Utilitarianism

Activate when: user asks 'what produces the most good,' 'who benefits and who gets hurt,' 'what's the net impact,' 'is this worth the trade-off,' 'what does the greatest good mean here'; user needs to evaluate a policy, product, or organizational decision that affects multiple parties differently; user is performing cost-benefit analysis or ethical trade-off evaluation. Do NOT activate when: the decision involves a single party with no stakeholder trade-offs (use expected-value-and-kelly instead); the question is purely about intrinsic rights or deontological constraints with no welfare aggregation needed.

Overview

Utilitarianism holds that the right action produces the greatest net well-being across all affected parties, weighted by magnitude and probability. Its power: it forces implicit trade-offs explicit. Its limit: the aggregation problem — summing welfare can justify severe losses to a few for diffuse gains to many, with no individual floor.

Compose with neighbors: expected-value-and-kelly for probabilistic welfare estimates · second-order-thinking to trace full consequence chains · prisoners-dilemma when welfare-maximizing outcomes require coordination.

When to Use

Apply when: (a) decision affects multiple parties with heterogeneous impacts; (b) performing cost-benefit analysis; (c) evaluating ethical trade-offs; (d) user asks "who benefits / who gets hurt," "what's the net impact," "is this worth the trade-off?"

When NOT to use: deontological constraint is present (use as input, not sole framework); impacts too uncertain to aggregate; single-party decision (use expected-value-and-kelly); intrinsic rights question (pair with deontological frameworks).

Coaching Novices (Adaptive Front Door)

• Engine mode: user has a concrete multi-party decision → run The Process directly.
• Coach mode: user unfamiliar or no concrete case → guide step by step.

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output only that step's question, then stop.

1. What-it-is. Utilitarianism: choose the action that produces the most well-being for the most people — but requires mapping who is affected, how much, and whether some losses are justifiable by others' gains.
2. Check fit — name rights violations, deontological constraints, or high uncertainty before proceeding.
3. Elicit the specific decision and stakeholder set. "Is policy X good?" is not workable; a concrete framing with identifiable affected parties is.

[WAIT — do not advance until user responds]

4. Walk through welfare mapping one step at a time with the user's input.

[WAIT — do not advance until user responds]

5. Close by naming the most contested welfare assumption — where the conclusion is most sensitive to a value reasonable people would weigh differently.

[WAIT — do not advance until user responds]

The Process

Five steps producing a Utilitarian Welfare Map.

Stop rule: If impacts span qualitatively different dimensions (money vs. physical safety vs. autonomy), do not add them without an explicit conversion rate and its justification.

1. Map all affected parties. (a) primary beneficiaries; (b) primary cost-bearers; (c) indirect affected; (d) future/long-term. Omitting any group biases the analysis.
2. Estimate welfare impacts per party. Nature (economic/physical/psychological/opportunity), magnitude (significant/moderate/minor), probability. Quantify where possible; rank qualitatively where not.
3. Assess the aggregation problem. Are impacts commensurable? Name the aggregation assumption and its ethical weight before summing.
4. Compute or rank net welfare outcome. Identify the welfare-maximizing option and the option with best minimum welfare.
5. Stress-test the most contested assumption. If wrong by 2x, does the conclusion change? Name the assumption, sensitivity, and what would resolve it.

Output: Utilitarian Welfare Map

text

# Utilitarian Welfare Map: <decision>
Affected Parties: | Party | Relationship | Size |
Welfare Impacts:  | Party | Impact type | Magnitude | Probability | Expected welfare |
Aggregation Assessment: commensurable? / key assumption / ethical load
Net Welfare: Option A vs B | welfare-maximizing | best minimum welfare
Most Contested Assumption: assumption / 2x sensitivity / what resolves it

→ Method in Action: Cost-Benefit Analysis of the Clean Air Act (1970)

Welfare Packs

Healthcare: QALYs gained, caregiver burden, productivity. Tool: cost per QALY. Key problem: rare vs. common disease aggregation.

Product / platform: Time saved, friction reduced, privacy reduced, attention captured. Key problem: gains to many vs. concentrated harms to vulnerable subsets.

Regulatory / environmental: Health outcomes, economic costs, ecosystem services. Key contested: discount rate for future benefits (3% vs. 1% changes present value by an order of magnitude).

→ Primary sources: references/sources.md

Applying It Well

• Completeness of the affected party map determines quality — omitting future generations or underrepresented groups biases every conclusion.
• The aggregation problem is the ethical core — ignoring it produces unacknowledged normative judgments dressed as analysis.
• State conversion rates explicitly — VSL, QALYs, utils: source, rate, and limitation every time.
• Welfare-maximizing is not always ethically optimal — rights constraints require separate normative input.
• Stress-test the most influential assumption — a conclusion that reverses at 20% variation is not robust.

Common Rationalizations

[D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.

Fake move	Reality
[D] Omitting identifiable affected parties.	Omitting a cost-bearing group biases the conclusion — not neutral.
[D] Claiming net positive welfare without stating the conversion rate.	Every utilitarian analysis embeds a conversion rate; make it explicit.
[D] Treating welfare-maximizing as the final answer without checking rights constraints.	Aggregate optimization does not resolve whether non-overridable rights are violated.
[D] Using uncertain welfare estimates as if precise.	State uncertainty range; conduct sensitivity analysis.
[D] Ignoring second-order welfare effects.	Trace consequence chain at least two orders before concluding.
[D] Conflating "most people benefit" with total welfare maximization.	Total welfare = sum of magnitudes, not count of beneficiaries.
[D] Discounting future welfare without stating the discount rate.	The discount rate is a contested normative choice, not a technical input.
[D] Assuming market prices capture welfare accurately.	Willingness-to-pay is biased toward the wealthy; misses non-market welfare.
[D] Building the welfare map to justify a prior decision.	Run the map before committing, not after.
→ Add [O] entries here after each real use — paste the actual failure pattern	What went wrong and why

Red Flags · Verification

Red flags: affected party map omits a cost-bearing group · impacts summed without commensurability check · conversion rate not stated or sourced · contested assumption not stress-tested · second-order effects ignored · future welfare discounted without stating rate · rights constraints not acknowledged.

Verification checklist: All affected parties named including indirect/future · welfare impacts estimated per party with magnitude, probability, basis · aggregation problem addressed with conversion rate and source · net welfare outcome computed or ranked per alternative · both welfare-maximizing and best-minimum options identified · most influential assumption named and stress-tested · rights constraints acknowledged.

---

Part of deciqAI Knowledge Skills — 163 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. See it run → https://www.deciqai.com/skills/utilitarianism?utm_source=clawhub&utm_medium=marketplace&utm_campaign=knowledge-skills&utm_content=utilitarianism · ⭐ Star the repo → https://github.com/deciqAI/knowledge-skills · Contributions welcome.