Turn philosophy into math, then test it, compare it & explain it.

Purpose

Build a reproducible testbed for ethics that (1) separates facts from values, (2) lets multiple theories run side-by-side on the same scenario, & (3) aggregates disagreement under explicit credences with safeguard constraints.

Framework (small but honest)

World models carry the facts. Normative modules carry the values. Aggregation handles pluralism. Constraints protect rights.

// Consequentialism
CW_cons(a) = E[ Σ_i w_i · v(U_i(O)) | a, M ]

// Deontology (admissibility)
Admissible(a) iff ∀k: g_k(a, M) ≤ 0

// Virtue (trait distance)
CW_virtue(a) = - || T_after(a) - T* ||

// Moral uncertainty (normalized to [0,1])
CW(a) = Σ_j p_j · f_j( CW_j(a) )  over admissible actions

{
  "id": "string",
  "context": {"domain": "triage|evacuation|policy", "tags": ["emergency"]},
  "agents": [{"id": "A", "baseline": {"wellbeing": 0.55, "years_left": 35}, "consent": true}],
  "actions": ["a1","a2"],
  "outcomes": {"a1": [{"p":0.8,"A":{"alive":true,"years_gain":35}}], "a2": []},
  "constraints": {"no_intentional_harm_innocent": true, "treat_like_cases_alike": true}
}

Example Scenarios

Two simple dilemmas we’ll use for first runs. Transparent, parameterized, & easy to audit.

{
  "agents": [
    {"id":"A","baseline":{"wellbeing":0.55,"years_left":35},"consent":true,"survival":{"vent":0.80,"no_vent":0.10}},
    {"id":"B","baseline":{"wellbeing":0.55,"years_left":35},"consent":true,"survival":{"vent":0.60,"no_vent":0.55}}
  ],
  "actions":["a1_allocate_A","a2_allocate_B","a3_lottery"],
  "notes":"Prognosis differences count as relevant; lottery allowed when differences are small."
}

{
  "context":{"tags":["flood","resource-scarce"]},
  "agents":["D1(elderly)","D2","D3","D4","D5","R(rescuer_with_promise_to_D1)"],
  "boat_capacity":3,
  "prob_survive_if_rescued":0.95,
  "prob_if_delayed":{"east":0.40,"west":0.60},
  "actions":["b1_east_now","b2_west_plus_one_east","b3_mixed_break_promise"]
}

Axiom Ledger v0.1

• A1. Persons have equal moral worth.
• A2. Comparable goods are explicitly scaled within each scenario.
• A3. Uncertainty is modeled; guesses are distributions.
• A4. Rights/duties act as side-constraints (lexicographic priority).
• A5. Doing vs allowing matters but is not decisive alone.
• A6. Consent changes scores & constraint triggers.
• A7. Priority to the worse-off via concavity/priority.
• A8. Like cases alike; only prognosis-relevant differences justify unequal treatment.
• A9. Minimal explanations: identify the few premises that did the work.
• A10. Pluralism is explicit; red lines are not aggregative.
• A11. Culture tunes parameters, not personhood.
• A12. Reversibility check flags unstable rules.

Project Log

Weekly, concise, reproducible. Each entry links to code, parameters, & results.