From sociology of quantification to ethics of quantification

The article opens by asking what price society pays for its pervasive reliance on numbers and quantification, highlighting how the COVID-19 pandemic made the demand for numbers acute and exposed limitations in models and metrics at the science–policy interface. It discusses headline-grabbing estimates (e.g., Imperial College projections) and the substantial uncertainties underpinning them, the transparency debates around models, and the risk of outsourcing political judgment to ostensibly neutral, model-generated numbers. The authors situate these concerns within broader issues such as the reproducibility crisis in science, perverse incentives, and the structural coupling of systems (science, economy, media, policy) that encourages media-driven numerification. The central research question is what common problems are shared across diverse forms of quantification (models, metrics, rankings, algorithms) and why an encompassing ethics of quantification is needed now. The purpose is to identify shared pitfalls, lock-ins, and unintended effects across domains and to outline principles and policy directions for a mature, socially responsible relationship between numbers and society.

The paper surveys interdisciplinary critiques of quantification across models, algorithms, metrics, and rankings. It contrasts perverse and virtuous quantifications, noting successful cases such as weather forecasting where uncertainty is communicated and societal use is mature, versus problematic uses in medicine, criminal justice, and higher education metrics. Drawing on Jasanoff, it distinguishes technologies of hubris (risk assessment, cost–benefit analysis promising control amid uncertainty) from technologies of humility (approaches that acknowledge ambiguity, indeterminacy, and ethical dimensions). It reviews the fragmented ethical landscape: AI ethics initiatives; statistical debates over core concepts; long-standing critiques of metric misuse; and the non-disciplinary nature of mathematical modeling. Espeland and Stevens’ five dimensions of quantification (work/bureaucracy, reactivity, discipline, authority, aesthetics) are highlighted, alongside calls to move from sociology to ethics of numbers. The review includes concerned readings and activism: O’Neil’s critique of algorithmic harms and labor impacts; Supiot’s governance by numbers and re-feudalization; French ‘statactivistes’ countering unjust metrics; the Data Justice Lab; the SSSQ network; hackathons revealing algorithmic bias; election auditing and anti-gerrymandering modeling; and public-facing model communication during COVID-19. It also examines quantification’s illumination/obfuscation duality (e.g., Aadhaar), and pandemic-era narrowing of attention to certain numbers at the expense of broader social costs and rights.

This is a perspective article based on conceptual analysis and a broad, cross-domain review of scholarly and activist literature concerning quantification in society, with illustrative examples from COVID-19 modeling, AI/algorithms, metrics, and rankings. The authors synthesize themes to identify common ethical issues, systemic obstacles (lock-ins, path dependencies), and propose principles and policy directions. No original empirical data or formal systematic review protocol is reported; rather, the method is argumentative synthesis drawing from interdisciplinary sources and case illustrations.

• Diverse forms of quantification (models, metrics, algorithms, rankings) share common ethical challenges: opacity, misplaced certainty, framing effects, reactivity that disciplines behavior, and the conferral of authority and power through numbers.
• Quantification is a social activity; claims of neutrality often mask value-laden choices. The technique is never neutral: methodological and disciplinary choices shape outcomes and narratives.
• The hubris–humility axis (Jasanoff) is key: hubristic quantifications promise control and exhaustiveness, often leading to overconfidence and depoliticization, whereas humble approaches foreground uncertainty, ethics, and inclusivity.
• COVID-19 exposed pitfalls of model use in high-uncertainty, high-stakes contexts and the risk of delegating political decisions to numbers. Examples include widely cited death toll projections (e.g., 510,000 UK and 2.2 million US deaths in a do-nothing scenario) and cost–benefit analyses employing value of a statistical life (e.g., US social distancing net benefit ≈ $5.2 trillion), which can obscure contested assumptions.
• There are virtuous exemplars (e.g., weather forecasting) where uncertainty communication and iterative updating enable beneficial societal integration.
• Obstacles to ethical quantification include techno-optimism, black-box algorithms, economic and institutional incentives, lock-ins and path dependencies (e.g., research assessment metrics), and the Collingridge dilemma hampering timely governance.
• Policy implications emerged across actors: organized labor’s role in contesting harmful metrics; institutional reforms to ensure independent evidence and counter ‘mercenary science’; proposals like an Office for Public Lobbying and OTA-like capabilities; and integrating responsible quantification into Europe’s RRI agenda.
• Proposed principles for an ethics of quantification: vigilance about explicit/implicit framings; using a humility–hubris compass; systematically probing missing numbers and blind spots; providing a home for data/model/stat-activism; and fostering quality as fitness for societal purpose via socially mediated, participatory approaches.

By foregrounding shared pitfalls and systemic drivers across quantification practices, the article addresses its central question: why a unified ethics of quantification is both necessary and urgent. The synthesis shows how numbers can simultaneously illuminate and obfuscate, legitimizing certain policy paths while excluding others through framing and claims of certainty. In pandemic conditions and other post-normal contexts, models influence decisions directly; thus, ethical scrutiny must encompass assumptions, uncertainties, and distributional impacts (winners/losers). The findings argue for shifting from technocratic faith in numbers toward participatory, humble, and transparent practices that embed social learning and accountability. This reframing has implications for the science–policy interface, research governance, and public trust, advocating institutional changes and cross-domain principles that can guide responsible use of quantification.

The paper consolidates calls to move from a sociology of quantification to an encompassing ethics of quantification, emphasizing commonalities across domains and the urgency heightened by COVID-19. It contributes a set of orienting principles (framing vigilance, humility–hubris compass, blind-spot probing, support for activism, and quality-as-fitness-for-purpose) and sketches policy avenues involving labor, institutions, and research governance (e.g., RRI enhancement, independent evidence infrastructures). Future work should operationalize these principles into practical guidelines, auditing frameworks, participatory processes for model/metric design and evaluation, and domain-specific codes of practice. Further research is needed on overcoming lock-ins, improving uncertainty communication, and developing institutions capable of democratic oversight of quantification in high-stakes settings.

The article is a perspective and argumentative synthesis rather than a systematic review or empirical study; coverage of literature is broad but not exhaustive. The heterogeneous domains considered (AI, statistics, modeling, metrics) are treated at a high level, and some examples are context-specific. Proposed principles and policy suggestions are normative and may face implementation challenges given institutional lock-ins, political economy constraints, and the Collingridge dilemma. The rapidly evolving context of the COVID-19 pandemic may also limit the generalizability of specific examples.