EU AI Subsidies Face a New Test: Measurable Happiness Gains
European Union technology policy has spent years trying to tame the risks of artificial intelligence. This paper argues that risk control alone is no longer enough. If public money, regulatory privilege, and access to shared infrastructure help sustain the AI industry, then policymakers should demand a public return measured not only in growth or efficiency, but in human happiness.
The proposal is deliberately provocative. Instead of treating social benefit as an abstract side effect of innovation, the authors argue that firms receiving subsidies should be required to predict concrete, measurable improvements in wellbeing and then be held accountable for whether those gains actually materialize. The framework is aimed at general-purpose AI systems, which are often marketed for economic productivity while their non-financial effects receive far less scrutiny.
A Policy Debates Centers on Public Value, Not Just Risk Control
The article frames current EU AI policy as structurally imbalanced. Regulation has largely concentrated on preventing harm, yet the industry also draws on substantial indirect support through tax breaks, public funding, privileged access to public lands, resources, data, infrastructure, and deregulatory advantages. In that setting, a narrow focus on mitigating downside risk can leave a much larger question unanswered: what public benefit justifies the subsidy in the first place?
The argument is not that safety should be relaxed. It is that public support should be contingent on delivering socially beneficial technologies rather than merely avoiding catastrophic misuse. A system can be legally compliant, commercially successful, and still offer little genuine value to the people financing it. In practical terms, that means subsidy decisions should ask what positive human outcomes an AI system is expected to produce, how those outcomes will be measured, and what happens if the promised gains fail to appear.
One sentence captures the tension sharply: public investment without public wellbeing is a one-way transfer.
Happiness Becomes the Third Criterion for AI Funding
The authors propose adding happiness as a third goal in EU AI policymaking, alongside the familiar emphasis on innovation and risk mitigation. The choice is not casual. Happiness is presented as a socially meaningful outcome that can anchor policy in lived human experience rather than in abstract performance claims. In a regulatory environment where firms can invoke productivity, convenience, or competitive advantage with little obligation to prove broader benefit, happiness introduces a direct test of whether AI improves daily life.
To make the concept workable, the paper turns to decades of psychological research. Rather than treating happiness as a vague sentiment, the framework relies on three widely accepted determinants: relatedness, competence, and autonomy. These traits give policymakers a more operational way to think about wellbeing. Relatedness concerns human connection and belonging. Competence reflects whether people feel effective and capable in their tasks. Autonomy captures the degree of control people retain over their decisions and actions.
The appeal of this structure is obvious. AI products that increase isolation, erode mastery, or reduce agency may raise output while quietly damaging wellbeing. Systems that support social connection, skill development, and meaningful choice would score differently. Happiness, in this model, is not a soft afterthought. It becomes a policy-relevant outcome tied to recognizable psychological mechanisms.
Relatedness, Competence, and Autonomy Offer a Practical Lens
The paper’s psychological framing matters because it avoids the trap of reducing wellbeing to a single subjective score. Happiness is multifactorial, and the three-part model provides a more grounded way to assess whether AI tools are improving or diminishing the conditions that support it. An application that automates away human interaction may be efficient yet socially hollow. A system that makes workers more capable without stripping away control can plausibly support both productivity and wellbeing. A platform that constrains choice while amplifying dependence may deliver commercial value at the expense of human autonomy.
This is where the policy argument becomes more exacting. If companies receive public support, they should not be allowed to define success solely by market adoption or technical performance. They should have to state the wellbeing mechanism in advance. Does the system strengthen relationships? Does it help users build competence? Does it preserve autonomy rather than narrow it? Those questions are concrete enough to be incorporated into subsidy criteria, even if the answers will require careful measurement.
A second blunt point follows from that logic: an AI system that cannot justify its social value should not expect public subsidy by default.
Accountability Would Shift the Burden of Proof onto Industry
The paper’s policy design places the burden on companies to predict and demonstrate benefit before receiving support. That would be a notable shift from the current model, in which firms often receive public backing on the assumption that innovation will eventually pay off for society. Under the proposed approach, subsidy would become conditional on a clearer social contract. Companies would need to articulate expected happiness gains in measurable terms and accept evaluation after deployment.
Such a model would not eliminate uncertainty. Happiness is harder to quantify than revenue, and psychological effects may vary by population, context, and product design. But the paper’s central claim is that uncertainty should not be an excuse for neglect. Public support already reflects a policy choice to socialize part of the risk. It follows that society can also demand a more explicit accounting of benefits.
The debate is likely to be uncomfortable for regulators and industry alike. Yet the premise is hard to dismiss: if AI development depends heavily on public subsidy, then public value should be part of the bargain. The article closes by pressing that point into EU policy architecture, where rights, values, competition, and innovation already collide. Adding happiness to the equation would not solve those tensions, but it would force them into the open and tie future support to something people can actually feel in their lives.
What the Proposal Leaves Unresolved
The paper does not provide a ready-made scoring system for happiness-based subsidies, and it does not solve the measurement problem inherent in translating wellbeing into regulatory criteria. It also leaves open how policymakers would compare happiness gains across different AI applications, or how to prevent companies from gaming the framework with shallow metrics and optimistic forecasts. Those questions are central, not peripheral.
Even so, the proposal makes a clear policy claim: AI regulation should not stop at harm prevention. If the industry is being sustained by public resources, then it should be expected to deliver demonstrable human benefit, with relatedness, competence, and autonomy serving as the core reference points for that benefit.