Socionics three-letter codes (ILE) name the top two Model A slots. Four-letter codes (ENTp) look like MBTI and are not MBTI. For introverts, function-based conversion flips the j/p letter — socionics LII (INTj) maps nearest to MBTI INTP — and this mapping is disputed. We display MBTI equivalents with ≈ and an asterisk, always. Any site that silently equates the two systems has made a choice it isn't telling you about.

The element symbols are Augustinavičiūtė's classical notation: triangle = intuition, circle = sensing, square = logic (matter), L-block = ethics (energy); filled = extraverted ("black"), outline = introverted ("white"). The four family hues are ours, one per family, purely for scanning — they carry no doctrinal meaning.

A type is its (base, creative) pair. Three operations generate everything: partner — same attitude, sibling element (Ne↔Se, Ti↔Fi); invert — same element, flipped attitude (Ne↔Ni); dualize — both at once (Ne↔Si). Every intertype relation is one recipe applied to the pair:

The test is MAGTI v1.0 (86 items), an instrument with MODEL A item design — every result also fit against MODEL G (§4) — scored in your browser through three channels: Model A fit (weight 1.0) from per-element strength and valuedness templates, dichotomy votes (0.7) across 4 Jungian + 8 Reinin dichotomies, and direct evidence (0.5) from quadra-climate and discriminator items — each z-normalized, combined, and softmaxed (τ=0.7) into the probabilities on your card. Unanswered items shrink votes conservatively; three strong dichotomy votes against the leader downgrade confidence one band. A test you can't audit is a horoscope with checkboxes — so audit it:

MAGTI-C, the MBTI conversion, uses the same engine plus one term: your MBTI opens a candidate set whose log-prior enters at weight 0.6 — against 2.2 of combined answer weight, so roughly four decisive answers outweigh the entire starting hypothesis. The engine also ranks types with the prior removed; when that answers-only ranking puts a non-candidate first by ≥ 0.2, your report shows a disagreement warning and confidence is capped. The priors, stated openly so you can disagree with them:

MODEL A is the classical model: Aušra Augustinavičiūtė's eight-slot architecture — the basis of the relations, the type pages, and the headline test verdict.

MODEL G is Viktor Gulenko's energy model (School of Humanitarian Socionics): the same sixteen types and eight elements, with the functions rearranged and read as energy rather than information. Every test result reports a Model G match alongside the Model A verdict, computed from the same answers. The structure we use:

The signature disagreement is the creative ⇄ demonstrative swap (highlighted rows): what Model A treats as strong-but-dismissed background, Model G promotes into the visible social mission. A person whose Model-A-demonstrative element is visibly used and valued is precisely who the two schools type differently — and the decisive template difference is the valuedness of position 8.

The match itself: your normalized element profile o = [Ŝ ; 0.8·V̂] is scored against each type's template under each model as match% = 50·(1+cos(o, τ)) — 100% a perfect profile match, 50% unrelated, below 50% actively opposite. Both models' templates are in the exported result and in the source of this page.

Model G is Victor Gulenko's energy model, used by the School of Humanitarian Socionics. It keeps the same sixteen types and eight elements but arranges the functions differently and reads them in terms of energy rather than information. This inventory's items were designed for Model A, so the Model G match is an approximation computed from the same evidence, not a diagnosis by Gulenko's school.

Augustinavičiūtė, The Dual Nature of Man · Gulenko's SHS publications · the community reference at wikisocion.github.io. Read the sources; don't take our word for it — the whole premise here is that authorities disagree.