Methodology

SELVET’s design principles

SELVET represents a new generation of personality testing that combines academic rigor with modern technology. Four core principles:

1. Scientific grounding: Questions from validated instruments (IPIP, PVQ)
2. Cross-framework mapping: Answers update scores across frameworks
3. Adaptive precision: Follow-up questions where they are needed most
4. Transparent methodology: You can see how the results are calculated

The item bank

SELVET uses 540 questions in total, across four frameworks:

How your scores are calculated

Big Five: domain and facet level

Scores are calculated across the 5 main domains, and across 30 facets (6 per domain) where the data supports it. Each facet is built from the facet codes on the questions themselves, not from their order in the test, and requires at least 4 answered items. A facet level is shown only once all six facets in a domain are measured — so the full version (300 questions) yields facets, while the short version (50) yields domains only. Reverse-scored questions are handled automatically, and raw scores are normalized to 0-100.

Thinking style: preference strength

For each dimension (E/I, S/N, T/F, J/P) both preference and strength are calculated. 50 = no preference, 100 = strong preference.

Motivation: ranking with a wing

All nine types are scored, the primary type is the highest score, and the wing is determined by comparing the neighboring types.

Values: a ranked hierarchy

Values are presented as a ranked list, centered on your personal average in order to show relative priorities.

Adaptive testing

When a score sits in the "uncertainty zone" (typically 45-55%), follow-up questions are activated, designed to discriminate sharply between the alternatives.

You can always choose to skip, but you are then told that the result is less certain in that area.

You can see the uncertainty in your result

SELVET reports not only point estimates, but confidence intervals as well. The Standard Error of Measurement is calculated from the number of questions and the reliability of the scale (Cronbach’s alpha).

A shareable personality code

SELVET generates a compact, shareable code that sums up the profile:

The code contains the thinking style (4 letters), the Enneagram type with its wing, and the two highest-ranked values. Designed to be short enough for social media, intriguing enough to spark curiosity, and comparable with other people’s codes.

AI-generated dashboard

The premium dashboard uses a large language model (LLM) to generate personalized insights. Prompt engineering ensures:

• Specificity: References concrete score combinations
• Contrastive statements: Includes what the profile does not indicate
• Action-oriented: Concrete suggestions to test

Quality assurance includes an automated check that the text refers to actual scores and avoids absolutes.