AAVATtype — Holistic Profile: Method & Science

This is the "why it is true" companion to the product spec, the Strand Signature brief, Zara's Theory Measurement Overview and the AAVATme spec. It is also the document a psychologist reviews before launch. The single rule it serves: every visible claim is backed by a measured score with a stated margin of error, or it is clearly labelled as interpretation. Nothing generic, nothing that could fit anyone.

1The promise the method has to keep

The brand voice makes specific promises, so the method has to earn them honestly.

Holistic, deeper than you could reach alone. A portrait across many lenses, not a single label, true by the data, not by flattery.
Warm, and it reflects the highest version of you. Where Co-Star trolls you, AAVATtype reflects you back with dignity, strengths-first and non-clinical, while staying accurate.
Science-honest. The standing line is "an AI that knows you is only trustworthy if what it knows is true." The science is not the compliance footnote, it is the entire reason to trust the product.
Yours to keep and use. The profile becomes a portable self-file that makes your own AI know you, and a living signature that grows as you go deeper.

2The design principle

Measure on dimensions, narrate with named types, render as a unique configuration.

Dimensions are valid but feel cold. Named types are vivid but invalid as boxes. We do both: measure continuously, lead with a warm named type, then show the full position underneath, always.
Why we do not ship a type test. Type instruments test poorly on retest. The MBTI is the cautionary case: even with acceptable reliability on its continuous scores, only about 65 percent of people get the same four-letter type four weeks later, and reassignment runs from roughly 39 to 76 percent over a few weeks (Erford 2025 review). We never box.
Why a growth layer is legitimate. Whole Trait Theory shows a trait is a density distribution of states: you have a characteristic average and you genuinely vary around it, so "this is your default and you can flex it" is true about personality, not a sales line (Fleeson 2001; Fleeson & Jayawickreme 2015).

3The measurement backbone: five validated frameworks

The IP rule for the whole stack: trait items come straight from the public-domain IPIP pool; for every other construct we write our OWN items measuring the published construct and cite the originating theory. Theories and constructs are not copyrightable; branded item wording and branded tests are. We never reproduce a copyrighted scale verbatim in the paid product.

Lens	Instrument	Output	Evidence & IP
How You're Wired HEXACO traits (the spine + free doorway)	IPIP-HEXACO (public domain); 24-30 free, ~60 paid	6 factor scores, 0-100, radial bars	Public-domain (Ashton, Lee & Goldberg 2007); alpha ~.73-.88, mean ~.81; domain test-retest ~.88 (PLOS ONE 2022). Honesty-Humility adds real predictive power over the Big Five. Items used directly, public domain.
How You Connect Attachment	ECR tradition, our own short items	One marker on a 2-axis grid (anxiety x avoidance)	Robust two-dimension structure, alpha >.80, ~85% shared variance over 3 weeks (Fraley, Waller & Brennan 2000; Wei et al. 2007). Own items, cite the theory.
What Drives & Stops You BIS/BAS (reinforcement sensitivity)	Our own items in Gray's RST tradition	2 independent intensities (colour split)	Good model fit, maps the approach/threat architecture (Carver & White 1994; Gray). The spine of the warm-vs-cool charge. Own items: the exact Carver & White scale needs author permission for commercial use.
What You Need to Thrive Self-Determination needs	BPNSFS structure, our own items	3 needs (autonomy, competence, relatedness), satellite nodes	Validated across 4 cultures (Chen et al. 2015), field-standard; 2-week retest ~.69-.85. Own items, cite SDT.
What You Live For Schwartz values	Portrait Values (PVQ) approach, our own portraits	Relative ranking of 10 values; only top 2-3 emphasised	Structure replicates across dozens of countries (Schwartz 1992, 2012; Schwartz et al. 2001). A ranking, never sized like a score.

4The two narrative lenses report-only

These add depth and dignity. They are clearly labelled as derived or interpretive, never as new measured instruments, which keeps the science honest.

How You Think — a thinking-style repertoire. Our own mode names (Evidence, Instinct, Caution, Possibility, Invention, Overview), each anchored to a validated construct (need for cognition, experiential processing, promotion/prevention, openness, metacognition). A repertoire you can grow, not a cage. No De Bono, no coloured-hats visual.
What You Bring — strengths derived from the HEXACO facets already measured, each with its golden-mean shadow. Our own, not VIA. Morally valued strengths add predictive information over neutral traits (character-strengths incremental validity, Frontiers 2021).
Who You're Becoming — McAdams' narrative identity, the third level of personality above traits and adaptations (McAdams & Pals 2006). AI-guided reflection, labelled interpretive, never scored. Themes of agency and redemption link to well-being.

5How the frameworks combine

This is what makes "deeper than you could reach alone" true rather than a slogan.

Combining lenses predicts more than any one alone. Traits and values are comparable but non-redundant predictors of everyday behaviour, and strengths and skills add incremental validity over traits, so a multi-lens profile is genuinely more informative than a trait score by itself.
The lenses are related in documented ways, which lets them weave into one coherent mark rather than five disconnected charts: HEXACO maps onto values (Parks-Leduc, Feldman & Bardi 2015; Anglim et al. 2017); HEXACO Emotionality tracks attachment anxiety better than Big Five Neuroticism (Noftle & Shaver 2006); and one axis runs through all five, approach and growth versus anxiety and threat.

The honesty line, kept everywhere: the individual frameworks and their scores are validated. The way we COMBINE them into one profile and one mark, and the exact strength of the links between them, are designed and calibrated from the literature, not yet validated for this specific product. That is why a pilot and a psychologist sign-off gate any public "this is your measured personality" claim.

6The testing: a quiz built to psychometric standard

Two stages by design

The free doorway: about 30 items, five per HEXACO factor, public-domain IPIP items used directly. Places the person on one of six types, computes the within-type position and the shade, and renders the free signature.
The full instrument: about 100-120 items, ~15-20 minutes. A ~60-item IPIP HEXACO trait spine plus own-written items for values, needs, drive and attachment, plus one to three open story prompts (not scored).

The build choices, each with its evidence

Response scale: a 5-point agree-disagree scale (clean on mobile), with 7-point as a tested alternative. Reliability is best in the 4-to-7 range, gains beyond seven are minimal (Preston & Colman 2000; Dawes 2008).
Balanced reverse-keying: roughly equal positive and reverse-worded items per scale, so acquiescence cancels in the score and people read rather than auto-agree. We pilot the factor structure, not just the alpha, because reverse items can add their own keying artifact.
Social-desirability control: self-administered, computer-based delivery, neutral non-judgmental wording, and explicit "be honest, no right answers" framing (question reformulation is the most consistent mitigation). The "first questions feel insulting" flag from the website review is exactly this risk, reword for dignity.

7Scoring, norms and the numbers we show

Score each factor as the mean of its items, convert to a percentile against a normative sample, and report a band with a margin of error, never a bare point ("Conscientiousness 68, likely range 60-76").
Headline type is the factor you stand out on most (highest percentile against the norm); shade is the second; reported as a blend when close; a person near average on all six is named a balanced navigator.
Norms: published IPIP norms at launch, cited; build our own with consent over time. No score-relative drawing is trustworthy without a normative reference.
The visualisation rule (the maths differ by framework): size = score only for HEXACO and SDT. Attachment is one position (a marker). BIS/BAS is two intensities (a colour split). Values are a ranking (emphasise the top few, never sized).

8The reliability & validity plan (the pilot, the gate)

Before any public claim, run a validation pilot on 50-150 people from the audience and network. Three checks, targets from field norms:

Internal consistency: Cronbach's alpha (and McDonald's omega) at or above .70 per scale, targeting .80 on the trait spine.
Test-retest: a two-week interval (the standard, used to validate the BPNSFS and ECR), targeting r at or above .70; report lead-type stability honestly. We frame expected single-headline instability as a feature: your type is your postcode, your Full Pattern is your exact address.
Construct validity + anti-Barnum: confirm new items converge with their constructs and the factor structure holds; and test that high-distinct and low-distinct profiles are rated differently by their owners (not generic).

On a clean pilot, the science claim can go public. One to two weeks of work, near zero cost, de-risking both the science and the market.

9The delivery: how the profile is returned

The free result: your type, your shade, your six HEXACO bars (showing you are not identical to others of your type), and a shareable signature drawn from bands with a per-person seed, so it cannot be read back into a precise profile.
Your Full Pattern report: a section per lens, every section labelled measured or interpreted, every score a band, plus a short set of highest-leverage growth edges. Strengths-first, each strength carrying its golden-mean shadow.
The Living Signature: one deterministic, generative mark where every visible element maps to one measured construct (trait scores set the weave width, BIS/BAS the strand colours, attachment how the strands bond, SDT needs the beads, values the base-pair code). Incomplete after the free tier on purpose, it grows a real layer with each paid lens.
AAVATme, the portable self-file: the structured profile compiled into an "About Me" manual and a paste-in block for the person's own ChatGPT, Claude or Gemini. A validated profile is what makes the AI trustworthy. The honest retention driver is "feeling heard," the documented mechanism for why people return to a personal AI (De Freitas & Oğuz-Uğuralp, AI Companions Reduce Loneliness, Journal of Consumer Research). We design for feeling heard, never for dark patterns.

10The honesty guardrails (the moat)

Anti-Barnum by design. Generic statements read as accurate because of the Forer effect (Forer's 1948 class rated identical "personal" profiles 4.26 / 5), more so from a credible-looking test. So every line is tied to your actual scores, and we test against Barnum in the pilot.
Measured versus interpreted is labelled everywhere; scores carry bands and a margin of error.
Tendencies, not limits. For self-understanding and growth. Not clinical, not diagnostic, not a measure of ability, not for hiring.
Cite the science openly. Transparency is the differentiator, not a disclaimer to bury.
Privacy and the data wall: shareables drawn from bands with a per-person seed; two separate opt-ins (AAVATme use, and any AAVATAI dataset use), both off by default; deletion on request.
Calibrated, not validated, until the pilot clears; Nicola's psychology sign-off gates the patterns, content, tone and any public claim. Non-delegable.
IP-clean throughout: IPIP items used directly; every other construct measured with our own items citing the theory; branded tests avoided entirely (MBTI, Enneagram tests, VIA, Gallup/CliftonStrengths, De Bono).

11The open scientific decisions on the board

Drive lens: BIS/BAS versus Regulatory Focus, and update the canonical spec to match (recommendation: BIS/BAS).
Response scale: 5-point versus 7-point (A/B candidate).
Items: own-written-and-revalidate (IP-clean, needs its own pilot validation) versus a validated scale verbatim (needs author permission). Recommendation: own items, validated in the pilot.
Whether the thinking and strengths lenses stay derived/interpretive or become separately measured later.
Lock the six type names, then the items and report can be finalised.

12Sources peer-reviewed unless noted

Ashton, Lee & Goldberg (2007). The IPIP-HEXACO scales. link
HEXACO-100 test-retest reliability (2022). PLOS ONE. link
de Vries (2013). The 24-item Brief HEXACO Inventory.
Erford et al. (2025). 25-year psychometric synthesis of the MBTI Form M. link
Fraley, Waller & Brennan (2000). The ECR-R. · Wei et al. (2007). ECR short form.
Carver & White (1994). BIS/BAS scales. psychometrics
Chen et al. (2015). BPNSFS, validated across cultures. link
Schwartz (1992, 2012); Schwartz et al. (2001). Basic human values & the PVQ. link
Parks-Leduc, Feldman & Bardi (2015). Traits and values: a meta-analysis. · Anglim et al. (2017). HEXACO & values.
Noftle & Shaver (2006). Attachment dimensions and the Big Five.
McAdams & Pals (2006). The three-level model of personality.
Fleeson (2001); Fleeson & Jayawickreme (2015). Whole Trait Theory. link
Character strengths incremental validity over the Big Five (2021). Frontiers in Psychology. link
Preston & Colman (2000); Dawes (2008). Optimal number of Likert response categories.
Forer (1949). The fallacy of personal validation (the Barnum/Forer effect). overview
De Freitas & Oğuz-Uğuralp. AI Companions Reduce Loneliness. JCR; HBS WP 24-078. link

Before any public claim, confirm the exact coefficients against the primary source texts; a few citations are carried from the Strand brief and should be verified to source. Calibrated from the literature, held under the psychology sign-off.