RIASEC interest-fit (Holland codes)

Status: Folded · Evidence: P · Family: Self and team awareness · Verdict: fold (2026-06-11)

Use instead: Decision Option Review

What it is

RIASEC is John Holland’s typology of vocational interests: six interest types - Realistic (doers), Investigative (thinkers), Artistic (creators), Social (helpers), Enterprising (persuaders), Conventional (organizers) - arranged on a hexagon so adjacent types are most alike and opposite types least alike. A person and a work environment are each described by their highest types (a two-or-three-letter Holland code), and the theory’s central claim is congruence: people whose interest profile matches their environment’s profile tend to be more satisfied and to perform and persist better.

As a thinking move, stripped of the inventory and the norm tables, the candidate reduces to a fixed-criteria fit check: take a role, an option, or a person; score it against the six pre-named interest dimensions; read off where the profile matches and where it clashes, using the hexagon’s adjacency to grade partial matches (same code congruent, adjacent codes a mild misfit, opposite codes a severe misfit). The deliverable is a fit matrix - the six types down the side, the thing being evaluated across the top, a match score per cell, and a where-it-fits / where-it-clashes summary with the caveat that interest fit predicts only a thin slice of the outcome. That fit matrix is the only part that could ship as an agent move; administering an interest inventory and computing a congruence index against occupational norm groups is instrument administration, not a reasoning move an agent performs.

When it helps / when it misleads

It helps when the evaluation criteria genuinely are the six interest dimensions and the decider wants a structured, repeatable read of fit rather than an ad-hoc gut call: triaging which of several roles best matches a stated interest profile, or naming the specific dimension a misfit sits on (“this role is high-Enterprising, the person profiles Investigative”) so the mismatch becomes discussable rather than vague. It is most defensible as a disengagement-risk and persistence lens, because that is closest to what the interest literature actually measures.

It misleads in several ways, and these are the walls. Fit is not competence: measured interests predict job performance only weakly (single-scale validity around r = .14 in Van Iddekinge et al. 2011), so interest fit must never function as an ability screen or a hiring gate. An unmeasured profile is only a hypothesis: all the supporting evidence comes from validated inventories, so an impressionistic code assigned from a resume or a conversation inherits the taxonomy but none of the measurement validity and must be held loosely. Do not promise satisfaction: the congruence-satisfaction link is modest and was long contested (Tsabari, Tziner and Meir 2005 put it near r = .17; Tinsley 2000 argued the fit model fails outright; Hoff et al. 2020 settle on rho = .19, real but far weaker than folk wisdom assumes). The hexagonal geometry that drives the adjacency read fits U.S. samples better than samples elsewhere (Rounds and Tracey 1996), so the partial-match grading weakens cross-culturally. And forcing a non-role decision onto R-I-A-S-E-C imports six irrelevant criteria and omits the ones that matter (cost, risk, reversibility): choosing among products, strategies, or vendors against criteria you select is decision-option-review, not an interest-fit question. Finally, typing a named colleague in a consequential decision without their input is a misuse, not a thinking move.

What the evidence says

Honest grade: M for the congruence construct, capped at P (transferred) for the shippable move. The registry’s preliminary M is the construct grade; the governing grade for an agent-run fit-check is P, and emitting the M would launder the inventory literature’s evidence onto a move it does not test.

What the research supports. The six-type structure replicates: Tracey and Rounds (1993, Psychological Bulletin) structurally meta-analyzed the model and found the hexagonal / circular ordering holds reasonably in U.S. samples (adjacent types correlate more highly than opposite types), with Rounds and Tracey (1996) finding weaker and less consistent fit cross-culturally. Interests predict outcomes: Van Iddekinge, Roth, Putka and Lanivich (2011, Journal of Applied Psychology; 74 studies) reported validities of .14 for job performance, .26 for training performance, and around -.15 to -.19 for turnover; Nye, Su, Rounds and Drasgow (2012, Perspectives on Psychological Science) summarized 60+ years showing interests predict performance, and their 2017 follow-up (92 studies, 1,858 correlations) found interest congruence a stronger performance predictor (baseline r around .32) than raw interest scores (around .16). Congruence relates to satisfaction, but modestly: Tsabari, Tziner and Meir (2005, 53 studies) found r around .17, and the most rigorous recent estimate, Hoff, Song, Wee, Phan and Rounds (2020, Journal of Vocational Behavior; 105 studies, k = 194, N = 39,602), found interest fit predicts overall job satisfaction at only rho = .19, their headline being that the relation is real but substantially weaker than the field had assumed.

What the research does NOT support, and the transfer. Every one of those studies measures a property of humans, generally using validated inventories: whether a person whose measured interests match their measured environment is more satisfied or performs better. None tests whether an agent that reasons about person-environment fit across six interest types decides or recommends better, and none tests the instrument-free version where both profiles are assigned by judgment. The evidence is congruence-as-construct, measured; the move is fit-reasoning-by-an-agent, impressionistic. The family rule is explicit: instrument validity evidence is not move-effectiveness evidence, and laundering one into the other is the failure this library exists to prevent. So the construct grades M; the move an agent could execute grades P, transferred and not agent-validated. Every numeric claim above maps to a named author and year; the widely-repeated claim that interest fit “doubles” engagement traces to no primary source and is excluded, as are the larger congruence-satisfaction figures that depend on title-inferred environment coding rather than pooled estimates.

Why it is / is not a skill here

Verdict: Fold into decision-option-review. This overturns the registry’s preliminary cand / build / M, with reasons.

The preliminary entry granted M and a build on “decades of vocational-psychology support,” noting distinctness from the option-evaluation skills was “to be confirmed.” Confirming it fails the bar on two independent grounds.

First, the tier. The M lives in the inventory literature - validated scores predicting outcomes - which is an adjacent claim, not evidence for an agent’s impressionistic fit read. The conservative governing grade for the move is P.

Second, and decisive, distinctness. The durable move, once the inventory is removed, is “score an entity against a fixed criterion set, grading partial matches by a domain rubric, and read the fit.” That is what decision-option-review (shipped; it absorbs Multi-Criteria Decision Analysis) already does: compare options against criteria and surface where each fits or clashes. The candidate-specific content is only which criteria - the six Holland types - plus a partial-match rubric (the hexagon adjacency). Preset criteria are a domain preset of the choose-your-criteria step, and an adjacency rule is a scoring rubric for partial matches, not a new operation. This is the exact shape of the already-recorded outside-in-inside-out-framing fold, where the sole candidate-specific content was which lenses and the registry folded it because a preset of an existing engine’s input-selection step is not a new mechanism. The six-row fit matrix and a criteria-by-options grid share their entire scoring engine, well above a fifth of the working mechanism, so the overlap ceiling is breached.

Anticipating the strongest counter (and the line the staged build took): that decision-option-review “has no person side, needs multiple options, and cannot supply the typology.” Each fails. The “person side” is a relabel - a profiled person is simply the entity being scored, the way an option is; MCDA routinely scores a single entity for a go/no-go fit, so “needs multiple options” is not a real wall; and “cannot supply the typology” confuses content with mechanism. morphological-analysis earned its build by adding a genuinely new operation (cross-consistency pruning of a Cartesian product) that the option matrix lacks; RIASEC adds no operation, only six preset criteria and an adjacency rubric. Content presets fold; new mechanisms ship. RIASEC is a content preset.

It is therefore not a Build: no separable artifact is uniquely RIASEC and not a fixed-criteria option matrix. It is not a Recipe: there is one move with preset criteria, not a chain. It is not a Reject: the fit-scoring move is real and worth locating, like the inversion-into-premortem and stakeholder-lens-into-parallel-perspectives folds. It also leans out-of-scope by domain - career and vocational guidance is an applied specialty nearer the sibling pm-skills library than a general thinking move - but the cleanest honest home for the generic mechanism is the fold into decision-option-review.

A second collision reinforces the fold: the candidate sits beside trait-lens-perspective (Big Five / HEXACO) in the same new family, and both apply a fixed psychometric typology as the scoring frame; shipping both would put two near-twin fixed-typology moves in one family. The branded sibling already points here - the Strong Interest Inventory entry says its generic core is Holland’s RIASEC and it “likely folds rather than shipping.” The chain resolves cleanly: Strong Interest Inventory folds to RIASEC, and RIASEC folds to decision-option-review.

Lineage and who to read

Origin: John L. Holland, “A Theory of Vocational Choice” (Journal of Counseling Psychology, 1959), developed across three editions of Making Vocational Choices: A Theory of Vocational Personalities and Work Environments (1973, 1985, 1997). The typology is generic and attributed to Holland; the proprietary instruments built on it - the Self-Directed Search (PAR) and the trademarked Strong Interest Inventory (originally E. K. Strong Jr., 1927; now The Myers-Briggs Company) - are the branded layer, not the model. The U.S. Department of Labor’s O*NET program assigned RIASEC profiles to its occupation database and offers a free Interest Profiler, making the environment-typing side public infrastructure.

To read the evidence honestly: Tracey and Rounds (1993) and Rounds and Tracey (1996) for the structure and its cross-cultural limits; Van Iddekinge et al. (2011) and Nye, Su, Rounds and Drasgow (2012, 2017) for what measured interests and congruence actually predict; Tsabari, Tziner and Meir (2005), Tinsley (2000), and especially Hoff, Song, Wee, Phan and Rounds (2020) for the modest, frequently-overstated congruence-satisfaction relation. The 2020 systematic review is the single best corrective to the popular “do what interests you and you will be satisfied” claim.