Treating people as data entrenches inequality: a multidisciplinary review finds that mainstream AI and design practices simplify sociocultural complexity into labels and metrics, reproducing historical injustices and shifting harms onto multiply marginalized groups; the paper urges participatory, disaggregated impact assessment and intersectional welfare frameworks for policy and economic analysis.

Main Finding

Applying intersectionality as an explicit analytic and methodological lens reveals that dominant computing and digital-media practices systematically reduce sociocultural complexity to data, code, and metrics. That reduction both stems from and reinforces epistemic limits in technology design and deployment, producing dehumanizing decisions and reifying longstanding social inequities rather than correcting them.

Key Points

• Intersectionality as organizing theory: Intersectionality foregrounds how multiple axes of identity and power (race, gender, class, disability, nationality, etc.) interact; using it in digital studies shifts attention from single-axis harms to compounded, context-dependent harms.
• Reductionism in computing: Engineering and data practices often treat people and culture as tractable, fungible inputs (labels, features, metrics). That reification erases context, nuance, and relational power dynamics.
• Political/ethical consequences: Framing social complexity as “solvable” by algorithms rationalizes decisions that are political (who counts, what gets measured) and ethical (who benefits, who is harmed), frequently producing disproportionate harms to multiply marginalized groups.
• Reproduction of inequity: Data, models, and design choices rarely interrogate historical and structural causes; instead they can reproduce and legitimize existing social hierarchies under veneers of efficiency and objectivity.
• Methodological gains from critical fields: Critical race studies, feminist tech studies, and digital anthropology provide epistemological tools (e.g., contextualized empiricism, participatory methods, narrative and ethnographic approaches) to reveal hidden assumptions in technological systems.
• Design and governance implications: The literature advocates for upstream consideration of people and environment (not only downstream mitigation), participatory co-design, and governance structures that mandate contextualized impact assessment.

Data & Methods

• Nature of the article: A multi-disciplinary literature review and theoretical synthesis rather than new quantitative empirical data.
• Domains surveyed: Critical race studies, feminist and queer theory, digital media studies, science & technology studies (STS), critical algorithm studies, and design research.
• Methods reported in the reviewed work:
  • Qualitative methods: ethnography, interviews, participant observation, narrative analysis.
  • Discursive and historical analyses: tracing the genealogy of technological practices and their social meanings.
  • Critical design and participatory action: co-design with affected communities, interventionist prototyping.
  • Algorithmic audits and case studies: empirical examinations exposing bias and failure modes.
  • Mixed-methods: combining qualitative contextualization with quantitative measurement, when present.
• Limits: The article synthesizes interpretive and critical scholarship rather than offering large-scale causal identification or new quantitative measurement. Empirical generalizability depends on the scope of reviewed case studies and disciplinary perspectives.

Implications for AI Economics

• Measurement and valuation
  • Standard metrics (clicks, engagement, LTV) omit contextual harms and non-market values (care, community cohesion). This leads to biased welfare assessments and mispriced goods/services.
  • Heterogeneous preferences and constraints across intersectional groups imply demand elasticities and willingness-to-pay estimates can be systematically biased if intersectional heterogeneity is ignored.
• Market outcomes and inequality
  • Algorithmic personalization and platform allocation can amplify segregation and unequal access to opportunities, reinforcing inequality along intersecting identity axes.
  • Labor markets (gig platforms, microwork) often externalize risks onto marginalized workers; intersectional analysis reveals differential bargaining power, exposure to surveillance, and precarity.
• Externalities and social cost
  • Omitted social costs (privacy harms, civic disempowerment, mental-health impacts, environmental burdens) are likely concentrated on marginalized communities—standard cost–benefit analyses understate true social costs.
• Risk, pricing, and insurance
  • Risk models trained on historical data may underprice systemic risks to vulnerable groups and misestimate tail risks stemming from structural vulnerabilities.
• Market power and governance
  • Epistemic capture: firms that control datasets and ontologies shape what is legible and valuable, potentially entrenching market power and blocking participatory governance.
  • Regulation designed around narrow notions of fairness or efficiency can miss intersectional harms; policies should require contextualized, disaggregated impact assessments.
• Research & policy recommendations for AI economists
  • Incorporate intersectional heterogeneity into empirical models: estimate treatment effect heterogeneity across multiple interacting dimensions, not just single covariates.
  • Use mixed methods: pair quantitative causal inference with qualitative, contextual inquiry to surface mechanisms and harms that numbers alone miss.
  • Develop enriched welfare frameworks: include social weights or distributional preferences sensitive to intersectional disadvantage and non-market values.
  • Design participatory valuation methods: co-produce metrics with affected communities, and measure outcomes prioritized by those communities (e.g., trust, dignity, safety).
  • Mandate algorithmic impact assessments that disaggregate outcomes by intersecting identities and require mitigation plans for disproportionate harms.
  • Expand cost–benefit analyses to internalize externalities affecting marginalized groups and the environment.
  • Promote multidisciplinary teams and governance: economists should partner with critical scholars, designers, and community representatives to ensure analyses capture sociotechnical realities.
• Open research questions
  • How to operationalize intersectional welfare weights in standard economic models?
  • Which econometric techniques best detect high-dimensional intersectional heterogeneity with realistic sample sizes?
  • What institutional designs (procurement rules, data trusts, impact assessments) most effectively translate intersectional analysis into practice?

Concluding note: For AI economics to produce accurate, just, and policy-relevant insights, it must move beyond purely technical efficiency metrics and embed intersectional, contextualized analysis into measurement, modeling, and governance.