not ask — we observe

A questionnaire asks people to describe themselves, and that is exactly what it gets: a description. People answer as they see themselves, and as they would like to be seen. In a hiring context there is the obvious added incentive to present the best possible version.

NeuroFrame works differently. The person plays a simulation game and makes dozens of decisions inside it: where to spend a limited resource, when to take a risk, what to do once the plan stops working. Throughout, the system records actions rather than answers.

What is measured is behaviour, not self-description. Playing to the test is close to impossible: the person does not know which behavioural signals are read, or how they add up. They are occupied by the task, not by the assessment.

Format: desktop or mobile device, 30–60 minutes, no observer present.

Four steps: from play to profile

Between the first click and the finished profile there are four steps. At none of them does the person answer questions about themselves.

The output is expressed in percentiles, showing where a person sits relative to the normative sample. The profile is then compared against the requirements of a specific role — not against other candidates in terms of better or worse.

How the processing works

• The simulation. The person becomes absorbed in the game and stops holding the assessment in mind. In the background the system collects thousands of digital traces: choices, pauses, reaction times, decisions revisited and revised.
• Processing. The stream of actions is turned into structure: what was chosen, what was passed over, how close the decisions came to optimal. Forgone options count too — what economists call opportunity cost.
• Comparison with norms. Behaviour is matched against the reference base: 14,850 people in real jobs and entrepreneurs from 500+ companies, 20+ industries and 25+ functional areas.
• The profile. A percentile score across eight parameters, usable for hiring, development, succession planning and team composition.

What kind of game it is

The genre is tower defence: hold a route along which waves of opponents advance, each wave harder than the last.

Behind the game shell sits a formal problem: dynamic allocation of limited resources under time pressure and rising load. It is the same problem an engineer faces at handover, an analyst faces before a deadline, and a manager faces with one budget and three directions to fund.

What actually produces a strong result is not published: the reference map of optimal play and the feature weights are closed. This is not secrecy for its own sake. If the key is known in advance, an implicit measure becomes a trainable skill, and the reference base loses its value for everyone assessed afterwards. Protecting the scoring key is a requirement of the professional assessment standards (AERA/APA/NCME; ITC Guidelines on Security of Tests).

Design principles

• Graduated difficulty. Levels get harder, so the assessment sees not only the current level of performance but the trajectory — that is, learnability.
• Rules are stated up front. Nothing is guesswork: what is measured is not knowledge of the rules but the quality with which known rules are applied under mounting load.
• An unfamiliar way of presenting the rules. How a person masters something unfamiliar — how often they return to the instructions, how quickly they grasp them — is itself informative.
• Incomplete information. Some properties of the environment have to be discovered along the way, as in real work.

The evidence model: from behaviour to parameter

The method is built in the Evidence-Centered Design paradigm. ECD does not permit the claim that a game measures leadership and asks you to take it on trust. It requires an explicit three-link chain: the quality being measured → the observable behaviour that serves as evidence of it → a task deliberately constructed so that the behaviour can appear.

The chain is built backwards. First, what we want to measure. Then, what observable behaviour would count as evidence for it. Only then is the episode designed in which that behaviour might emerge — or fail to. The absence of an expected behaviour is evidence too.

A few examples follow. They are deliberately stated at the level of a type of behaviour, never a specific move in the game.

The full evidence model — all eight parameters, the signals and the logic by which they are captured — is disclosed to the client's methodologists and decision-makers under a confidentiality agreement. Transparency towards the client and security of the key against those being assessed are different requirements, and they do not conflict.

Example links

• A person masters unfamiliar rules quickly, moving from trial attempts to consistently sound decisions — evidence of learnability.
• Decision quality holds up as load increases, and after a setback the approach changes rather than the effort stopping — evidence of persistence.
• A workable solution is reached with a minimum of superfluous actions — evidence of mental efficiency.

What the method rests on

NeuroFrame does not invent its own theory of personality. It takes established models and uses them as a frame for interpreting behaviour, rather than as a source of questionnaire items.

The underlying models

• The Big Five — an empirical model of personality developed since the 1930s. Here it serves as an interpretative frame for analysing behaviour, not as a questionnaire.
• Reinforcement Sensitivity Theory (RST, Jeffrey Gray) — a neuropsychological model of how sensitive a person is to reward, and how sensitive to punishment and uncertainty. Risk appetite is derived from it as the balance of two systems, not as abstract boldness.
• The PEN model (Hans Eysenck) — one of the oldest biologically grounded models of temperament; it helps decompose behaviour into stable dimensions of emotional stability and social activity.
• Cybernetic Big Five Theory (CB5T) — treats personality traits as parameters of a behavioural control system.
• Evidence-Centered Design — the engineering methodology of the assessment itself: how an in-game event becomes evidence for a psychological construct.
• Stealth assessment (Shute, 2011) — the concept of evidence-based assessment embedded in play without interrupting it.

Why this is sounder than a questionnaire

The point is not that questionnaires are bad. The point is that self-report carries built-in limits which behavioural observation sidesteps.

Four reasons

• The measure is implicit. The person is solving a task, not describing themselves. Socially desirable distortion is markedly lower than in self-report: candidates scoring systematically higher than incumbent employees is a robustly replicated effect in questionnaires, particularly on conscientiousness and emotional stability.
• Ecological validity. The conditions resemble real work: incomplete information, limited resources, rising pressure. That is closer to the job than agreeing or disagreeing with a statement.
• Potential, not only present competence. The trajectory through the game shows capacity to grow — something a static snapshot of answers cannot give at all.
• Exposure control. The operational key is closed to those being assessed, so coaching to the score does not work.

What the method does not do

Honest limits matter more than broad promises — not least because the client will test them in practice anyway.

The limits

• It does not make psychological diagnoses and does not detect illness.
• It does not measure knowledge, experience or professional skills — only cognitive and personality characteristics.
• It does not replace interviews, reference checks or managerial judgement.
• It does not return a hire or no-hire verdict: it sets out an observation, its implication and the available options.
• It does not guarantee any individual's success — it raises the probability of a sound decision and lowers risk.