socionics lab

decode the psyche.

a panel of AI agents researches a character, or you, argues over the evidence, and settles on a socionics type.

start analysishow it works

the process

every analysis runs the same phases, modeled on how people reach consensus: gather the facts, form independent views, check, decide. you can add a debate between the agents.

  1. phase 1: the scout

    compiles a factual dossier on the subject, or on your answers: behavior, decisions, quotes. it does not assign a type.

    reads
    a character and their source, or your own answers
    produces
    a dossier of facts, quotes and relationships

  2. phase 2: the council

    three agents type the subject independently, each through one lens: reinin dichotomies, quadra values, or the function model. a direct read of the whole picture votes alongside them.

    reads
    reinindichotomiesquadraquadra valuesfunctionsmodel a
    produces
    three independent predictions

  3. phase 3: the debate (optional)

    off unless you turn it on in settings. each council agent reads the other two predictions, then agrees, pushes back, or revises its own call. it's there to read; the vote decides the type either way.

    reads
    the other two agents' predictions
    produces
    agreement, pushback, or a revised call

  4. phase 4: the validator

    fact-checks the council's theoretical claims against a socionics reference and flags the errors it finds.

    reads
    every claim the council made about theory
    produces
    corrections. it never picks a type itself

  5. phase 5: synthesis

    counts the votes (the direct read counts double), then the manager explains the result. the confidence score comes from how much they agreed.

    reads
    everything above
    produces
    one type, and a score for how much the agents agreed

the theory

socionics is a personality framework based on information metabolism: how the psyche takes in and processes different kinds of information.

model a

8 cognitive functions

each type has 8 functions arranged in a specific order. the leading function defines core perspective, while the vulnerable function is the point of least resistance.

  • Neintuition of possibilities
  • Niintuition of time
  • Sevolitional sensing
  • Sisensing of sensations
  • Tebusiness logic
  • Tistructural logic
  • Feethics of emotions
  • Fiethics of relations

square logic, angle ethics, triangle intuition, circle sensing. filled is extraverted, outlined is introverted.

quadras

4 value clusters

types in the same quadra share core values and communication styles, and tend to understand each other easily.

Alpha
ILE SEI ESE LII
Beta
EIE LSI SLE IEI
Gamma
SEE ILI LIE ESI
Delta
IEE SLI LSE EII

dichotomies

reinin traits

beyond jung's 4 dichotomies, socionics uses 15 trait pairs derived from type mathematics to tell similar types apart.

  • extraverted/introverted
  • intuitive/sensing
  • logical/ethical
  • rational/irrational
  • static/dynamic
  • positivist/negativist
  • asking/declaring
  • tactical/strategic
  • constructivist/emotivist
  • process/result
  • yielding/obstinate
  • carefree/farsighted
  • merry/serious
  • judicious/decisive
  • aristocratic/democratic

intertype

relationship dynamics

duality, conflict, supervision, mirror: socionics maps 14 distinct relationship patterns between types.

ILE
  1. Neposition 1, leading, leading
  2. Tiposition 2, creative
  3. Seposition 3, role
  4. Fiposition 4, vulnerable
  5. Siposition 5, suggestive, matches SEI's leading function
  6. Feposition 6, activating
  7. Niposition 7, ignoring
  8. Teposition 8, demonstrative
SEI
  1. Siposition 1, leading, leading
  2. Feposition 2, creative
  3. Niposition 3, role
  4. Teposition 4, vulnerable
  5. Neposition 5, suggestive, matches ILE's leading function
  6. Tiposition 6, activating
  7. Seposition 7, ignoring
  8. Fiposition 8, demonstrative
duality, ILE and SEI. each type's leading function is the other's suggestive one, so each supplies what the other seeks.

the research

classical socionics theory, applied by several AI agents that have to agree with each other before you get an answer.

multi-agent consensus

rather than one model making the call, several specialist agents analyze the same subject through different theoretical lenses. disagreement is surfaced, not hidden.

adversarial validation

a dedicated validator challenges the council's conclusions against theory. it catches logical inconsistencies that a single pass would miss.

confidence calibration

the final score reflects uncertainty. when agents disagree or the evidence is thin, the score drops. we don't pretend to a certainty we don't have.

a note on limitations

confidence scores reflect agent agreement, not objective accuracy.

this tool is an experiment in applying multi-agent AI to socionics analysis. it is not a replacement for trained practitioners, and it doesn't claim to be more accurate than human typists.

socionics typing is a nuanced practice that benefits from dialogue, lived experience and theoretical depth that no current AI can replicate. use this as a starting point for exploration, not a final determination. even high-confidence results deserve a second look.

what you get

analyses run on your own OpenRouter key, so you see and pay for exactly what each one costs.

  • your own OpenRouter key

    you control what analyses cost. the key is stored encrypted and only used for your own runs.

    1. sign in with Google
    2. add your OpenRouter key
    3. pay OpenRouter directly

  • a link for every result

    each analysis gets its own result page with a link you can share.

  • history with search

    every analysis you run is saved to your account. search past ones by name, type or source.

begin analysis.

pick a character, or type yourself.

start analysis