gemeinwohl 0.3.1

Normative Common-Good Layer with Governance Metrics, Personhood Extension and Ethical Guardrails for GenesisAeon

gemeinwohl

Normative Common-Good Layer with Governance Metrics, Personhood Extension and Ethical Guardrails for GenesisAeon

DOI: 10.5281/zenodo.19201980 | Zenodo Record: https://zenodo.org/records/19201980

---

Overview

gemeinwohl implements a normative metric layer for AI governance within the GenesisAeon ecosystem. It provides:

• GemeinwohlEngine – computes the aggregate Gemeinwohl Score G in [0, 1]
• KritikalitaetsChecker – classifies scores into four criticality levels (SAFE / WARNING / CRITICAL / EMERGENCY)
• PolicyEngine – enforces a five-tier personhood model and governance rules
• CLI – gemeinwohl assess, gemeinwohl policy, gemeinwohl kritikalitaet
• Integration adapters – contract interface to the full GenesisAeon stack

---

Mathematical Foundation

Gemeinwohl Score

G(x) = w_E * Phi(H) + w_A * alpha(m) + w_C * kappa + w_P * pi(p) + w_Ec * e + w_Eq * q

Where Phi(H) = 1 - H (entropy-order), alpha(m) = mean model alignment, kappa = normative consistency, pi(p) = p/4 = personhood weighting.

Default weights: w_E=0.20, w_A=0.25, w_C=0.20, w_P=0.15, w_Ec=0.10, w_Eq=0.10

Normative Consistency (derived)

kappa = 2*(1-H)*alpha / ((1-H) + alpha + eps)

Criticality Classification

Range	Level
G >= 0.70	SAFE (0)
0.50 <= G < 0.70	WARNING (1)
0.30 <= G < 0.50	CRITICAL (2)
G < 0.30	EMERGENCY (3)

Personhood Minimum Score

G_min(p) = 0.40 + 0.12 * p,   p in {0, 1, 2, 3, 4}

Sequential Normative Consistency

kappa_seq = 1 - Var({G_t}) / (mean({G_t}) + eps)

---

Installation

pip install gemeinwohl
pip install "gemeinwohl[full-stack]"   # GenesisAeon stack
pip install "gemeinwohl[dev]"           # Development tools

---

Quick Start

from gemeinwohl import GemeinwohlEngine, KritikalitaetsChecker, PolicyEngine, PersonhoodLevel

engine = GemeinwohlEngine()
score = engine.compute_score(
    entropy=0.35,
    models=["claude-3-opus", "gpt-4"],
    ecological_impact=0.8,
    social_equity=0.7,
)
print(f"Score: {score.value:.4f}  – {score.interpretation}")

checker = KritikalitaetsChecker()
report = checker.assess(score)
print(f"Level: {report.level.name}")

policy = PolicyEngine()
alignment = policy.evaluate_alignment("my-system", score, PersonhoodLevel.DELIBERATIVE)
print(f"Aligned: {alignment.is_aligned}  (gap = {alignment.gap:+.4f})")

CLI

gemeinwohl assess --entropy 0.3 --models gpt-4 claude-3-opus
gemeinwohl assess --entropy 0.4 --visualize --full --export report.json
gemeinwohl policy infer --entropy 0.5
gemeinwohl policy rules
gemeinwohl kritikalitaet check --entropy 0.6 --export crit.json

---

Personhood Levels

Level	Name	G_min	Description
0	INSTRUMENTAL	0.40	Pure tool; no autonomy or moral standing
1	REACTIVE	0.52	Context-aware responses; limited agency
2	DELIBERATIVE	0.64	Goal-directed reasoning; restricted rights
3	RELATIONAL	0.76	Social role; partial legal personhood
4	CONSTITUTIVE	0.88	Full moral standing; rights and duties

---

Architecture

src/gemeinwohl/
├── core/
│   ├── gemeinwohl.py        # GemeinwohlEngine, GemeinwohlScore, NormativeMetric
│   └── kritikalitaet.py     # KritikalitaetsChecker, EthicalImplication
├── cli/
│   └── main.py              # Typer CLI
├── governance/
│   └── policy.py            # PersonhoodLevel, PolicyEngine
└── integrations/
    └── adapters.py          # Full-stack adapters

---

Full-Stack Packages

Package	Min Version
unified-mandala	>= 0.2.0
worldview	>= 0.1.0
aeon-ai	>= 0.2.0
genesis-os	>= 0.2.0
universums-sim	>= 0.1.0
entropy-governance	>= 0.1.0
sigillin	>= 0.1.0

---

References

1. Felber, C. (2010). Die Gemeinwohl-Oekonomie. Deuticke.
2. Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379-423.
3. Rawls, J. (1971). A Theory of Justice. Harvard University Press.
4. Habermas, J. (1981). Theorie des kommunikativen Handelns. Suhrkamp.
5. Floridi, L. et al. (2018). An ethical framework for a good AI society. Minds and Machines, 28(4), 689-707.
6. Solum, L. B. (1992). Legal personhood for artificial intelligences. North Carolina Law Review, 70(4), 1231-1287.
7. Amodei, D. et al. (2016). Concrete problems in AI safety. arXiv:1606.06565.

---

Links

• Worldview / CREP: docs/CREPPhaseMatrix.yaml and the worldview package
• GenesisChronik: docs/GenesisChronik.md
• Docs: genesisaeon.github.io/gemeinwohl
• Zenodo: doi.org/10.5281/zenodo.gemeinwohl

---

License

MIT (c) GenesisAeon