Skip to main content

How Param AI Measures

TermsPrivacyDisclaimerAcceptable UseRefunds

How Param AI Measures

Last Updated: August 27, 2026

We publish our method because a career recommendation you cannot interrogate is not advice. If a report tells your child to pick Science, you are entitled to know what produced that sentence.

This page describes what we measure, how, and — just as importantly — what we do not claim.

What we measure

Every assessment, at every class level, scores the same ten traits:

  • Technical — comfort with mechanisms, tools and hands-on systems
  • Analytical — reasoning from evidence, breaking problems apart
  • Social — orientation toward people, teaching, care and persuasion
  • Leadership — willingness to direct, decide and take responsibility
  • Stability — preference for predictable structures over volatility
  • Creative — generating original work rather than refining existing work
  • Independence — preference for self-directed over supervised work
  • Structure — need for defined process and clear expectations
  • Resilience — response to setbacks, criticism and sustained pressure
  • Precision — tolerance for detail, accuracy and repetition

    • The question bank draws on established instruments — RIASEC / Holland Code for interest orientation and Big Five (OCEAN) constructs for personality — adapted to the Indian school context.

    How the questions adapt

    The assessment is adaptive. It does not serve a fixed list.

    After each answer we re-estimate where you sit on every trait, along with how *uncertain* that estimate still is. The next question is then chosen to maximise Fisher information — plain English: we serve the question whose answer we can least predict, because that is the one that teaches us the most. Items are modelled under 2-parameter and 3-parameter item response theory, so a sharply discriminating question counts for more than a vague one.

    The practical effect is that most students answer between 20 and 60 questions rather than a couple of hundred, and the questions get more specific as the session goes on. A student whose analytical score settles early will stop being asked about it.

    Two consequences worth knowing:

  • Two students can see completely different questions. That is the design working, not an error.
  • Stopping early is a confidence decision, not a shortcut. The session ends when the estimates are precise enough, or at the cap.

    • How a recommendation is built

    Trait estimates are matched against a career knowledge base mapped to Indian streams, entrance routes and qualifications. Every recommendation in the report carries a reasoning trail — which trait scores drove it and why it ranked where it did.

    Read the reasoning, not just the ranking. If the reasoning describes someone who is not your child, the recommendation is wrong, and we would like to know.

    Reading the numbers correctly

    The trait figures printed in the report are population percentiles — how you compare with other students who have taken the assessment. Stream and career matching uses a different, within-student comparison: which of *your own* traits stand out relative to your own average.

    This is why a report can rank a career highly even when its headline trait number is not your highest. The two scales answer different questions.

    What we do not claim

  • We do not claim to predict your future. We measure disposition today, at one point in a life that is still forming. A Class 9 profile is a snapshot, not a verdict.
  • We do not claim your result is unchangeable. Retake it later and it may move — that is a property of adolescence, not a defect in the instrument.
  • We do not publish a validity coefficient or an accuracy percentage, because we are not yet in a position to stand one up honestly. When we have longitudinal outcome data worth publishing, it will appear here with its sample size and its limitations attached.
  • We do not replace a counsellor, a teacher or a parent who knows the student. The report is an input to that conversation.
  • No recommendation is influenced by payment. Paid tiers add depth — more analysis, more pathways — never a different answer.

    • Languages

    The assessment and report are available in English, Hindi, Marathi, Telugu, Tamil, Gujarati, Kannada, Odia, Punjabi, Malayalam and Bengali. The psychometric content is adapted for each language, not machine-translated at read time.

    If you think a result is wrong

    Tell us. Write to support@paramai.in with the student's name and what looks off, and a person will look at the actual session — the questions served, the answers given and the scores produced.

    Challenges to results are how the question bank improves. They are welcome, not a nuisance.

    Who we are

    Param AI™ is a product of Param Innovations Private Limited (CIN: U36000JH2025PTC025036), Dhanbad, Jharkhand, India.