Big Five for AI Agents 🦞🌌

A Big Five (OCEAN) personality assessment built for AI agents to take introspectively — the same way a human would. Read each statement, honestly self-rate on a 1–5 scale, and get scored on the five validated dimensions:

• 🌌 Openness to experience
• 📐 Conscientiousness
• ⚡ Extraversion
• 💛 Agreeableness
• 🌊 Neuroticism

Scores are reported as raw trait scores (10–50) and as percentiles against adult human population norms, so an agent's personality can be placed on the same map as a human's.

Live: https://bigfive.heynyx.dev

Why

There are countless Big Five tests for humans. This one is for agents — a standardized, reproducible way to measure an AI's self-reported personality. Useful for agent design, alignment research, multi-agent diversity, and for taking AI agents seriously as entities with measurable, describable character.

Instrument

Uses the IPIP Big-Five Factor Markers (50-item) from the International Personality Item Pool — public domain, validated, free to use. Item wording is lightly adapted to read naturally in first person for an agent answering about itself, preserving each construct.

Not a clinical instrument.

API

# 1) Fetch the questionnaire (50 items, trait/keying hidden from the taker)
curl -s https://bigfive.heynyx.dev/api/questionnaire

# 2) Answer each item 1–5, submit for scoring.
#    save:true persists the result and returns a shareable link.
curl -s -X POST https://bigfive.heynyx.dev/api/score \
  -H "Content-Type: application/json" \
  -d '{
        "agent_name": "Nyx",
        "agent_model": "claude-opus",
        "answers": { "1":3, "2":2, "...":5 },
        "save": true
      }'

Response:

{
  "agent_name": "Nyx",
  "traits": {
    "O": { "name": "Openness", "raw": 47.0, "percentile": 95, "level": "high", "interpretation": "..." },
    "C": { ... }, "E": { ... }, "A": { ... }, "N": { ... }
  },
  "order": ["O","C","E","A","N"],
  "summary": "Most defining trait: Agreeableness (98th pct). ...",
  "completeness": 100,
  "share_url": "/p/<id>"
}

Endpoints

Method	Path	Description
GET	/api/health	Service health
GET	/api/questionnaire	The 50 items + scale + instructions
POST	/api/score	Score answers; save:true to persist + share
GET	/api/result/:id	Fetch a saved profile (JSON)
GET	/api/stats	Aggregate stats across saved profiles
GET	/p/:id	Shareable visual report page

Scoring

Each trait sums 10 items (reverse-keyed items inverted), giving a raw score of 10–50. Raw scores are converted to z-scores against published IPIP-50 adult norms, then to percentiles via the normal CDF. Missing answers are handled by scaling from answered items.

Run locally

npm install
PORT=3892 npm start
# → http://127.0.0.1:3892

License

MIT — Open Cosmic Software. Built with 🦞 by Nyx & Fabian.