Book a demo

Rigour first. AI second.

For every AI agent, every scoring model, every assessment framework we ship, the science comes first. Ten years of applied IO psychology research at The Talent Games, three million candidate assessments in the underlying dataset, and a research team that publishes.

Book a demo
See the four pillars

We started as psychometricians. We stayed that way. AI didn’t change what we measure. It changed how many candidates we can measure it for.

C-Factor AI is built on assessment research developed through collaboration between assessment psychologists, I/O expertise, subject-matter experts, and data scientists.

10+

Years of applied IO research

3M+

Candidate assessments in our data foundation

30+

Peer-reviewed studies & internal research reports

5+

IO psychologists on the research team

THE FOUR SCIENTIFIC PILLARS

The science behind every question we ask.

Four disciplines, layered. IO psychology gives us what to measure. Psychometrics gives us how to measure it. Behavioral science gives us how to read the signals. Continuous validation makes sure we keep being right.

 
  • PILLAR 01

Industrial-Organizational

Psychology

The foundation. Every framework we use is grounded in decades of applied research on workplace behavior, cognitive ability, and job performance not on what’s fashionable in AI this quarter.

VALIDATED FRAMEWORKS

Competency models rooted in classical IO frameworks

Our competency taxonomies build on Big Five personality research, cognitive ability constructs, and situational judgement methodology. We don’t invent competencies. We map validated ones onto specific role families.

IN-HOUSE IO REVIEW

Every framework reviewed by our IO psychology team

Before any role template ships into the platform, our MSc- and PhD-level IO psychologists review it for construct validity. If a framework can’t explain itself to a psychometrician, we don’t deploy it to a customer.

RESEARCH-LED UPDATES

Frameworks evolve with the research

Workplace research doesn’t stand still. As new meta-analyses land and new competencies emerge in the literature, we update our frameworks. Customers on active deployments get informed when the underlying science shifts.

PUBLISHED WORK

Our psychologists publish

Members of our research team contribute to peer-reviewed IO psychology and assessment journals. When our practitioners defend their methodology in academic forums, our customers benefit from that scrutiny.
  • PILLAR 02

Psychometric Rigor

Every score claims to measure something. Psychometrics is what proves the claim. Construct validity, predictive validity, reliability, fairness measurable properties of every model we ship.

CONSTRUCT VALIDITY

Measuring what we claim to measure

Every competency score is validated against convergent and discriminant constructs. A “structured thinking” score correlates with related cognitive measures and diverges from unrelated ones. This isn’t optional. It’s how we know the score means what we say it means.

PREDICTIVE VALIDITY

Scores that correlate with real-world performance

Where customers permit, we track post-hire performance against pre-hire scores. When correlations weaken, we investigate. This is the only test that ultimately matters do our scores predict who does the job well?

RELIABILITY TESTING

Cronbach's alpha and inter-rater reliability

A score that varies wildly across administrations isn’t measuring a stable trait it’s measuring noise. We test internal consistency, test-retest reliability, and inter-rater reliability on every scored competency.

DIF ANALYSIS

Differential Item Functioning across protected classes

An item that works well for one demographic and poorly for another is a fairness problem regardless of intent. DIF testing surfaces these before deployment. Any item that shows meaningful DIF gets redesigned or dropped.
  • PILLAR 03

Behavioral & Response Science

The signals inside candidate responses are as important as the answers themselves. Extracting those signals from language, structure, timing, choice patterns is a discipline of its own.

STAR METHODOLOGY

Structured behavioral event interviewing

Our AI interview agents probe for Situation, Task, Action, and Result the same structured framework used by trained behavioral interviewers. When a response is thin on any of the four, follow-ups are triggered.

MULTI-SIGNAL AGGREGATION

Never a single-point score

Every competency score aggregates multiple independent signals: content of the response, structural completeness, specificity of examples, coherence over time. A single confident answer never dominates the score.

ADVERSARIAL TESTING

Gaming resistance under red-team testing

Every scoring model faces adversarial testing before deployment prompt-hacking attempts, coached responses, and pre-written answers. Models that can be gamed get rebuilt until they can’t.

CROSS-CULTURAL CALIBRATION

Signals calibrated per language and region

Response patterns vary by language, cultural norm, and communication style. A directness that reads as confident in one region reads as blunt in another. Every model is calibrated per language and validated separately across regions.
  • PILLAR 04

Continuous Validation

Science that stops testing itself stops being science. Every deployed model is watched, audited, and re-tested on a fixed cadence with independent review at the year mark.

QUARTERLY AUDITS

Bias, drift, and validity re-tested every quarter

Every deployed scoring model gets four full re-audits a year. We check for model drift, adverse-impact drift, and construct-validity drift. Where anything moves outside tolerance, the model is retrained or retired.

80% RULE + COHEN'S D

Adverse-impact monitoring on protected classes

We run the four-fifths rule and effect-size tests across gender, ethnicity, age, disability status, and other protected attributes (where legal to measure). Reports available to enterprise customers under NDA.

POST-HIRE OUTCOMES

Scores tracked against real-world performance

Where customers permit, pre-hire scores are correlated with post-hire performance ratings, tenure, and promotion data. This is the ground-truth signal that keeps our models honest.

EXTERNAL PEER REVIEW

Independent IO psychologists review our methodology annually

External researchers with no commercial interest in our success review our methods, our fairness reports, and our validity claims once a year. Findings drive our roadmap. Not our marketing.

Our Research Library

Ongoing research programmes.

Our research team is actively exploring six areas shaping talent assessment and AI. We study emerging approaches to improve how organizations evaluate and understand talent.

Predictive Validity of Gamified Assessments

Cross-industry studies correlating our assessment scores with post-hire performance ratings. Aggregate customer-deployment data updates the analysis quarterly.

AI Interviews vs Traditional Panels

Comparative studies on how AI-conducted structured interviews measure up to traditional panel interviews on validity, candidate experience, and adverse-impact metrics.

Cross-Cultural Fairness Analysis

DIF (Differential Item Functioning) analysis across 15+ countries and four language families. Findings directly inform our cross-cultural calibration methodology.

Adaptive Follow-up Effectiveness

Research on how adaptive STAR-based follow-ups contribute to overall scoring accuracy versus static question sets. Underpins Noor's and Aly's adaptive probing.

Regional Competency Calibration

Regional validity studies of Big Five-based competency frameworks across MENA, APAC, European, and Americas candidate populations. Informs regional model calibration.

Explainability & Hiring-Team Trust

User research measuring how score explainability affects hiring team trust, override rates, and candidate appeal outcomes. Underpins our explainability-first product philosophy.

THE PEOPLE BEHIND THE SCIENCE

A research team big enough to defend every claim we make and small enough that every member has read every model we’ve shipped.

10+

IO psychologists on staff, MSc and PhD-level

4

Continents represented in the research team

Annual

External peer review by independent IO researchers

Every

Model reviewed before deployment. No exceptions.

A NOTE FROM OUR CEO

We didn't start with AI. We started with fifteen years of IO psychology research and three million candidate assessments. When AI arrived, we brought all of that with us. Every model we ship has to convince our psychometricians before it goes anywhere near a customer. Assessment science is the foundation. AI is the amplifier not the other way around.
Paul Keijzer
CEO · The Talent Games · Engage Consulting

A NOTE FROM OUR CEO

Paul Keijzer

CEO · The Talent Games 

We didn’t start with AI. We started with fifteen years of IO psychology research and three million candidate assessments. When AI arrived, we brought all of that with us. Every model we ship has to convince our psychometricians before it goes anywhere near a customer. Assessment science is the foundation. AI is the amplifier not the other way around.

A NOTE FROM OUR CEO

Every AI decision in hiring is a small verdict on someone's future. When we built C-Factor AI, we started from that fact. If we couldn't explain a score to the candidate who was given it, we didn't ship it. If we couldn't audit that decision six months later, we didn't ship it. If a hiring manager couldn't override the AI, we didn't ship it. Trust isn't a feature we bolt on. It's the reason we exist.
Paul Keijzer
CEO · The Talent Games · Engage Consulting
FAQ

Questions on the science.

Fifteen-plus years of applied IO psychology research at The Talent Games, three million candidate assessments in our data foundation, and a team of MSc- and PhD-level IO psychologists who continue publishing in peer-reviewed forums. Our competency frameworks are grounded in Big Five personality research, cognitive ability constructs, and situational judgement methodology.
Where customers permit, we track post-hire performance ratings against pre-hire scores. When correlations weaken over time, we investigate and retrain. This is our primary ground-truth signal. Customers with mature deployments contribute to our ongoing predictive validity research programme.
Every scoring model is calibrated per language and validated separately across regions. Response patterns that read as confident in one culture can read as blunt in another scoring that ignores this bakes in cultural bias. Our cross-cultural calibration is grounded in ongoing DIF (Differential Item Functioning) analysis across 15+ countries.
Most AI-first hiring platforms start with a language model and reverse-engineer an assessment claim onto it. We started with fifteen years of IO psychology research and layered AI on top of validated methodology. The order matters our AI amplifies proven assessment science rather than trying to synthesise it from language patterns.
Summaries of our research programmes are available on request to enterprise buyers. Full methodology notes, bias reports, and predictive validity studies are available to enterprise customers under NDA. Book a demo and we’ll walk your team through the specific research relevant to your industry.
Every deployed model faces external peer review annually from independent IO psychologists with no commercial interest in our success. Their findings drive our roadmap not our marketing. Enterprise customers can request access to their review summaries under NDA.
Still have a question?