DEV Community

Cover image for We Built a Scientifically Designed Online IQ Test (and Explained the Methodology Behind It)
Emir Baycan
Emir Baycan

Posted on

We Built a Scientifically Designed Online IQ Test (and Explained the Methodology Behind It)

Over the past months we built a project called:

https://whats-your-iq.com/

Most online IQ tests are either entertainment quizzes or opaque scoring systems where you answer some questions and receive a number without understanding how it was calculated.

So instead of building another quiz, we focused on building something different:

A transparent intelligence assessment platform where the methodology is fully documented.

You can read the full methodology here:

https://whats-your-iq.com/en/our-iq-test-methodology


The Problem With Most Online IQ Tests

Most online tests use extremely simple scoring models.

Typical approach:

  • count correct answers
  • divide by total questions
  • convert to a number that looks like an IQ score

The issue is that real cognitive assessments do not work this way.

Professional psychometric tests account for:

  • question difficulty
  • guessing probability
  • measurement error
  • reliability
  • response validity

Without these elements, the score has little statistical meaning.


Our Approach

Instead of simple raw scoring, the system uses modern psychometric measurement methods.

The methodology includes:

  • Item Response Theory (IRT 3PL)
  • reliability estimation
  • measurement error modeling
  • response validity checks
  • confidence interval reporting

These frameworks are commonly used in standardized assessments.


Item Response Theory (IRT)

Instead of counting correct answers, the system estimates cognitive ability using Item Response Theory.

IRT models the probability that a person with a certain ability level will answer each question correctly.

Each item includes parameters such as:

  • difficulty
  • discrimination
  • guessing probability

The final ability estimate is calculated using Maximum A Posteriori (MAP) estimation based on the full response pattern.


Reliability of the Test

Reliability is estimated using Cronbach's Alpha.

Current estimate:

α ≈ 0.94
Enter fullscreen mode Exit fullscreen mode

Values above 0.90 indicate very high internal consistency.


Calibrated Question Bank

The test is built on a calibrated item bank using Item Response Theory (IRT).

Current dataset:

  • 73 psychometrically calibrated questions
  • 10,000+ cleaned response samples used for parameter estimation

Responses used for calibration go through a data quality pipeline before being included in the dataset.

The cleaning process removes noisy or invalid responses such as:

  • extremely fast responses indicating random answering
  • inconsistent response patterns
  • incomplete test sessions
  • duplicate or automated submissions

After filtering the dataset, the remaining responses are used to estimate item parameters including:

  • difficulty
  • discrimination
  • guessing probability

As the dataset grows, the item parameters and score norms are recalibrated using the updated response pool, improving the stability and accuracy of the ability estimates.

This continuous calibration process allows the test to become more reliable as more valid responses are collected.


Validity Checks

The system also includes response validity indicators.

Examples include:

  • person-fit analysis
  • response pattern detection
  • rapid guessing detection
  • response speed analysis

These checks help detect invalid response behavior.


Cognitive Domains Measured

The assessment evaluates multiple areas of reasoning.

These include:

  • logical reasoning
  • spatial intelligence
  • verbal comprehension
  • working memory
  • processing speed
  • quantitative reasoning
  • pattern recognition
  • creative reasoning

Score Interpretation

Scores follow the standard IQ distribution model:

Mean = 100
Standard deviation = 15
Enter fullscreen mode Exit fullscreen mode

Percentiles are estimated using this theoretical distribution.


Important Transparency Note

The test follows modern psychometric principles but it is not a clinical psychological assessment.

It does not replace professionally administered tests such as:

  • WAIS
  • Stanford-Binet

Those require administration by licensed psychologists.


Why We Published the Full Methodology

Most online IQ tests hide their scoring methods.

We chose to publish ours openly so users can understand:

  • how the score is calculated
  • what the test measures
  • what the limitations are

Full methodology:

https://whats-your-iq.com/en/our-iq-test-methodology


Try the Test

If you're curious about the system, you can try it here:

https://whats-your-iq.com/

Top comments (0)