Algorithmic authority alone does not secure legitimacy: AI can improve efficiency but decision legitimacy in organizations hinges on transparency, explainability and maintained human judgment; without these, algorithms may both bolster and erode stakeholders' trust.

Title: Decision Legitimacy in AI-Enabled Organizations: A Multilevel Framework of Algorithmic Authority and Human Judgment Author: Aslı Tenderis (2026), Journal of Social Science and Human Research Studies

Main Finding

The paper develops a multilevel, theoretical framework arguing that decision legitimacy in AI-enabled organizations is an emergent, socially constructed outcome of the dynamic interplay between algorithmic authority and human judgment. Legitimacy is multidimensional (procedural, distributive, cognitive) and contingent on boundary conditions—transparency/explainability, perceived fairness, and accountability. Algorithmic authority can both increase and erode legitimacy; human judgment is essential to interpret, contextualize, and anchor accountability. The paper presents 10 propositions specifying when hybrid human–AI systems will enhance versus undermine legitimacy.

Key Points

• Conceptualization
  • Decision legitimacy is distinct from organizational legitimacy and is framed along three dimensions:
    • Procedural legitimacy: fairness/transparency of processes (linked to explainability).
    • Distributive legitimacy: fairness of outcomes (linked to bias/equity).
    • Cognitive legitimacy: understandability and taken-for-granted acceptance.
  • Algorithmic authority: legitimacy attributed to algorithmic outputs due to perceived objectivity, consistency, and data-driven rationality.
  • Human judgment: contextual sensitivity, ethical reasoning, and accountability attribution that moderates algorithmic effects.
• Paradox and contingency
  • Algorithmic systems generate an "algorithmic legitimacy paradox": the same features (consistency, objectivity) that increase legitimacy can undermine it via opacity, bias, and responsibility gaps.
  • The legitimacy effect of algorithmic authority is conditional on transparency/explainability, fairness, and accountability structures.
• Multilevel perspective
  • Legitimacy is produced and contested across individual, group/organizational, and societal/institutional levels.
• Propositions (selected highlights)
  • P1–P4: Algorithmic authority increases procedural and distributive legitimacy when processes/outcomes are transparent and unbiased; it decreases legitimacy when opaque or biased.
  • P5–P6: Human judgment enhances cognitive legitimacy and perceived accountability.
  • P7–P8: Complementarity between algorithms and human oversight strengthens legitimacy; absence of oversight weakens it.
  • P9–P10: Relationships vary across levels; sustained legitimacy requires joint satisfaction of transparency, fairness, and accountability.
• Governance implications
  • Responsible AI should be treated as a dynamic organizational capability. Practices include explainability mechanisms, fairness audits, structured human–AI complementarities, and multilevel governance to manage trade-offs (efficiency vs fairness; transparency vs control).
• Limitations acknowledged by author
  • Framework is theoretical; empirical testing and operationalization of constructs are needed.

Data & Methods

• Approach: Conceptual/theoretical development via literature synthesis.
  • Draws on institutional theory, socio-technical systems thinking, behavioral decision-making, and information-systems research.
  • Integrates micro (cognitive), meso (organizational governance), and macro (institutional norms) literatures.
• Outputs:
  • Multidimensional conceptual model and a set of 10 formal propositions (P1–P10).
  • A figure illustrating multilevel interactions (conceptual).
• No original empirical data or quantitative analysis presented.
• Recommended next-step methodologies (implied): experiments, surveys, field studies, case studies, audits, multilevel empirical designs.

Implications for AI Economics

• Adoption & investment incentives
  • Firms’ adoption of AI will depend not only on productivity gains but on legitimacy-related costs/benefits (e.g., investments in explainability, audits, human oversight). Economic models should endogenize firms’ investment in transparency and accountability as components of productive capacity and market acceptance.
• Market competition & differentiation
  • Legitimacy (procedural/distributive/cognitive) can be a competitive differentiator. Firms that credibly signal fairness and accountability may capture higher demand or lower policy risk. This creates incentives for signaling (certifications, independent audits) and potential market segmentation.
• Labor markets & task allocation
  • Human judgment remains necessary for legitimacy; economic models of task reallocation should account for complementarities between AI and human oversight, leading to sustained demand for certain managerial, interpretive, and supervisory roles despite automation.
• Regulation & compliance costs
  • Regulatory requirements (transparency, fairness audits, human-in-the-loop mandates) create compliance costs and alter industry equilibrium. Welfare analysis should weigh efficiency gains from automation against social welfare losses from biased or opaque systems and the costs of corrective governance.
• Information asymmetries & principal–agent problems
  • Opacity of algorithmic systems heightens information asymmetries between firms, regulators, and stakeholders. Economists can model how firms’ private incentives (profit from opaque, high-performance models) conflict with social welfare, motivating interventions (mandated disclosure, liability rules).
• Measurement and empirical strategy for economic research
  • Need for operationalizable metrics for procedural, distributive, and cognitive legitimacy (e.g., transparency scores, bias/outcome disparity measures, survey-based acceptance metrics).
  • Empirical approaches: randomized field experiments on explainability/human oversight; difference-in-differences exploiting regulatory changes; structural models linking investments in explainability to market outcomes; agent-based simulations to capture multilevel dynamics.
• Welfare and distributional consequences
  • Distributive legitimacy concerns map directly to distributional outcomes—algorithmic allocation mechanisms can generate or entrench inequality. Economic evaluations should consider distributional weights and long-term institutional effects.
• Policy design & mechanism design opportunities
  • Policymakers can use mechanism-design tools to build incentive-compatible reporting, auditing, and liability structures that align private algorithmic innovation with social legitimacy criteria.
• Research agenda for AI economists (practical suggestions)
  • Quantify the trade-off frontier between efficiency and legitimacy-related costs.
  • Model firms’ optimal choice of human oversight intensity and explainability investment.
  • Estimate consumer and employee demand elasticities for legitimacy-enhancing features (transparency, fairness guarantees).
  • Evaluate market-level outcomes under different regulatory regimes (disclosure mandates, liability rules, certification markets).
  • Study long-run equilibrium impacts on innovation when legitimacy externalities (reputational spillovers) are internalized.

Overall, the paper offers a structured conceptual foundation that economists can translate into formal models and empirical tests to evaluate how legitimacy considerations reshape incentives, adoption patterns, labor allocation, regulation, and welfare in AI-augmented markets and organizations.