Methodology

Dishadarshaq scores every assessment through four pillars (Heart, Brain, Mind, Environment) using a deterministic engine pinned to a versioned scoring bundle. Every session is reconstructable from an append-only event log; every result has a SHA-256 hash that does not drift across runs.

Heart pillar — MIRT

Compensatory multidimensional scoring over RIASEC, Big Five, and India-localized values. Classical alpha-corrected SEM with literature priors during pilot mode; recalibrated from production data after N ≥ 500 per cohort.

Brain pillar — DINA

Five-attribute cognitive diagnostic over 32 latent classes: number sense, proportional/algebraic reasoning, quantitative data interpretation, verbal inference, logical/deductive reasoning. Bayes-modal estimation with monotonicity constraints replaces naive EM after pilot data is collected.

Mind pillar — classical with guardrails

WHO-5 Wellbeing Index plus India-validated SDQ for K-12. Workplace EQ (non-clinical) for Enterprise. Student-facing labels are restricted to three non-clinical values; a centralized resolver and a CI grep audit prevent any clinical token from reaching a student-facing surface.

Environment pillar — deterministic rules

Weighted features over school, board, locality, parental education, SES, language, and internet access. Dimensions below 40% feature completeness are hidden from the primary result and surfaced as quality flags.

Replay invariance

Scoring is pure-function. Identical inputs produce byte-identical canonical JSON and byte-identical result hashes across hundreds of runs. The contract is enforced by a custom go vet analyzer that rejects time.Now(), unseeded random, and forbidden I/O imports inside scoring code.

Pilot status

Until per-cohort norms reach pre-registered sample-size targets, every report carries a "pilot mode" banner. The Open Science Framework pre-registration is published; updates land in the technical manual change log.