Who Legitimizes the AI Algorithm? Leadership, Volatility, and the Governance of Algorithmic Authority
Listen to a review of this article:
Abstract: Artificial intelligence systems increasingly function as decision-making infrastructures that allocate access, classify individuals, and distribute life chances at organizational scale. While ethical AI governance frameworks proliferate, they overwhelmingly assume stable conditions: reliable infrastructure, coherent regulatory institutions, and baseline organizational legitimacy. This article reconceptualizes ethical AI governance as a legitimacy production challenge rather than a technical compliance problem, arguing that under conditions of volatility—infrastructural fragility, institutional flux, and contested social consent—principles and documentation alone cannot sustain governable algorithmic authority. Drawing on legitimacy theory, leadership scholarship, and algorithmic accountability research, the article develops a three-dimensional volatility typology and proposes the Sensing–Stabilizing–Legitimizing capability framework. This leadership-centered model specifies how organizations build contestability, accountability, and procedural justice into AI systems when background stability conditions fail. The contribution is integrative-conceptual: theorizing volatility as an explicit governance variable and positioning ethical AI governance as strategic leadership capability rather than delegated technical function.
The organizational power of artificial intelligence has fundamentally shifted from support function to decision infrastructure. Algorithmic systems now determine creditworthiness, hiring eligibility, healthcare access, social benefit allocation, and content visibility at scales and speeds that exceed human oversight capacity. This represents more than technological advancement; it constitutes a reorganization of organizational authority itself. Where human judgment once mediated consequences, computational procedures now classify, rank, exclude, and recommend largely beyond direct scrutiny or immediate contestation.
The governance imperative is urgent because algorithmic decisions create new regimes of visibility and invisibility, determining who is seen as legitimate, trustworthy, deserving, or risky. These classifications carry material consequences: access to credit shapes economic mobility; hiring algorithms influence career trajectories; welfare targeting systems determine basic survival resources; content moderation decisions affect political discourse and social connection. When these systems fail, they fail systematically, concentrating harm among populations already experiencing marginalization while operating with an appearance of objectivity that masks embedded biases and structural inequities.
Current governance approaches, despite rapid maturation, inadequately address a critical reality: the conditions under which AI systems operate are frequently unstable. Responsible AI frameworks articulate normative goals—fairness, transparency, accountability—but typically assume consistent infrastructure, predictable regulatory environments, and baseline organizational legitimacy. AI assurance mechanisms provide verification tools—model cards, algorithmic audits, risk frameworks—but presuppose traceable data provenance, reliable monitoring systems, and enforceable standards. Regulatory initiatives demonstrate policy innovation yet struggle with implementation uncertainty and jurisdictional fragmentation. Where stability holds, technical compliance may suffice. Where volatility prevails, governance frameworks lose operational traction.
The Volatility-Governance Nexus
Table 1: Ethical AI Governance and Algorithmic Authority Framework
Governance Dimension | Volatility Type | Key Leadership Capability | Organizational Risk | Stakeholder Impact | Evidence-Based Response | Implementation Strategy |
Infrastructural Volatility | Instability in technical foundations (connectivity, power, data infrastructure) | Sensing | Operational brittleness, monitoring gaps, and degradation of data quality | Exclusion without explanation and systematic disadvantage for underrepresented populations | Establish anticipatory mechanisms such as drift detection and community feedback channels | Build redundancy into monitoring systems and institutionalize harm visibility through multiple feedback channels |
Institutional Volatility | Regulatory uncertainty, fragmented oversight, and shifting enforcement priorities | Stabilizing | Regulatory intervention, compliance costs, and performative rather than operational governance | Weakened recourse pathways and inability to challenge algorithmic decisions | Create minimum viable safeguards including clear decision rights and escalation protocols | Develop adaptive governance documentation and establish distributed leadership with authority to pause deployments |
Socio-political Volatility | Fluctuating trust, contested legitimacy, and politicized interpretations of fairness | Legitimizing | Legitimacy collapse, innovation constraints, and strategic crises | Cumulative erosion of trust and perception of systems as instruments of surveillance or discrimination | Design contestability and procedural justice into systems through accessible recourse and transparent communication | Implement multi-channel recourse pathways, plain-language explanations, and independent review mechanisms |
Volatility manifests across three critical dimensions that directly impact governance capacity. Infrastructural volatility involves instability in the technical foundations AI systems depend upon: connectivity, power reliability, data infrastructure, identity verification systems. When these elements are inconsistent, monitoring gaps emerge, data quality degrades, and exclusions compound precisely among populations least represented in training datasets. Institutional volatility encompasses regulatory uncertainty, fragmented oversight, and shifting enforcement priorities. Organizations navigate incomplete rules, inconsistent signals, and unpredictable compliance thresholds, creating conditions where governance becomes performative rather than operational. Socio-political volatility reflects fluctuating trust, contested legitimacy, and politicized interpretations of fairness. Even technically sound systems provoke backlash when perceived as instruments of surveillance, extraction, or discrimination.
These dimensions interact and amplify one another. Infrastructural fragility concentrates exclusion harms that erode trust. Institutional ambiguity weakens recourse pathways that communities depend upon to challenge algorithmic decisions. Socio-political contestation accelerates regulatory change that creates additional compliance uncertainty. Together, these dynamics transform ethical AI governance from a technical assurance problem into a leadership challenge of legitimacy production: the continuous organizational work required to sustain social consent for algorithmic authority when conditions are unstable.
Reframing the Governance Challenge
This article argues that ethical AI governance under volatility is fundamentally a matter of building legitimacy infrastructure—the organizational capabilities that make algorithmic decisions contestable, accountable, and procedurally just when infrastructure wobbles, institutions shift, and social consent remains contested. This reframing distinguishes governance as leadership work from governance as compliance documentation. Where responsible AI frameworks specify normative goals and assurance mechanisms verify technical performance, legitimacy infrastructure addresses the organizational design question: What must leaders build to sustain governable algorithmic authority when stability cannot be assumed?
The answer requires integrating three leadership capabilities. Sensing establishes anticipatory mechanisms that surface emerging harms before they escalate into crises: drift detection, exclusion pattern monitoring, community feedback channels. Stabilizing builds minimum viable safeguards that remain operational despite infrastructure constraints and institutional uncertainty: clear decision rights, escalation protocols, adaptive documentation practices. Legitimizing designs contestability and procedural justice into algorithmic systems: accessible recourse pathways, genuine explanations, visible accountability mechanisms.
The contribution is theoretically integrative rather than empirically generalizable. By specifying how different volatility dimensions produce distinct governance failure modes and how leadership capabilities respond to those failures, the article offers analytical precision for comparative research and practical guidance for organizational implementation. The following sections develop this framework systematically, demonstrate its application through illustrative scenarios, and derive a research agenda for empirical validation across contexts.
The Algorithmic Authority Landscape
Defining Algorithmic Authority in Organizational Contexts
The phrase "ethical AI" frames governance as a moral overlay on technical systems, implying that ethics constrains technology from outside. This framing obscures the governance challenge. What requires governance is not artificial intelligence as computational technique but algorithmic authority: the organizational capacity of automated systems to classify individuals, assess risk, allocate resources, and determine access in ways that shape life chances and organizational outcomes.
Algorithmic authority is institutional rather than purely technical. It embeds in organizational routines, vendor relationships, performance metrics, regulatory expectations, and public interpretations of fairness. Credit scoring systems exercise authority not simply through statistical predictions but through their position in financial service ecosystems where algorithmic outputs determine approval or denial with limited explanation. Hiring algorithms exercise authority not merely by ranking candidates but by shaping who enters organizational consideration at all. Welfare targeting systems exercise authority not only through eligibility calculations but through their role in distributing survival resources where errors cascade into household crises.
This authority reorganizes organizational power in consequential ways. Decisions that previously required human judgment—with attendant accountability, documentation, and contestation—now occur through computational procedures difficult to audit and harder to challenge. The shift is not merely efficiency gains through automation but a fundamental change in how organizational decisions are made, who can scrutinize them, and whether affected parties can meaningfully contest outcomes they perceive as unjust.
Prevalence and Organizational Adoption Patterns
Algorithmic systems now mediate decisions across virtually every organizational domain. Financial services deploy credit scoring, fraud detection, and loan origination algorithms that determine access to capital. Human resources functions use applicant screening, performance assessment, and promotion prediction systems that shape workforce composition. Healthcare organizations implement diagnostic support, treatment recommendation, and resource allocation algorithms affecting patient care. Public agencies adopt welfare targeting, benefit fraud detection, and service eligibility systems impacting vulnerable populations. Platform companies deploy content moderation, recommendation engines, and user scoring systems that structure information access and social connection.
Adoption accelerates because algorithmic systems promise scalability, consistency, and efficiency. Organizations facing resource constraints turn to automation to process greater volumes with fewer personnel. Competitive pressures drive adoption as algorithmic capabilities become table stakes in digital-first markets. Regulatory compliance itself sometimes encourages algorithmic decision-making where documentation and auditability appear easier with computational systems than discretionary human judgment.
Yet adoption often outpaces governance maturity. Organizations deploy systems before establishing clear accountability structures, monitoring mechanisms, or recourse pathways. The innovation imperative—the competitive pull toward algorithmic capabilities—encounters limited organizational capacity to detect harms, respond to failures, or sustain legitimacy when systems malfunction. This gap between deployment speed and governance readiness is where volatility transforms from operational challenge into strategic crisis.
Organizational and Stakeholder Consequences of Ungoverned Algorithmic Authority
Organizational Performance and Strategic Risks
The organizational consequences of governance failures extend far beyond regulatory penalties or reputational damage. When algorithmic systems produce systematic exclusions, prediction errors, or procedural injustices, organizations face compound risks that directly impact strategic objectives.
Regulatory intervention represents the most visible organizational risk. Governance failures trigger investigations, enforcement actions, and compliance requirements that constrain strategic flexibility. The European Union's implementation of the AI Act demonstrates how rapid regulatory responses to algorithmic harms can fundamentally alter operational models. Organizations that treated governance as afterthought discover that catch-up compliance is exponentially more costly than integrated governance from deployment.
Legitimacy collapse creates strategic crises that technical fixes cannot resolve. When communities perceive algorithmic systems as instruments of discrimination or exploitation, opposition mobilizes rapidly. Civil society investigations, media coverage, and public controversy erode organizational reputation in ways that cascade beyond the specific system at issue. Trust reserves depleted through governance failures prove difficult to rebuild, particularly when organizations respond with technical explanations that communities interpret as evasion rather than accountability.
Operational brittleness emerges when organizations over-rely on algorithmic systems without building governance redundancy. When systems fail—data pipelines break, models drift, infrastructure degrades—organizations lacking human oversight capacity or fallback procedures experience service disruptions that amplify rather than isolate failures. The operational risk is particularly acute when algorithmic authority concentrates in vendor-supplied systems where organizations lack visibility into model logic or control over deployment conditions.
Innovation constraints paradoxically result from inadequate governance. Organizations experiencing governance failures face stakeholder resistance to subsequent algorithmic deployments regardless of use case or risk profile. The innovation pull that initially drove adoption reverses into innovation constraint when organizational legitimacy for algorithmic authority erodes. Strategic initiatives requiring algorithmic capabilities stall when trust is insufficient to sustain implementation.
Research evidence supports these risk patterns. Organizations experiencing algorithmic governance failures demonstrate measurably reduced stakeholder trust, increased regulatory scrutiny, and constrained strategic flexibility compared to organizations with mature governance infrastructures. The strategic implication is clear: ethical governance is not peripheral to algorithmic strategy but constitutive of it. Organizations cannot execute algorithmic strategies they lack legitimacy to implement.
Stakeholder and Community Impacts
The individual and community consequences of governance failures are both immediate and cumulative. Algorithmic decisions that lack transparency, contestability, or procedural justice harm stakeholders in ways that extend beyond individual instances to shape systemic inequities.
Exclusion without explanation represents the most fundamental harm. Individuals denied credit, employment, benefits, or access receive algorithmic rejections without understanding why decisions occurred or how to improve prospects. The opacity compounds the material harm: not only does someone not receive a loan, they cannot learn what would make them creditworthy. Not only does an applicant face hiring rejection, they cannot discern whether bias influenced assessment. The procedural injustice—being judged by systems one cannot comprehend or challenge—creates psychological harm alongside material disadvantage.
Compounding marginalization occurs when algorithmic systems systematically disadvantage populations already experiencing structural inequity. Predictive policing concentrates enforcement in communities already over-policed. Credit algorithms deny access to populations historically excluded from financial systems. Hiring screens filter out candidates from demographic groups under-represented in training data. The algorithmic reproduction of historical patterns legitimizes continued inequality under the guise of objective assessment.
Recourse futility emerges when contestation pathways exist nominally but prove ineffective practically. Individuals can appeal algorithmic decisions but lack information necessary to construct meaningful challenges. Organizations provide explanation but frame it in technical language inaccessible to non-specialists. Review processes exist but rarely overturn algorithmic outputs, signaling that contestation is performative rather than genuine.
Cumulative erosion of trust extends beyond individual harms to shape community relationships with institutions. When algorithmic systems repeatedly produce outcomes perceived as unjust, communities lose confidence in organizational fairness. The erosion affects not only the specific systems at issue but institutional legitimacy broadly. Public agencies lose trust essential for service delivery effectiveness. Private organizations face community resistance that constrains strategic initiatives regardless of potential benefit.
These stakeholder consequences are not peripheral externalities but central governance failures. Organizations exercise algorithmic authority with implicit social contracts: that decisions will be explainable, challengeable, and substantively fair. When volatility undermines those commitments, algorithmic authority itself becomes ungovernable.
Evidence-Based Organizational Responses to Governance Challenges
Transparent Communication and Stakeholder Engagement
Research on organizational legitimacy demonstrates that procedural transparency significantly impacts stakeholder perceptions of fairness even when substantive outcomes remain unchanged. Stakeholders assess legitimacy not only through results but through the processes producing those results and their accessibility to scrutiny. For algorithmic systems, this means transparency must extend beyond technical disclosures to encompass decision-making authority, accountability structures, and genuine engagement mechanisms.
Effective transparency approaches include:
Plain-language explanations that describe algorithmic decision logic without requiring technical expertise, enabling affected individuals to understand why particular outcomes occurred
Proactive disclosure of algorithmic system deployment, specifying use cases, decision thresholds, and override conditions before controversies emerge
Stakeholder advisory mechanisms that involve community representatives, civil society organizations, and affected populations in governance design rather than treating engagement as post-deployment consultation
Regular algorithmic impact reporting disaggregated by relevant demographic dimensions, making exclusion patterns visible to both internal governance bodies and external stakeholders
Accessible documentation of governance structures, including decision rights, escalation pathways, and accountability assignments, so stakeholders can identify who holds responsibility when failures occur
Microsoft established an external AI ethics advisory board with authority to audit internal deployments and publish findings, creating institutional commitment to transparency that extends beyond corporate communications. The governance model demonstrates that transparency becomes credible when external parties can verify organizational claims independently.
Organizations implementing transparency as governance infrastructure rather than public relations discover that disclosure disciplines internal decision-making. Knowing that algorithmic deployments will face external scrutiny creates organizational incentives for governance rigor during design rather than remediation after failures. The strategic benefit is not merely reputational but operational: transparency surfaces governance gaps early when corrections are manageable rather than late when crises are imminent.
Procedural Justice and Accessible Contestability
Procedural justice research establishes that people accept unfavorable decisions more readily when they perceive decision-making processes as fair: providing voice, demonstrating neutrality, showing respect, and building trust. For algorithmic decisions, procedural justice requires more than technical accuracy; it demands that affected individuals can understand decisions, challenge outcomes, and receive genuine consideration of appeals.
Key procedural justice mechanisms include:
Multi-channel recourse pathways that accommodate different access capabilities, including offline options for populations with limited digital connectivity
Explanation formats calibrated to user context rather than uniform technical disclosures, recognizing that meaningful explanation varies by stakeholder knowledge and information needs
Independent review mechanisms where appeals receive consideration from parties not incentivized to defend original algorithmic outputs
Transparent outcomes tracking for contestation processes, demonstrating whether challenges produce substantive reconsideration or merely formalities
Community liaison functions that help individuals navigate challenge procedures and translate between technical system operations and lived experiences of harm
Omidyar Network's Responsible Data program funded civil society organizations to provide algorithmic advocacy services for communities facing automated decision systems. The intermediary model recognizes that individual contestation proves insufficient when power asymmetries favor organizations with technical expertise and legal resources over individuals experiencing harm.
Organizations building procedural justice into algorithmic governance discover that contestability improves system performance beyond ethical considerations. When challenge mechanisms surface systematic errors—model drift, proxy discrimination, data quality degradation—organizations gain operational intelligence that internal monitoring might miss. Communities experiencing harm often detect failures before organizational metrics register problems. Procedural justice mechanisms thus function as distributed sensing systems that enhance both governance and performance.
Algorithmic Impact Assessment as Living Practice
Research on risk management demonstrates that prospective impact assessment, when integrated into deployment decisions rather than treated as compliance documentation, meaningfully reduces harm incidence and improves organizational responsiveness to emerging risks. For algorithmic systems, impact assessment must function as living governance practice rather than one-time exercise, continuously updating as contexts change and new harms surface.
Effective impact assessment practices include:
Pre-deployment assessments that explicitly evaluate who might be harmed when systems function as designed, not merely when they malfunction technically
Differential impact analysis across demographic dimensions, infrastructure access levels, and digital literacy capacities, surfacing exclusion risks concentrated among vulnerable populations
Monitoring infrastructure that tracks assessment assumptions over time, detecting when operational conditions diverge from design expectations
Triggered reassessment protocols specifying conditions requiring governance review: complaint volume thresholds, drift detection, regulatory changes, infrastructure disruptions
Stakeholder validation of impact findings through community engagement, testing whether organizational assessments align with lived experiences of system interactions
UK Information Commissioner's Office guidance on data protection impact assessments explicitly requires organizations to revisit assessments when processing conditions change materially, establishing impact evaluation as continuous governance work rather than static compliance artifact.
Organizations implementing dynamic impact assessment discover governance advantages beyond harm reduction. When assessment updates trigger deployment pauses or system modifications, organizations build organizational muscle for responsible innovation speed: the capacity to move quickly on deployments demonstrating acceptable risk while slowing or halting deployments where impacts remain uncertain or harms emerge. This calibrated approach contrasts with both reckless deployment and blanket risk aversion, enabling innovation while maintaining governance accountability.
Distributed Leadership and Clear Decision Rights
Leadership scholarship on organizational agility demonstrates that clear decision rights and distributed authority enable faster response to emerging challenges while maintaining accountability. For algorithmic governance, this means establishing who can pause deployments, who owns risk acceptance decisions, and who holds authority to modify systems when harms surface.
Critical leadership structure elements include:
Designated risk owners at appropriate organizational levels with clear authority to accept, reject, or modify algorithmic deployment proposals
Escalation protocols specifying triggers that move governance decisions to senior leadership rather than leaving them embedded in technical teams
Cross-functional governance bodies that integrate data science, legal, operations, and community engagement perspectives rather than siloing governance in compliance functions
Override authorities empowering frontline staff to deviate from algorithmic recommendations when contextual knowledge suggests outputs are inappropriate
Accountability visibility through governance dashboards that make decision rights, review outcomes, and system modifications transparent to both internal stakeholders and external oversight bodies
Salesforce established an Office of Ethical and Humane Use of Technology reporting directly to executive leadership with budget authority and deployment veto power. The organizational positioning signals that algorithmic governance holds strategic priority rather than existing as technical advisory function.
Organizations that clarify decision rights discover that accountability improves not only governance but innovation effectiveness. When governance responsibilities are ambiguous, risk-averse teams delay deployments unnecessarily while risk-tolerant teams proceed despite governance concerns. Clear authorities enable both: rapid deployment where risks are acceptable and deliberate caution where uncertainties remain unresolved. The leadership benefit is organizational learning: as governance decisions accumulate, organizations build institutional knowledge about risk calibration under different volatility conditions.
Capability Building and Organizational Literacy
Research on organizational learning demonstrates that technical expertise alone proves insufficient for governance effectiveness. Non-technical leaders require sufficient understanding of algorithmic systems to ask appropriate oversight questions, assess governance trade-offs, and hold technical teams accountable. Frontline staff need capability to recognize when algorithmic outputs diverge from contextual appropriateness, justifying overrides or escalations.
Key capability-building approaches include:
Leadership education programs that build senior executive understanding of algorithmic risks, governance mechanisms, and strategic implications without requiring technical implementation expertise
Frontline training that develops staff capacity to recognize algorithmic failure modes, exercise override authority appropriately, and document cases where system outputs prove inadequate
Cross-functional rotation exposing data scientists to operational contexts and operations staff to data science perspectives, building shared understanding of governance challenges
Community education initiatives that help stakeholders understand algorithmic system functions, limitations, and contestation options, reducing information asymmetries that compound power imbalances
Governance capability assessment as standard organizational practice, identifying knowledge gaps and resourcing learning programs rather than assuming that technical skills alone ensure governance competence
Partnership on AI developed stakeholder education resources specifically designed for non-technical audiences including community advocates, journalists, and policymakers. The initiative recognizes that governance accountability requires distributed capability rather than expertise concentration among technical specialists.
Organizations investing in governance literacy discover that capability building generates benefits beyond immediate oversight improvements. When non-technical leaders understand algorithmic constraints and possibilities, strategic conversations improve: executives make better deployment decisions, resource allocation becomes more appropriate, and governance integrates into business planning rather than emerging as afterthought. The organizational maturity signal is when governance questions arise naturally during strategy discussions rather than requiring specialized prompting.
Building Long-Term Governance Capacity Under Volatility
Institutionalizing Anticipatory Sensemaking
Volatility makes reactive governance insufficient because by the time failures become visible through conventional metrics, harms have often become systemic rather than isolated. Organizations must develop capabilities that surface emerging governance challenges before they escalate into crises—what the framework terms sensing: anticipatory sensemaking that institutionalizes harm visibility.
Under infrastructural volatility, sensing requires building redundancy into monitoring systems. Organizations cannot assume clean data pipelines or consistent system uptime will reveal drift, exclusion patterns, or silent failures. Effective sensing mechanisms include multiple feedback channels: quantitative monitoring supplemented by qualitative community reporting; automated drift detection complemented by human pattern recognition; internal audit functions reinforced by external independent review. The principle is that no single monitoring approach proves sufficient when infrastructure is unstable; governance requires diverse information sources that remain functional even when primary systems degrade.
Under institutional volatility, sensing extends to regulatory signals and policy development. Organizations must build dedicated capacity that tracks legislative discussions, regulatory consultations, enforcement patterns, and policy debates as they emerge rather than reacting to finalized rules. This prospective approach enables proactive governance adaptation rather than reactive scrambling when requirements change. Organizations with mature sensing capabilities participate in policy development processes, contributing technical expertise and operational insights that improve regulatory design while positioning themselves strategically for implementation.
Under socio-political volatility, sensing demands attention to relational and reputational signals that conventional dashboards systematically miss. Complaint volumes, civil society scrutiny, community trust indicators, and media coverage patterns provide early warnings of legitimacy erosion before it manifests in crisis. Organizations must establish structured mechanisms for capturing these signals: regular stakeholder engagement sessions, civil society liaison functions, community advisory bodies, and systematic media monitoring. The governance integration challenge is elevating these qualitative signals to decision-relevant status rather than treating them as peripheral public relations concerns.
The organizational design implication is that sensing cannot remain ad hoc. It requires institutional commitment: dedicated roles, resource allocation, escalation protocols connecting sensing functions to decision-making authority, and executive accountability for response when signals indicate emerging governance challenges.
Adaptive Governance Documentation
Institutional volatility creates a paradox for governance documentation. Compliance requires evidence that appropriate safeguards exist, yet documentation becomes obsolete rapidly when rules shift, enforcement priorities change, or operational conditions diverge from design assumptions. Organizations navigating this paradox must treat documentation not as static compliance artifacts but as living instruments that evolve alongside changing conditions.
Adaptive documentation practices center on what the framework terms stabilizing: establishing minimum viable safeguards that remain operational despite instability. Rather than comprehensive documentation that proves unmanageable to maintain, organizations prioritize essential elements that enable accountability regardless of volatility: clear decision rights, documented risk acceptance, escalation protocols, and recourse mechanisms. These governance foundations function across different regulatory regimes and operational contexts because they address fundamentals: who decided, what alternatives were considered, who bears responsibility when failures occur.
Effective adaptive practices include regular review cycles triggered not merely by calendar intervals but by substantive changes in operating conditions. When infrastructure configurations shift, when regulations update, when complaint patterns change, documentation review becomes automatic governance practice. The review asks whether existing safeguards remain appropriate given new conditions or whether modifications are necessary to sustain governance effectiveness.
Critically, adaptive documentation requires organizational cultures that view governance updates as responsible stewardship rather than compliance burden. When updating documentation carries stigma—suggesting initial work was inadequate—organizations face incentives to maintain obsolete safeguards rather than acknowledge changing realities. Mature governance cultures recognize that volatility demands continuous adaptation and that documentation evolution signals governance health rather than governance failure.
The practical benefit extends beyond compliance. When documentation remains current, it functions as operational resource rather than archaeological artifact. Staff confronting novel situations can reference governance guidance confident it reflects present conditions. Auditors and regulators encounter documentation that describes actual practices rather than idealized intentions. The governance credibility that results from alignment between documented commitments and operational reality creates organizational capacity to sustain legitimacy even when failures occur.
Embedding Procedural Justice in Operational Practice
Socio-political volatility means technical accuracy alone cannot sustain algorithmic authority. Communities with fluctuating trust, historical experiences of institutional discrimination, and contested interpretations of fairness assess legitimacy through procedural dimensions that exceed substantive outcomes. Organizations must therefore invest in what the framework terms legitimizing: institutional work that designs contestability and procedural justice into algorithmic systems as operational requirements rather than aspirational additions.
This requires fundamentally reorienting the governance question from "Is the algorithm accurate?" to "Is the algorithmic authority exercised in ways that sustain social consent?" Accuracy matters, but insufficient accuracy alone explains governance failures. Systems can be technically sound yet organizationally ungovernable when affected parties cannot understand decisions, challenge outcomes, or access meaningful recourse.
Embedding procedural justice operationally means making contestability available throughout system lifecycles. At deployment, this involves establishing clear recourse pathways, publishing explanation formats, and resourcing review mechanisms before first decisions occur. During operations, it requires monitoring contestation patterns for systematic issues, tracking whether challenges receive substantive consideration rather than perfunctory responses, and closing feedback loops so individuals see tangible evidence that their challenges influenced outcomes. When failures emerge, procedural justice demands transparent accountability: acknowledging harms, explaining causes, specifying remediation, and demonstrating organizational learning that reduces future recurrence.
The organizational challenge is resourcing procedural justice adequately. Contestability infrastructure requires people, processes, and systems distinct from the algorithmic deployments themselves. Organizations face temptations to under-resource these functions as non-core expenses. Yet research demonstrates that procedural justice investment directly impacts legitimacy sustainability. Organizations cutting corners on recourse accessibility, explanation quality, or review independence discover that legitimacy collapses prove exponentially more costly to address than procedural justice would have been to maintain.
The strategic insight is that procedural justice functions as legitimacy insurance. Under stable conditions, robust contestability mechanisms may see limited use. Under volatile conditions—when legitimacy is contested and trust is fragile—procedural justice becomes the organizational capability determining whether algorithmic authority remains governable or triggers backlash that constrains strategic flexibility. Organizations that treated procedural justice as optional discover during crises that rebuilding legitimacy from zero proves far more difficult than sustaining it through continuous investment.
Conclusion
This article reconceptualizes ethical AI governance as legitimacy infrastructure: a leadership capability system enabling organizations to exercise algorithmic authority in ways that remain contestable, accountable, and socially tolerable when conditions are unstable. The central theoretical contribution integrates legitimacy theory, leadership scholarship, and algorithmic accountability research to demonstrate that volatility is not ambient turbulence but a causal governance condition producing distinct failure modes that principles and technical controls alone cannot address.
The Sensing–Stabilizing–Legitimizing framework specifies the leadership work required to sustain governable algorithmic authority under three volatility dimensions. Infrastructural instability demands anticipatory mechanisms that surface harms early and build redundancy into monitoring. Institutional uncertainty requires adaptive safeguards that remain operational despite regulatory flux and enforcement ambiguity. Socio-political contestation necessitates procedural justice mechanisms that make algorithmic decisions explainable, challengeable, and visibly accountable even when trust is fragile.
For organizational leaders, the practical implication is unambiguous: ethical AI governance cannot be delegated to technical teams, compliance functions, or vendor partners as peripheral responsibility. When algorithmic systems mediate consequential decisions affecting stakeholder life chances, governance becomes strategic leadership capability determining whether innovation remains sustainable or collapses into legitimacy crisis. Leaders who treat governance as afterthought discover that algorithmic failures return as strategic constraints precisely when competitive pressures make recovery most difficult.
The research agenda challenges scholars to test the framework's propositions empirically: whether redundant sensing reduces undetected exclusions under infrastructural volatility; whether adaptive documentation sustains legitimacy longer under institutional volatility; whether procedural justice investment outperforms technical accuracy investment under socio-political volatility. Comparative research across sectors, geographies, and volatility configurations will refine understanding of when particular governance approaches prove effective and where boundary conditions limit their applicability.
As algorithmic authority continues expanding in organizational scope, scale, and consequence, the capacity to build and maintain legitimacy infrastructure becomes not merely a governance competency but the defining leadership challenge of the algorithmic era. Organizations that master this capability can innovate responsibly, deploying AI systems that deliver value while sustaining the social consent required for continued operation. Organizations that fail face escalating cycles of deployment, failure, crisis, and strategic constraint that ultimately render algorithmic innovation self-defeating.
The deeper insight transcends AI governance specifically: legitimacy is not a byproduct of good governance but its precondition. Organizations cannot govern what they have not first legitimized. Under volatility, this means building acceptance infrastructure alongside technological capability, treating contestability as operational requirement rather than public relations gesture, and recognizing that leadership work sustaining legitimacy constitutes strategic investment rather than compliance cost. The algorithmic future belongs not to organizations with the most sophisticated technical capabilities but to those with the governance maturity to exercise algorithmic authority responsibly even when conditions remain fundamentally uncertain.
Research Infographic
References
Bamberger, P. A., & Pratt, M. G. (2010). Moving forward by looking back: Reclaiming unconventional research contexts and samples in organizational scholarship. Academy of Management Journal, 53(4), 665–671.
Bennett, N., & Lemoine, G. J. (2014). What a difference a word makes: Understanding threats to performance in a VUCA world. Business Horizons, 57(3), 311–317.
Bitektine, A. (2011). Toward a theory of social judgments of organizations: The case of legitimacy, reputation, and status. Academy of Management Review, 36(1), 151–179.
Cornelissen, J. P. (2017). Editor's comments: Developing propositions, a process model, or a typology? Addressing the challenges of writing theory without a boilerplate. Academy of Management Review, 42(1), 1–9.
Crawford, K. (2021). Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.
Deephouse, D. L., Bundy, J., Tost, L. P., & Suchman, M. C. (2017). Organizational legitimacy: Six key questions. In R. Greenwood, C. Oliver, T. B. Lawrence, & R. E. Meyer (Eds.), The SAGE handbook of organizational institutionalism (2nd ed., pp. 27–54). SAGE Publications.
Edelman, L. B. (2016). Working law: Courts, corporations, and symbolic civil rights. University of Chicago Press.
Eisenhardt, K. M. (1989). Building theories from case study research. The Academy of Management Review, 14(4), 532–550.
Eubanks, V. (2018). Automating inequality: How high-tech tools profile, police, and punish the poor. St. Martin's Press.
Floridi, L., Cowls, J., Beltrametti, P., & Vayena, E. (2018). AI4People—an ethical framework for a good AI society: Opportunities, risks, principles, and recommendations. Minds and Machines, 28(4), 689–707.
Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., Daumé, H., & Crawford, K. (2021). Datasheets for datasets. Communications of the ACM, 64(12), 86–92.
George, G., Corbishley, C., Khayesi, J. N. O., Haas, M. R., & Tihanyi, L. (2016). Bringing Africa in: Promising directions for management research. Academy of Management Journal, 59(2), 377–393.
Graham, S. (2011). Disrupted cities: When infrastructure fails. In S. Graham (Ed.), Disrupted cities: When infrastructure fails (pp. 1–2). Routledge.
Graham, S., & Thrift, N. (2007). Out of order: Understanding repair and maintenance. Theory, Culture & Society, 24(3), 1–25.
Hällgren, M., Rouleau, L., & de Rond, M. (2018). A matter of life or death: How extreme context research matters for management and organization studies. Academy of Management Annals, 12(1), 111–153.
Heifetz, R., Grashow, A., & Linsky, M. (2009). The practice of adaptive leadership: Tools and tactics for changing your organization and the world. Harvard Business Press.
Jobin, A., Ienca, M., & Vayena, E. (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1(9), 389–399.
Kroll, J. A., Huey, J., Barocas, S., Felten, E. W., Reidenberg, J. R., Robinson, D. G., & Yu, H. (2017). Accountable algorithms. University of Pennsylvania Law Review, 165(3), 633–705.
Langley, A. (1999). Strategies for theorizing from process data. The Academy of Management Review, 24(4), 691–710.
Lawrence, T. B., Suddaby, R., & Leca, B. (Eds.). (2011). Institutional work: Actors and agency in institutional studies of organizations. Cambridge University Press.
MacLean, T. L., & Behnam, M. (2010). The dangers of decoupling: The relationship between compliance programs and unethical behavior. Academy of Management Journal, 53(6), 1499–1520.
Maitlis, S., & Christianson, M. (2014). Sensemaking in organizations: Taking stock and moving forward. Academy of Management Annals, 8(1), 57–125.
Martin, K., & Waldman, A. (2023). Are algorithmic decisions legitimate? The effect of process and outcomes on perceptions of legitimacy of AI decisions. Journal of Business Ethics, 183(3), 653–670.
Mitchell, M., Wu, S., Zaldivar, A., Barnes, P., Vasserman, L., Hutchinson, B., … Gebru, T. (2019). Model cards for model reporting. Proceedings of the Conference on Fairness, Accountability, and Transparency, 220–229.
Mittelstadt, B. (2019). Principles alone cannot guarantee ethical AI. Nature Machine Intelligence, 1(11), 501–507.
National Institute of Standards and Technology. (2023). Artificial intelligence risk management framework (AI RMF 1.0). U.S. Department of Commerce.
Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. New York University Press.
Pasquale, F. (2015). The black box society: The secret algorithms that control money and information. Harvard University Press.
Raji, I. D., Smart, A., White, R. N., Mitchell, M., Gebru, T., Hutchinson, B., & Barnes, P. (2020). Closing the AI accountability gap: Defining an end-to-end framework for internal algorithmic auditing. Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, 33–44.
Schoemaker, P. J. H., Heaton, S., & Teece, D. (2018). Innovation, dynamic capabilities, and leadership. California Management Review, 61(1), 15–42.
Scott, W. R. (2014). Institutions and organizations: Ideas, interests, and identities (4th ed.). SAGE Publications.
Selbst, A. D., Boyd, D., Friedler, S. A., Venkatasubramanian, S., & Vertesi, J. (2019). Fairness and abstraction in sociotechnical systems. Proceedings of the Conference on Fairness, Accountability, and Transparency, 59–68.
Siggelkow, N. (2007). Persuasion with case studies. Academy of Management Journal, 50(1), 20–24.
Suchman, M. C. (1995). Managing legitimacy: Strategic and institutional approaches. The Academy of Management Review, 20(3), 571–610.
Taeihagh, A. (2021). Governance of artificial intelligence. Policy and Society, 40(2), 137–157.
Teece, D. J., Peteraf, M., & Leih, S. (2016). Dynamic capabilities and organizational agility: Risk, uncertainty, and strategy in the innovation economy. California Management Review, 58(4), 13–35.
Uhl-Bien, M., & Arena, M. (2018). Leadership for organizational adaptability: A theoretical synthesis and integrative framework. The Leadership Quarterly, 29(1), 89–104.
Weick, K. E. (1995). Sensemaking in organizations. SAGE Publications.
Whetten, D. A. (1989). What constitutes a theoretical contribution? The Academy of Management Review, 14(4), 490–495.
Zietsma, C., & Lawrence, T. B. (2010). Institutional work in the transformation of an organizational field: The interplay of boundary work and practice work. Administrative Science Quarterly, 55(2), 189–221.
Jonathan H. Westover, PhD is Chief Research Officer (Nexus Institute for Work and AI); Associate Dean and Director of HR Academic Programs (WGU); Professor, Organizational Leadership (UVU); OD/HR/Leadership Consultant (Human Capital Innovations). Read Jonathan Westover's executive profile here.
Suggested Citation: Westover, J. H. (2026). Who Legitimizes the AI Algorithm? Leadership, Volatility, and the Governance of Algorithmic Authority. Human Capital Leadership Review, 36(3). doi.org/10.70175/hclreview.2020.36.3.4
