{ "schema_version": "welfare-voi-pilot-v1", "generated_at": "2026-09-12", "status": "AI-assisted methods pilot; not an official Unjournal evaluation or funding recommendation", "scope": "Six contrasting papers selected from the public prioritization dashboard", "method": { "causal_chain": [ "research output", "belief or tool change", "policy, funding, or research decision change", "natural outcomes", "welfare" ], "funding_formula": "relevant budget × research-caused allocation-share change × incremental outcome per dollar versus displaced alternative × welfare per outcome − costs/harms", "coverage_formula": "eligible population × research-caused coverage change × incremental outcome per covered unit × duration × welfare weight − costs/harms", "research_tool_formula": "downstream decisions × P(tool/evidence improves) × P(improvement changes decision) × outcome difference × welfare weight − costs/harms", "provenance_tags": [ "source_estimate", "analyst_assumption", "unknown" ], "scenario_rule": "Low, central, and high entries are sensitivity scenarios. They are not statistical confidence intervals unless a source explicitly supplies an interval.", "aggregation_rule": "Do not produce a common score or rank when essential causal terms are unknown. Preserve natural outcomes and conditional calculations.", "income": { "utility": "CRRA", "eta_sensitivity": [ 1, 1.5, 2 ], "reference_consumption_ppp_per_person_year": 1000, "unit": "reference-consumption-doubling equivalent" }, "health_bridge": { "healthy_year_equivalents_per_consumption_doubling_year": 0.4347826086956522, "derivation": "GiveWell assigns 1 unit to a consumption doubling for one person-year and 2.3 units to averting one full YLD; 1/2.3 ≈ 0.435.", "status": "analyst convention" }, "animal": { "source": "Rethink Priorities Moral Weight Project", "rule": "Use probability-of-sentience-adjusted 5th/50th/95th welfare ranges and do not multiply by sentience again." }, "catastrophic_risk": { "rule": "Show research-attributable change in absolute risk × welfare at stake, with present-generation and long-future ledgers separate." }, "evaluation_counterfactual": "Value of commissioning an Unjournal evaluation is publication and use of the paper with versus without the evaluation. It is separate from the paper's welfare relevance." }, "headline_result": "Only the Kenya cash case supports a sourced distribution-sensitive welfare conversion per affected household. None of the six currently supports a defensible total expected-welfare estimate for the research product, because research-caused decision change and the displaced alternative are missing. The prescription, animal, and x-risk cases include conditional sensitivity calculations to expose the missing assumptions.", "papers": [ { "paper_id": "prescription-errors-india", "title": "Reducing Prescription Errors Through Information Intervention: A Field Experiment in Healthcare Operations", "url": "https://arxiv.org/abs/2609.09673", "case_role": "Direct health intervention in India with a source-reported mortality extrapolation", "cause_area": "Global health and health-system operations", "assessment_level": "quantified_sensitivity_not_welfare_estimate", "causal_chain": { "research_output": "Randomized platform experiment and difference-in-differences analysis of a non-mandatory drug-interaction alert.", "belief_or_tool_change": "EMR vendors or health systems become more willing to use passive, informative alerts and refine the interface.", "decision_change": "Additional deployment or earlier deployment beyond what would otherwise occur.", "natural_outcome": "Fewer flagged drug-drug-interaction prescription errors; downstream admissions and deaths are extrapolated.", "welfare_bridge": "Prevented morbidity and mortality expressed first as healthy-life-year equivalents; patient cost incidence remains separate." }, "pathway_type": "coverage", "quantitative_inputs": [ { "name": "prescriptions", "description": "Prescriptions in the experiment", "unit": "prescriptions", "provenance_tag": "source_estimate", "value": 2810000, "low": null, "central": null, "high": null, "source_url": "https://arxiv.org/abs/2609.09673", "note": "" }, { "name": "physicians", "description": "Physicians in the experiment", "unit": "physicians", "provenance_tag": "source_estimate", "value": 1700, "low": null, "central": null, "high": null, "source_url": "https://arxiv.org/abs/2609.09673", "note": "" }, { "name": "ddi_error_reduction", "description": "Relative reduction in flagged DDI errors", "unit": "proportion", "provenance_tag": "source_estimate", "value": 0.086, "low": null, "central": null, "high": null, "source_url": "https://arxiv.org/abs/2609.09673", "note": "" }, { "name": "full_scale_lives", "description": "Paper's annual India-wide potential-lives-saved extrapolation", "unit": "potential deaths/year", "provenance_tag": "source_estimate", "value": 134, "low": null, "central": null, "high": null, "source_url": "https://arxiv.org/abs/2609.09673", "note": "Appendix arithmetic uses population 1.307bn, 3.7% annual hospitalization, 1% DDI attribution, 8.6% reduction, and 0.32% mortality among serious adverse drug reactions. It is not an experimentally observed mortality effect." }, { "name": "research_caused_coverage_share", "description": "Share of the paper's India-wide reference scale additionally implemented because of the research", "unit": "proportion", "provenance_tag": "analyst_assumption", "value": null, "low": 0.001, "central": 0.01, "high": 0.1, "source_url": null, "note": "Illustrative sensitivity values; no evidence in the paper identifies this counterfactual research effect." }, { "name": "extrapolation_survival", "description": "Fraction of the mortality extrapolation surviving clinical validation and delivery constraints", "unit": "proportion", "provenance_tag": "analyst_assumption", "value": null, "low": 0.25, "central": 0.5, "high": 0.8, "source_url": null, "note": "" }, { "name": "healthy_years_per_death_prevented", "description": "Healthy-life-year equivalents per prevented death", "unit": "years", "provenance_tag": "analyst_assumption", "value": null, "low": 20, "central": 30, "high": 40, "source_url": null, "note": "" }, { "name": "net_costs_harms", "description": "Implementation costs and harms in the same welfare unit", "unit": "healthy-year equivalents/year", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" } ], "arithmetic": { "formula": "source full-scale potential deaths × research-caused coverage share × extrapolation-survival factor × healthy years per prevented death − costs/harms", "intermediate_lives": [ { "scenario": "low", "expression": "134 * 0.001 * 0.25", "result": 0.0335, "unit": "potential deaths prevented/year", "status": "scenario_only" }, { "scenario": "central", "expression": "134 * 0.01 * 0.5", "result": 0.67, "unit": "potential deaths prevented/year", "status": "scenario_only" }, { "scenario": "high", "expression": "134 * 0.1 * 0.8", "result": 10.72, "unit": "potential deaths prevented/year", "status": "scenario_only" } ], "scenarios": [ { "scenario": "low", "expression": "134 * 0.001 * 0.25 * 20.0", "result": 0.67, "unit": "gross healthy-life-year equivalents/year", "status": "scenario_only" }, { "scenario": "central", "expression": "134 * 0.01 * 0.5 * 30.0", "result": 20.1, "unit": "gross healthy-life-year equivalents/year", "status": "scenario_only" }, { "scenario": "high", "expression": "134 * 0.1 * 0.8 * 40.0", "result": 428.8, "unit": "gross healthy-life-year equivalents/year", "status": "scenario_only" } ] }, "natural_outcome_estimate": { "status": "scenario_only", "result": [ { "scenario": "low", "expression": "134 * 0.001 * 0.25", "result": 0.0335, "unit": "potential deaths prevented/year", "status": "scenario_only" }, { "scenario": "central", "expression": "134 * 0.01 * 0.5", "result": 0.67, "unit": "potential deaths prevented/year", "status": "scenario_only" }, { "scenario": "high", "expression": "134 * 0.1 * 0.8", "result": 10.72, "unit": "potential deaths prevented/year", "status": "scenario_only" } ], "interpretation": "The experimental outcome is a reduction in flagged errors. The potential deaths are a secondary extrapolation, and the research-caused deployment scale is an analyst sensitivity input." }, "welfare_estimate": { "status": "scenario_only_gross", "unit": "healthy-life-year equivalents/year", "scenarios": [ { "scenario": "low", "expression": "134 * 0.001 * 0.25 * 20.0", "result": 0.67, "unit": "gross healthy-life-year equivalents/year", "status": "scenario_only" }, { "scenario": "central", "expression": "134 * 0.01 * 0.5 * 30.0", "result": 20.1, "unit": "gross healthy-life-year equivalents/year", "status": "scenario_only" }, { "scenario": "high", "expression": "134 * 0.1 * 0.8 * 40.0", "result": 428.8, "unit": "gross healthy-life-year equivalents/year", "status": "scenario_only" } ], "why_not_full_estimate": "Costs, patient morbidity, actual platform reach, counterfactual rollout, and generalization outside the study are not identified." }, "income_eta_sensitivity": { "status": "not_applicable", "note": "Avoided patient spending could be distribution-weighted later, but incidence is not reported." }, "evaluation_value": { "status": "not_estimated", "qualitative_assessment": "Potentially material because the paper is a new preprint and an evaluation could test the clinical meaning and mortality conversion before diffusion. The probability that an Unjournal evaluation changes deployment, and the outcome difference it would cause, are unknown." }, "uncertainty_and_double_counting": [ "The probability and size of a decision change caused by the research relative to the evidence already available.", "The incremental outcome versus the policy, funding, or research option displaced by that change.", "Implementation costs, harms, spillovers, and correlations among uptake, scale, and effectiveness.", "Do not add the $4.8m hospitalization-savings extrapolation to health gains without identifying who bears the costs and whether it represents real resources rather than transfers or prices." ], "sources": [ "https://arxiv.org/abs/2609.09673" ] }, { "paper_id": "pill-infant-health-us", "title": "The Multigenerational Effects of Legal Access to the Pill on Infant Health", "url": "https://www.nber.org/papers/w35722", "case_role": "Historical US policy study with measured infant-health proxies but an unmeasured research-to-current-policy link", "cause_area": "Reproductive health and infant health", "assessment_level": "structured_not_estimated", "causal_chain": { "research_output": "Historical US evidence linking legal pill access for minors to outcomes in the next generation.", "belief_or_tool_change": "Current policy analysts update estimates of long-run benefits from adolescent contraceptive access.", "decision_change": "A court, legislature, health agency, or funder changes an access rule or implementation decision because of the incremental evidence.", "natural_outcome": "Birth weight and low-birth-weight incidence among affected births; medical savings are model-derived.", "welfare_bridge": "Requires a durable health effect per changed birth plus incidence of medical savings and affected households' consumption; these are not available in the pilot source." }, "pathway_type": "coverage", "quantitative_inputs": [ { "name": "birth_weight_effect", "description": "Increase in birth weight among affected births", "unit": "grams", "provenance_tag": "source_estimate", "value": null, "low": 12, "central": 14, "high": 16, "source_url": "https://www.nber.org/papers/w35722", "note": "" }, { "name": "low_birth_weight_reduction", "description": "Relative reduction in low birth weight among affected births", "unit": "proportion", "provenance_tag": "source_estimate", "value": null, "low": 0.027, "central": 0.0305, "high": 0.034, "source_url": "https://www.nber.org/papers/w35722", "note": "" }, { "name": "medical_savings_2000_black_cohort", "description": "Estimated medical savings for the 2000 Black birth cohort", "unit": "2025 USD", "provenance_tag": "source_estimate", "value": null, "low": 113800000, "central": 114550000, "high": 115300000, "source_url": "https://www.nber.org/papers/w35722", "note": "This is a modeled cost result, not cash paid to beneficiaries." }, { "name": "research_caused_policy_change", "description": "Current or future policy coverage caused by this paper beyond the prior literature", "unit": "affected births", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "durable_health_gain", "description": "Lifetime health effect per affected birth", "unit": "healthy-year equivalents/birth", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "cost_incidence", "description": "Who receives or bears the medical-resource savings", "unit": "distribution", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" } ], "arithmetic": { "formula": "affected births caused by the research × incremental health outcome per affected birth × duration × welfare weight − costs/harms", "scenarios": [ { "scenario": "conditional_1_percent", "expression": "$113.8m–$115.3m × 0.01", "result": 1145500, "unit": "midpoint medical-cost savings per 1% of a 2000-Black-birth-cohort-equivalent policy effect", "status": "conditional_only" } ] }, "natural_outcome_estimate": { "status": "source_policy_effect_only", "result": { "birth_weight_grams": [ 12, 16 ], "relative_low_birth_weight_reduction": [ 0.027, 0.034 ] }, "interpretation": "These are effects of historical legal access, not effects caused by publication of this paper." }, "welfare_estimate": { "status": "not_estimated", "why_not_full_estimate": "The incremental current decision effect, number and distribution of affected births, durable health consequences, and cost incidence are unknown." }, "income_eta_sensitivity": { "status": "not_estimated", "note": "A US-wide or within-US disadvantage label is not a beneficiary consumption distribution. Medical savings cannot be treated as household consumption without incidence." }, "evaluation_value": { "status": "not_estimated", "qualitative_assessment": "A new evaluation could test identification and the cost conversion while revisions remain possible. Its expected welfare value still needs the probability that the evaluation changes a live policy or corrects a consequential error." }, "uncertainty_and_double_counting": [ "The probability and size of a decision change caused by the research relative to the evidence already available.", "The incremental outcome versus the policy, funding, or research option displaced by that change.", "Implementation costs, harms, spillovers, and correlations among uptake, scale, and effectiveness.", "Do not count birth-weight gains, avoided low birth weight, medical savings, later education, and earnings as independent benefits without modeling their overlap." ], "sources": [ "https://www.nber.org/papers/w35722" ] }, { "paper_id": "tracking-inequality-italy", "title": "Tracking Inequality: Teachers and the Allocation of Educational Opportunities", "url": "https://www.nber.org/papers/w35701", "case_role": "High-income-country education intervention with a concrete mechanism but no accessible long-run welfare effect", "cause_area": "Education and inequality", "assessment_level": "structured_not_estimated", "causal_chain": { "research_output": "Italian administrative, vignette, belief-elicitation, and field evidence on teacher recommendations.", "belief_or_tool_change": "Education systems update beliefs about forecast error and use feedback on recommendations.", "decision_change": "Additional teachers receive feedback and change recommendations beyond the counterfactual.", "natural_outcome": "Recommendations and demanding-track enrollment for high-achieving disadvantaged students; no detectable short-run academic harm in the abstract.", "welfare_bridge": "Requires completion, later consumption, wellbeing, displacement, and beneficiary consumption; these are unmeasured in the accessible pilot source." }, "pathway_type": "coverage", "quantitative_inputs": [ { "name": "eligible_population", "description": "High-achieving disadvantaged students reached because of the research", "unit": "students", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "research_caused_coverage_change", "description": "Additional share of teachers receiving effective feedback because of the research", "unit": "proportion", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "incremental_enrollment", "description": "Increase in demanding-track enrollment per newly covered student", "unit": "proportion", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": "https://www.nber.org/papers/w35701", "note": "The NBER abstract reports a direction but not the coefficient." }, { "name": "long_run_consumption_effect", "description": "Causal consumption gain from changed tracking, net of redistribution and displacement", "unit": "real PPP/person/year", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "duration", "description": "Duration of the long-run consumption or wellbeing effect", "unit": "years", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" } ], "arithmetic": { "formula": "eligible students × research-caused coverage change × incremental enrollment × long-run outcome × duration × welfare weight − displacement/costs", "scenarios": [ { "scenario": "all", "status": "not_estimated", "reason": "Every scale-to-welfare term is currently unknown." } ] }, "natural_outcome_estimate": { "status": "not_estimated", "interpretation": "The source abstract establishes direction and subgroup concentration, but does not provide the coefficients needed for a natural-unit rollout calculation." }, "welfare_estimate": { "status": "not_estimated", "why_not_full_estimate": "Short-run enrollment is not a lifetime benefit. Italy's income level and within-country disadvantage do not by themselves identify welfare weight or global opportunity cost." }, "income_eta_sensitivity": { "status": "not_estimated", "required_inputs": [ "beneficiary baseline real consumption", "causal consumption effect", "duration", "displacement effects" ] }, "evaluation_value": { "status": "not_estimated", "qualitative_assessment": "Independent review may be useful while the working paper is new, particularly on subgroup selection and long-run claims. No evidence yet quantifies the chance that review changes implementation or the value of the resulting decision." }, "uncertainty_and_double_counting": [ "The probability and size of a decision change caused by the research relative to the evidence already available.", "The incremental outcome versus the policy, funding, or research option displaced by that change.", "Implementation costs, harms, spillovers, and correlations among uptake, scale, and effectiveness.", "Selective-track entry may partly redistribute scarce places or positional earnings; those effects should be netted out rather than counted as gross social gain." ], "sources": [ "https://www.nber.org/papers/w35701", "https://www.socialscienceregistry.org/trials/10758" ] }, { "paper_id": "cash-transfers-kenya-ge", "title": "General Equilibrium Effects of Cash Transfers: Experimental Evidence From Kenya", "url": "https://doi.org/10.3982/ECTA17945", "case_role": "Direct LMIC consumption intervention with enough outcome data for a conditional distribution-sensitive conversion", "cause_area": "Global development and consumption", "assessment_level": "conditional_unit_estimate_not_research_total", "causal_chain": { "research_output": "Large randomized cash-transfer saturation experiment with recipient and local spillover measurement.", "belief_or_tool_change": "Funders update the expected direct and general-equilibrium effects of large cash transfers.", "decision_change": "Funding shifts toward cash transfers from a specified alternative because of the incremental evidence.", "natural_outcome": "Annualized household consumption expenditure and assets; local output multiplier is not itself household welfare.", "welfare_bridge": "CRRA utility sensitivity over per-person consumption, normalized to a doubling from $1,000 PPP per person-year." }, "pathway_type": "funding", "quantitative_inputs": [ { "name": "transfer_households", "description": "Poor households receiving about $1,000 nominal", "unit": "households", "provenance_tag": "source_estimate", "value": 10500, "low": null, "central": null, "high": null, "source_url": "https://doi.org/10.3982/ECTA17945", "note": "" }, { "name": "household_size", "description": "Average baseline household size", "unit": "people/household", "provenance_tag": "source_estimate", "value": 4.3, "low": null, "central": null, "high": null, "source_url": "https://doi.org/10.3982/ECTA17945", "note": "" }, { "name": "baseline_consumption", "description": "Control low-saturation mean annualized household expenditure", "unit": "USD PPP/household/year", "provenance_tag": "source_estimate", "value": 2536.01, "low": null, "central": null, "high": null, "source_url": "https://doi.org/10.3982/ECTA17945", "note": "" }, { "name": "recipient_consumption_gain", "description": "Total recipient-household effect including spatial spillovers", "unit": "USD PPP/household/year", "provenance_tag": "source_estimate", "value": 338.57, "low": null, "central": null, "high": null, "source_url": "https://doi.org/10.3982/ECTA17945", "note": "" }, { "name": "duration", "description": "Years for which this measured annualized gain persists", "unit": "years", "provenance_tag": "analyst_assumption", "value": null, "low": 1, "central": 1, "high": 1, "source_url": null, "note": "The pilot uses one year only. Longer-run persistence is a separate empirical input." }, { "name": "reference_consumption", "description": "Normalization baseline for consumption-doubling equivalents", "unit": "USD PPP/person/year", "provenance_tag": "analyst_assumption", "value": 1000, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "relevant_budget", "description": "Budget whose allocation is changed by this research", "unit": "USD", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "research_caused_allocation_share", "description": "Counterfactual allocation change caused by the research", "unit": "proportion", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "incremental_return_vs_displaced_alternative", "description": "Cash-transfer outcome minus the displaced option's outcome", "unit": "welfare units/USD", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" } ], "arithmetic": { "formula": "relevant budget × research-caused allocation-share change × incremental welfare per dollar versus displaced alternative", "conditional_conversion": "For a recipient household: 4.3 × [u_eta((2536.01+338.57)/4.3) − u_eta(2536.01/4.3)] / [u_eta(2000) − u_eta(1000)]", "scenarios": [ { "eta": 1.0, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 0.7774002138482107, "reference_consumption_doubling_equivalents_per_1000_households_year": 777.4002138482107, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 338.00009297748295, "status": "conditional_gross_conversion" }, { "eta": 1.5, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 1.1610601826524947, "reference_consumption_doubling_equivalents_per_1000_households_year": 1161.0601826524946, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 504.8087750663021, "status": "conditional_gross_conversion" }, { "eta": 2.0, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 1.7174733980412717, "reference_consumption_doubling_equivalents_per_1000_households_year": 1717.4733980412718, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 746.7275643657704, "status": "conditional_gross_conversion" } ] }, "natural_outcome_estimate": { "status": "conditional_gross_conversion", "result": { "annualized_consumption_gain_ppp_per_recipient_household": 338.57, "per_1000_similar_recipient_households_ppp": 338570.0 }, "interpretation": "The paper also estimates non-recipient gains and a 2.58 expenditure multiplier. Those are excluded here to avoid treating output as welfare and double counting the same spending chain." }, "welfare_estimate": { "status": "conditional_gross_conversion", "unit": "reference-consumption-doubling equivalents per year", "scenarios": [ { "eta": 1.0, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 0.7774002138482107, "reference_consumption_doubling_equivalents_per_1000_households_year": 777.4002138482107, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 338.00009297748295, "status": "conditional_gross_conversion" }, { "eta": 1.5, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 1.1610601826524947, "reference_consumption_doubling_equivalents_per_1000_households_year": 1161.0601826524946, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 504.8087750663021, "status": "conditional_gross_conversion" }, { "eta": 2.0, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 1.7174733980412717, "reference_consumption_doubling_equivalents_per_1000_households_year": 1717.4733980412718, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 746.7275643657704, "status": "conditional_gross_conversion" } ], "why_not_full_estimate": "The total scale caused by the paper, duration, program cost, and return of the displaced alternative are missing." }, "income_eta_sensitivity": { "status": "calculated", "eta_values": [ { "eta": 1.0, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 0.7774002138482107, "reference_consumption_doubling_equivalents_per_1000_households_year": 777.4002138482107, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 338.00009297748295, "status": "conditional_gross_conversion" }, { "eta": 1.5, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 1.1610601826524947, "reference_consumption_doubling_equivalents_per_1000_households_year": 1161.0601826524946, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 504.8087750663021, "status": "conditional_gross_conversion" }, { "eta": 2.0, "reference_consumption_ppp_per_person_year": 1000.0, "baseline_consumption_ppp_per_person_year": 589.7697674418605, "gain_ppp_per_person_year": 78.73720930232558, "reference_consumption_doubling_equivalents_per_household_year": 1.7174733980412717, "reference_consumption_doubling_equivalents_per_1000_households_year": 1717.4733980412718, "health_year_equivalents_per_1000_households_year_under_0_435_bridge": 746.7275643657704, "status": "conditional_gross_conversion" } ], "interpretation": "Because baseline consumption is below the $1,000 reference, greater curvature raises this gain relative to a doubling at the reference level. This is a normalization-sensitive moral comparison, not an empirical effect." }, "evaluation_value": { "status": "not_estimated_low_incremental_case", "qualitative_assessment": "The paper is published in Econometrica, has a public replication package, is already used in GiveWell's cash analysis, and the dashboard records a historical deprioritization. A conventional new evaluation is unlikely to be the best marginal scrutiny project unless it targets the 2026 corrigendum or contested long-run evidence." }, "uncertainty_and_double_counting": [ "The probability and size of a decision change caused by the research relative to the evidence already available.", "The incremental outcome versus the policy, funding, or research option displaced by that change.", "Implementation costs, harms, spillovers, and correlations among uptake, scale, and effectiveness.", "Do not add the local output multiplier to household consumption utility: expenditures, revenues, wages, and profits can be different views of the same circulation of funds." ], "sources": [ "https://doi.org/10.3982/ECTA17945", "https://zenodo.org/records/16548593", "https://www.givewell.org/international/technical/programs/givedirectly-cash-for-poverty-relief-program" ] }, { "paper_id": "vegetarian-availability-cafeterias", "title": "Impact of increasing vegetarian availability on meal selection and sales in cafeterias", "url": "https://doi.org/10.1073/pnas.1907207116", "case_role": "Observed-choice animal-welfare intervention with a clear behavioral outcome but missing production and species composition", "cause_area": "Animal welfare and food systems", "assessment_level": "quantified_natural_outcome_with_conditional_animal_illustration", "causal_chain": { "research_output": "Observational and randomized cafeteria evidence on doubling vegetarian availability.", "belief_or_tool_change": "Caterers and advocates update beliefs about a low-friction availability intervention using observed purchases.", "decision_change": "More cafeterias change menus because of the evidence.", "natural_outcome": "Fewer meat meals selected; the randomized study estimates a 7.8 percentage-point change.", "welfare_bridge": "Requires meat species and mass, supply response, animal-life duration, realized welfare relative to neutral, and animal products in substitutes." }, "pathway_type": "coverage", "quantitative_inputs": [ { "name": "study_meals", "description": "Meals observed across three Cambridge cafeterias", "unit": "meals", "provenance_tag": "source_estimate", "value": 94644, "low": null, "central": null, "high": null, "source_url": "https://doi.org/10.1073/pnas.1907207116", "note": "" }, { "name": "experimental_meals", "description": "Meals in the randomized cafeteria study", "unit": "meals", "provenance_tag": "source_estimate", "value": 7712, "low": null, "central": null, "high": null, "source_url": "https://doi.org/10.1073/pnas.1907207116", "note": "" }, { "name": "experimental_effect", "description": "Increase in vegetarian sales when availability doubled from 25% to 50%", "unit": "percentage points", "provenance_tag": "source_estimate", "value": 7.8, "low": null, "central": null, "high": null, "source_url": "https://doi.org/10.1073/pnas.1907207116", "note": "" }, { "name": "research_caused_meal_opportunities", "description": "Meal choices newly exposed to the intervention because of this research", "unit": "meals", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "meat_mass_per_meal", "description": "Chicken meat displaced in the chicken-only illustration", "unit": "kg/meal", "provenance_tag": "analyst_assumption", "value": 0.15, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "edible_yield_per_chicken", "description": "Edible meat yield in the chicken-only illustration", "unit": "kg/bird", "provenance_tag": "analyst_assumption", "value": 1.8, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "chicken_lifespan", "description": "Chicken lifespan in the illustration", "unit": "days", "provenance_tag": "analyst_assumption", "value": 42.0, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "production_response", "description": "Farm production reduction per unit of demand reduction", "unit": "proportion", "provenance_tag": "analyst_assumption", "value": null, "low": 0.25, "central": 0.5, "high": 1.0, "source_url": null, "note": "" }, { "name": "welfare_deficit_fraction", "description": "Share of the negative chicken welfare range averted by preventing the modeled production", "unit": "proportion", "provenance_tag": "analyst_assumption", "value": null, "low": 0.1, "central": 0.5, "high": 1.0, "source_url": null, "note": "" }, { "name": "chicken_welfare_range", "description": "Probability-of-sentience and subjective-time-adjusted chicken welfare range relative to humans", "unit": "relative range", "provenance_tag": "source_estimate", "value": null, "low": 0.002, "central": 0.332, "high": 0.869, "source_url": "https://rethinkpriorities.org/research-area/welfare-range-estimates/", "note": "RP 5th/50th/95th percentiles. Sentience is already included, so it is not multiplied again." } ], "arithmetic": { "formula": "meal opportunities × 0.078 fewer meat meals/opportunity × meat kg/meal ÷ edible kg/bird × life-years/bird × production response × welfare-deficit fraction × RP welfare range", "natural_unit_reference": { "scenario": "conditional_100000_meals", "expression": "100000 × 0.078", "result": 7800.0, "unit": "fewer meat meals", "status": "conditional_only" }, "chicken_only_scenarios": [ { "scenario": "low", "expression": "100000 * 0.078 * 0.15 / 1.8 * 42/365 * 0.25 * 0.1 * 0.002", "result": 0.0037397260273972607, "unit": "illustrative human-healthy-year welfare equivalents per 100,000 meal opportunities", "status": "scenario_only" }, { "scenario": "central", "expression": "100000 * 0.078 * 0.15 / 1.8 * 42/365 * 0.5 * 0.5 * 0.332", "result": 6.207945205479453, "unit": "illustrative human-healthy-year welfare equivalents per 100,000 meal opportunities", "status": "scenario_only" }, { "scenario": "high", "expression": "100000 * 0.078 * 0.15 / 1.8 * 42/365 * 1.0 * 1.0 * 0.869", "result": 64.99643835616439, "unit": "illustrative human-healthy-year welfare equivalents per 100,000 meal opportunities", "status": "scenario_only" } ] }, "natural_outcome_estimate": { "status": "conditional", "result": { "per_100000_exposed_meals": 7800.0, "unit": "fewer meat meals" }, "interpretation": "This directly scales the randomized 7.8 percentage-point estimate and assumes the effect transports to the new setting." }, "welfare_estimate": { "status": "scenario_only_not_species_complete", "unit": "human-healthy-year welfare equivalents per 100,000 meal opportunities", "scenarios": [ { "scenario": "low", "expression": "100000 * 0.078 * 0.15 / 1.8 * 42/365 * 0.25 * 0.1 * 0.002", "result": 0.0037397260273972607, "unit": "illustrative human-healthy-year welfare equivalents per 100,000 meal opportunities", "status": "scenario_only" }, { "scenario": "central", "expression": "100000 * 0.078 * 0.15 / 1.8 * 42/365 * 0.5 * 0.5 * 0.332", "result": 6.207945205479453, "unit": "illustrative human-healthy-year welfare equivalents per 100,000 meal opportunities", "status": "scenario_only" }, { "scenario": "high", "expression": "100000 * 0.078 * 0.15 / 1.8 * 42/365 * 1.0 * 1.0 * 0.869", "result": 64.99643835616439, "unit": "illustrative human-healthy-year welfare equivalents per 100,000 meal opportunities", "status": "scenario_only" } ], "why_not_full_estimate": "The paper does not report the species and mass of meat displaced, production response, animal welfare state, or added egg/dairy harms in vegetarian meals." }, "income_eta_sensitivity": { "status": "not_applicable" }, "evaluation_value": { "status": "not_estimated", "qualitative_assessment": "Observed choices are more informative than hypothetical preferences, but this 2019 PNAS paper has had years for scrutiny. A new evaluation would need a specific live user or a synthesis role to have a strong counterfactual case." }, "uncertainty_and_double_counting": [ "The probability and size of a decision change caused by the research relative to the evidence already available.", "The incremental outcome versus the policy, funding, or research option displaced by that change.", "Implementation costs, harms, spillovers, and correlations among uptake, scale, and effectiveness.", "The chicken-only illustration is not a claim about the actual menu. Do not add separate climate or health benefits unless overlap and attribution are modeled." ], "sources": [ "https://doi.org/10.1073/pnas.1907207116", "https://doi.org/10.17863/CAM.41328", "https://rethinkpriorities.org/research-area/welfare-range-estimates/" ] }, { "paper_id": "xpt-existential-risk", "title": "Forecasting Existential Risks: Evidence From a Long-Run Forecasting Tournament", "url": "https://forecastingresearch.org/research/existential-risk-persuasion-tournament", "case_role": "Catastrophic-risk research tool with enormous conditional stakes but no identified research-attributable risk change", "cause_area": "Catastrophic risk and forecasting", "assessment_level": "conditional_risk_sensitivity_not_welfare_estimate", "causal_chain": { "research_output": "A multi-stage tournament eliciting forecasts and rationales from 80 domain experts and 89 superforecasters.", "belief_or_tool_change": "Decision-makers learn the scale and structure of disagreement, and forecasting practice improves.", "decision_change": "Risk-reduction funding or policy changes relative to the best available evidence without XPT.", "natural_outcome": "The paper reports beliefs, including 6% versus 1% median extinction forecasts by 2100; it does not observe catastrophe-risk reduction.", "welfare_bridge": "Research-attributable absolute risk reduction × welfare at stake, with present-generation and long-future terms kept separate." }, "pathway_type": "research_tool", "quantitative_inputs": [ { "name": "experts", "description": "Participating domain experts", "unit": "people", "provenance_tag": "source_estimate", "value": 80, "low": null, "central": null, "high": null, "source_url": "https://forecastingresearch.org/research/existential-risk-persuasion-tournament", "note": "" }, { "name": "superforecasters", "description": "Participating superforecasters", "unit": "people", "provenance_tag": "source_estimate", "value": 89, "low": null, "central": null, "high": null, "source_url": "https://forecastingresearch.org/research/existential-risk-persuasion-tournament", "note": "" }, { "name": "expert_extinction_forecast", "description": "Median expert total extinction forecast by 2100", "unit": "probability", "provenance_tag": "source_estimate", "value": 0.06, "low": null, "central": null, "high": null, "source_url": "https://forecastingresearch.org/research/existential-risk-persuasion-tournament", "note": "" }, { "name": "superforecaster_extinction_forecast", "description": "Median superforecaster total extinction forecast by 2100", "unit": "probability", "provenance_tag": "source_estimate", "value": 0.01, "low": null, "central": null, "high": null, "source_url": "https://forecastingresearch.org/research/existential-risk-persuasion-tournament", "note": "" }, { "name": "research_attributable_risk_change", "description": "Absolute catastrophe or extinction risk reduction caused by the research through decisions", "unit": "probability", "provenance_tag": "unknown", "value": null, "low": null, "central": null, "high": null, "source_url": null, "note": "The 1% and 6% forecasts are beliefs about risk, not effects on risk." }, { "name": "present_generation_population", "description": "Round reference population in the present-generation illustration", "unit": "people", "provenance_tag": "analyst_assumption", "value": 8000000000.0, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "present_generation_remaining_years", "description": "Average remaining healthy-year reference", "unit": "years/person", "provenance_tag": "analyst_assumption", "value": 30, "low": null, "central": null, "high": null, "source_url": null, "note": "" }, { "name": "long_future_welfare_at_stake", "description": "Moral and empirical scenarios for future welfare conditional on survival", "unit": "healthy-year equivalents", "provenance_tag": "analyst_assumption", "value": null, "low": 1000000000000000.0, "central": 1e+20, "high": 1e+30, "source_url": null, "note": "Illustrative orders of magnitude only; no central moral estimate is asserted." } ], "arithmetic": { "formula": "research-attributable absolute risk reduction × welfare at stake", "present_generation_sensitivity": [ { "absolute_risk_reduction": 1e-09, "welfare_at_stake_hly": 240000000000.0, "result_hly": 240.00000000000003 }, { "absolute_risk_reduction": 1e-07, "welfare_at_stake_hly": 240000000000.0, "result_hly": 24000.0 }, { "absolute_risk_reduction": 1e-05, "welfare_at_stake_hly": 240000000000.0, "result_hly": 2400000.0 } ], "long_future_sensitivity_matrix": [ { "absolute_risk_reduction": 1e-09, "welfare_at_stake_hly": 1000000000000000.0, "result_hly": 1000000.0000000001 }, { "absolute_risk_reduction": 1e-09, "welfare_at_stake_hly": 1e+20, "result_hly": 100000000000.0 }, { "absolute_risk_reduction": 1e-09, "welfare_at_stake_hly": 1e+30, "result_hly": 1.0000000000000001e+21 }, { "absolute_risk_reduction": 1e-07, "welfare_at_stake_hly": 1000000000000000.0, "result_hly": 100000000.0 }, { "absolute_risk_reduction": 1e-07, "welfare_at_stake_hly": 1e+20, "result_hly": 10000000000000.0 }, { "absolute_risk_reduction": 1e-07, "welfare_at_stake_hly": 1e+30, "result_hly": 1e+23 }, { "absolute_risk_reduction": 1e-05, "welfare_at_stake_hly": 1000000000000000.0, "result_hly": 10000000000.0 }, { "absolute_risk_reduction": 1e-05, "welfare_at_stake_hly": 1e+20, "result_hly": 1000000000000000.1 }, { "absolute_risk_reduction": 1e-05, "welfare_at_stake_hly": 1e+30, "result_hly": 1e+25 } ] }, "natural_outcome_estimate": { "status": "source_beliefs_only", "result": { "expert_median_extinction_probability_2100": 0.06, "superforecaster_median_extinction_probability_2100": 0.01 }, "interpretation": "The disagreement is an output of the research. It is not an estimate that XPT reduces risk by five percentage points." }, "welfare_estimate": { "status": "not_estimated", "conditional_sensitivity": { "present_generation": [ { "absolute_risk_reduction": 1e-09, "welfare_at_stake_hly": 240000000000.0, "result_hly": 240.00000000000003 }, { "absolute_risk_reduction": 1e-07, "welfare_at_stake_hly": 240000000000.0, "result_hly": 24000.0 }, { "absolute_risk_reduction": 1e-05, "welfare_at_stake_hly": 240000000000.0, "result_hly": 2400000.0 } ], "long_future": [ { "absolute_risk_reduction": 1e-09, "welfare_at_stake_hly": 1000000000000000.0, "result_hly": 1000000.0000000001 }, { "absolute_risk_reduction": 1e-09, "welfare_at_stake_hly": 1e+20, "result_hly": 100000000000.0 }, { "absolute_risk_reduction": 1e-09, "welfare_at_stake_hly": 1e+30, "result_hly": 1.0000000000000001e+21 }, { "absolute_risk_reduction": 1e-07, "welfare_at_stake_hly": 1000000000000000.0, "result_hly": 100000000.0 }, { "absolute_risk_reduction": 1e-07, "welfare_at_stake_hly": 1e+20, "result_hly": 10000000000000.0 }, { "absolute_risk_reduction": 1e-07, "welfare_at_stake_hly": 1e+30, "result_hly": 1e+23 }, { "absolute_risk_reduction": 1e-05, "welfare_at_stake_hly": 1000000000000000.0, "result_hly": 10000000000.0 }, { "absolute_risk_reduction": 1e-05, "welfare_at_stake_hly": 1e+20, "result_hly": 1000000000000000.1 }, { "absolute_risk_reduction": 1e-05, "welfare_at_stake_hly": 1e+30, "result_hly": 1e+25 } ] }, "why_not_full_estimate": "The research-attributable change in action and risk is unknown; long-future population, welfare, discounting, and moral weights are worldview choices." }, "income_eta_sensitivity": { "status": "not_applicable" }, "evaluation_value": { "status": "already_evaluated", "qualitative_assessment": "The Unjournal already published two evaluations and a synthesis in 2024. Commissioning the same evaluation again has little incremental value absent a distinct update, such as resolved near-term questions or a new decision-use study." }, "uncertainty_and_double_counting": [ "The probability and size of a decision change caused by the research relative to the evidence already available.", "The incremental outcome versus the policy, funding, or research option displaced by that change.", "Implementation costs, harms, spillovers, and correlations among uptake, scale, and effectiveness.", "Keep present-generation and long-future value separate. Do not treat disagreement, forecast level, funding volume, or citations as risk reduction." ], "sources": [ "https://forecastingresearch.org/research/existential-risk-persuasion-tournament", "https://unjournal.pubpub.org/pub/evalsumforecastingexistentialrisk/release/7" ] } ], "cross_case_assessment": { "defensible_total_welfare_estimates": 0, "conditional_or_sensitivity_calculations": 4, "structured_not_estimated": 2, "implication_for_current_0_to_10_scores": "The pilot does not recover or validate the existing 0-10 impact ratings. It shows why a decimal rating cannot be described as a calculated welfare estimate without additional evidence and explicit moral inputs." }, "framework_sources": [ "https://www.givewell.org/international/technical/programs/givedirectly-cash-for-poverty-relief-program", "https://www.givewell.org/how-we-work/our-criteria/cost-effectiveness/moral-weights", "https://rethinkpriorities.org/research-area/an-introduction-to-the-moral-weight-project/", "https://rethinkpriorities.org/research-area/welfare-range-estimates/" ] }