Scholars argue AI success should be measured by human flourishing, not capability, and introduce Flourishing Metrics and 'Return on Flourishing' to reorient AI governance and investment; the framework is formal and policy-oriented but remains conceptual and untested.

Main Finding

The paper argues that human flourishing — a multidimensional state of well-being spanning physical, emotional, relational, financial, character/meaning, and planetary domains — should be the primary success criterion and objective function for AI systems and post‑AGI economic design. It introduces (1) Flourishing Metrics, a scalable measurement architecture combining validated subjective measures with behavioural, administrative, and contextual indicators organized as a Flourishing State Vector, and (2) Return on Flourishing (RoF), a distribution‑sensitive evaluation metric that compares the counterfactual flourishing gains produced by an intervention (or AI system) to its resources, risks, and opportunity costs. The framework is normative and conceptual, intended as a decision architecture for investment, governance, assurance, and objective‑function design rather than as a ready‑to‑deploy reward function.

Key Points

• Motivation
  • Current AI evaluation prioritizes capability, productivity, engagement, and financial return — proxies that can be misaligned with actual human well‑being.
  • As AI scales (and potentially AGI emerges), optimizing narrow proxies risks amplifying systemic misalignment and societal harms.
• Flourishing Metrics
  • Operationalizes flourishing as a multidimensional state (e.g., happiness/life satisfaction, physical/mental health, meaning/purpose, character, social relationships, financial security, and planetary context).
  • Recommends combining subjective self‑reports with behavioural, administrative, and contextual indicators to improve validity and detect early warning signals.
  • Emphasizes distributional measurement: who gains and who loses matters; avoid compensatory aggregation that masks severe losses in particular domains/groups.
• Return on Flourishing (RoF)
  • Extends ROI logic: RoF = (counterfactual change in flourishing attributable to an intervention) / (resources, risks, opportunity costs).
  • Designed to be distribution‑sensitive and include non‑compensatory safeguards (i.e., unacceptable losses in any critical dimension or group cannot be outweighed by gains elsewhere).
  • Proposed as a means to guide corporate investment, public policy, assurance, work redesign, and post‑deployment monitoring.
• Safeguards and Governance
  • The framework requires democratic specification of weights, independent validation, empirical calibration, and protections against Goodhart effects.
  • Not presented as an automatic reward function for an autonomous AGI; rather, as a societal decision architecture and accountability tool.
• Empirical context and concerns
  • Theoretical and emerging empirical literature shows risks (e.g., sycophantic chatbots, emotional dependence, erosion of social skills).
  • Empirical evidence is sparse and often limited; a few RCTs suggest certain AI designs can harm social flourishing.
• Limitations (authors’ caveats)
  • The work is a purposive integrative synthesis, not an exhaustive systematic review or empirical estimation.
  • Proposed equations and weights are normative and uncalibrated; no field trials or randomized validations have been completed.

Data & Methods

• Research design
  • Integrative conceptual synthesis drawing from flourishing science, welfare economics, human‑centred AI, alignment literature, and AI assurance/impact assessment.
  • Purposive rather than systematic literature selection based on conceptual relevance and methodological utility.
• Analytical procedure (four stages)
• Compare established multidimensional accounts of flourishing to identify recurrent domains and measurement principles.
• Assess limits of subjective measures and the need for behavioural/administrative/contextual indicators.
• Formalize the Flourishing State Vector and construct a distribution‑sensitive aggregate flourishing function.
• Extend the formalism into Return on Flourishing and sketch how it could inform appraisal, assurance, and objective‑function design.
• Formalization
  • Flourishing is treated as a dynamic system variable (multidimensional vector; distribution‑sensitive aggregation; non‑compensatory constraints).
  • RoF is defined analytically as a counterfactual attribution of flourishing change relative to resource expenditure and risk.
• Empirical status
  • No new empirical data collected; proposed constructs and equations are presented as testable hypotheses.
  • Authors call for empirical calibration, independent validation, pilot implementations, and randomized evaluations before high‑stakes deployment.

Implications for AI Economics

• Reorienting objective functions
  • Economic systems and AI governance should move beyond productivity/financial metrics toward explicit optimization for human flourishing. This reframes what counts as economic progress, especially under high productivity/abundance scenarios.
• Investment appraisal and corporate decision‑making
  • Add RoF to appraisal toolkits: measure expected counterfactual gains in flourishing per unit of resource/risk to compare projects, not just expected ROI or productivity.
  • Use distribution‑sensitive criteria and non‑compensatory floors (e.g., no project that causes severe losses for vulnerable groups should be approved even if aggregate RoF is positive).
• Labor economics and work redesign
  • RoF provides a way to evaluate AI‑enabled task automation and workplace redesign by quantifying effects on workers’ multidimensional well‑being (beyond wages and productivity).
  • Could inform policy on retraining, job redesign, worktime, and institutional supports where automation changes the nature of meaningful work.
• Macro and post‑AGI governance
  • If AGI materially increases productive capacity, policy priorities shift from sheer output maximization to distributional allocation that maximizes flourishing (e.g., public goods, leisure, health, environmental restoration).
  • RoF could help set tradeoffs between consumption, redistribution, environmental limits, and investments in social infrastructure.
• Measurement and information infrastructures
  • Implementing RoF requires new data systems: representative subjective surveys, behavioural/admin signals, and environmental/contextual indicators, plus mechanisms for independent auditing and democratic specification of weights.
  • Economists should invest in causal evaluation methods (RCTs, quasi‑experimental designs) to estimate counterfactual flourishing impacts of AI interventions.
• Incentives and Goodhart risks
  • Careful design is needed to avoid turning flourishing metrics into brittle proxies; safeguard with distributional rules and multiple indicator types to resist gaming.
  • Incentives for firms and model developers must be aligned (e.g., procurement, regulation, tax/subsidy structures, liability regimes) so that profitability does not trump RoF.
• Policy and regulatory tools
  • RoF provides a normative basis for regulation: e.g., requiring RoF assessments for high‑impact AI deployments, mandating pilots and monitoring, or conditioning approvals on minimum RoF and distributional safeguards.
• Research agenda for AI economics
  • Empirically calibrate flourishing domains and aggregation weights across contexts and cultures.
  • Develop methods to attribute counterfactual flourishing changes to AI systems (causal inference under adoption endogeneity).
  • Design institutional governance models for democratic specification, auditing, and enforcement of RoF‑based criteria.
  • Explore macroeconomic models where the social welfare function is multidimensional (flourishing) rather than GDP‑based, and analyze long‑run growth, distribution, and stability implications.

Actionable short list for economists and policymakers - Pilot RoF assessments in corporate AI pilots and public programs; include preregistered evaluation designs. - Build measurement systems combining surveys with administrative and behavioural data; ensure representative coverage and subgroup analysis. - Require distributional reporting and set non‑compensatory safeguards for critical dimensions/groups. - Fund research to operationalize counterfactual attribution methods for flourishing impacts. - Embed RoF considerations into procurement, subsidy, and regulatory regimes to reshape incentives toward human‑centered outcomes.