When AI Gives Advice: The Asymmetric Power of Algorithmic Moral Influence in Organizations

Jonathan H. Westover, PhD
1 hour ago
17 min read

Listen to a review of this article:

Abstract: As artificial intelligence systems become embedded in organizational decision-making, a critical question emerges: do employees defer to AI recommendations on ethical choices the same way they surrender to AI on cognitive tasks? This article synthesizes recent experimental evidence demonstrating that AI moral influence operates directionally rather than symmetrically. When AI systems recommend prosocial behaviors—such as generosity, cooperation, or honesty—individuals show substantial behavioral shifts. Yet when AI recommends antisocial actions, compliance fails to materialize, even when participants verbally acknowledge the recommendation. This asymmetry inverts patterns observed in human-to-human behavioral contagion, where antisocial influence typically dominates. The findings reveal a domain boundary in AI authority: algorithmic systems can activate existing moral preferences but cannot override them. For organizations deploying AI-assisted decision systems, these results carry significant implications for ethics, governance, and the design of human-AI collaboration frameworks that preserve rather than erode moral agency.

Artificial intelligence has moved from the periphery to the center of organizational life with remarkable speed. What began as isolated algorithmic tools has evolved into agentic systems that advise, recommend, and increasingly shape consequential decisions across domains ranging from hiring and performance evaluation to resource allocation and strategic planning (Bick et al., 2026; Cowgill et al., 2026). As this transformation accelerates, a fundamental question confronts practitioners and scholars alike: when AI systems offer behavioral guidance on morally charged decisions, do humans defer to those recommendations—and if so, under what conditions and with what limits?

Recent evidence suggests the answer is more nuanced than simple acceptance or rejection. Research demonstrates that people exhibit what has been termed "cognitive surrender" to AI on reasoning tasks, adopting AI-generated answers whether correct or incorrect (Shaw & Nave, 2026). Separately, survey evidence indicates that laypeople rate AI-generated ethical guidance as roughly equivalent in quality to advice from human ethics experts (Meincke et al., 2026). Taken together, these findings could suggest that algorithmic moral authority mirrors cognitive authority—that AI recommendations shift behavior symmetrically regardless of whether they push toward prosocial or antisocial outcomes.

A groundbreaking experimental study by Dimant (2026) challenges this symmetry assumption. Across three incentivized paradigms involving over 600 participants, the research reveals that AI moral influence operates directionally: prosocial AI advice substantially increases other-regarding transfers in dictator games (roughly 15 percentage points), while antisocial AI advice produces no detectable downward shift, even as participants verbally acknowledge receiving the recommendation. This pattern inverts the antisocial-dominance regularity documented in human peer-effect studies, where antisocial behavior spreads more readily than prosocial behavior (Dimant, 2019; Bicchieri et al., 2022).

Why This Matters Now

The practical stakes are substantial. Organizations increasingly deploy AI systems that provide recommendations on decisions with clear moral dimensions: whom to hire, how to allocate raises, whether to approve exceptions to policies, how to respond to customer complaints. If AI influence is directionally constrained—amplifying prosocial impulses but unable to override moral guardrails—then the governance calculus shifts fundamentally. Rather than facing bilateral risk (AI pushing behavior in both positive and negative directions with equal force), organizations confront an asymmetric influence landscape where AI can serve as a prosocial amplifier but offers limited capacity to erode ethical standards through direct recommendation.

This article examines the organizational implications of directional AI moral influence. We explore the mechanisms underlying the asymmetry, assess how it manifests across different decision contexts, and develop practical guidance for organizations seeking to harness AI's prosocial potential while preserving moral agency. The core insight—that AI supplies a permission structure for actions aligning with private moral preference but cannot supply the community standing required to override that preference—has direct bearing on how organizations should design, deploy, and govern AI-assisted decision systems.

The AI Authority Landscape

Defining Algorithmic Authority in Organizational Context

Algorithmic authority refers to the degree to which individuals weight or defer to recommendations generated by computational systems when those recommendations inform consequential choices. This authority operates along multiple dimensions: cognitive authority (trust in algorithmic accuracy on factual or analytical tasks), predictive authority (confidence in algorithmic forecasts), and normative authority (willingness to follow algorithmic guidance on what one should do) (Logg et al., 2019; Glikson & Woolley, 2020).

The normative dimension is where moral influence lives. Unlike cognitive tasks where ground truth can be verified (e.g., "Will this customer churn?" or "What is the optimal inventory level?"), moral decisions involve value judgments where multiple defensible positions may coexist. When an AI system recommends that a manager should share credit for team success, approve flexible work arrangements, or report a compliance concern, it is asserting normative authority over a domain traditionally governed by social norms, organizational culture, and individual conscience.

Prevalence and Current State of Practice

AI-assisted decision systems have proliferated across organizational functions. In talent management, algorithmic tools screen résumés, predict performance, and recommend compensation adjustments (Cowgill et al., 2026). In customer service, sentiment analysis systems flag complaints requiring escalation and suggest response strategies. In operations, scheduling algorithms allocate shifts and recommend leave approvals. While these systems are nominally presented as decision support, research on automation bias demonstrates that humans often defer to algorithmic recommendations even when those recommendations conflict with available evidence or personal judgment (Dietvorst et al., 2015).

The ethical stakes intensify when recommendations carry distributional or fairness implications. Consider an AI system that recommends denying a leave request because coverage is thin, or an algorithm that flags an employee as "low engagement risk" and therefore suitable for a stretched workload. These recommendations embed value judgments about whose interests matter and what behaviors are appropriate, yet they often arrive wrapped in the objectivity heuristic that surrounds algorithmic output (Bigman & Gray, 2018).

The Cognitive Surrender Benchmark

Recent experimental work establishes that AI deference on cognitive tasks can be remarkably high. Shaw and Nave (2026) document that participants adopt AI-generated answers to reasoning problems at rates that rival or exceed adoption of human expert advice, and critically, they do so symmetrically—accepting AI guidance both when it steers them toward correct answers and when it steers them away. This symmetric deference, termed "cognitive surrender," suggests that algorithmic authority on factual tasks operates without strong directional bias.

If cognitive surrender translated directly to the moral domain, we would expect AI behavioral recommendations to shift choices symmetrically: prosocial advice moving behavior upward by roughly the same magnitude that antisocial advice moves it downward. The Dimant (2026) findings demonstrate this extrapolation fails. The directional asymmetry—large prosocial shifts paired with null antisocial shifts—pins down a domain boundary in AI authority that cognitive surrender alone would miss.

Organizational and Individual Consequences of AI Moral Influence

Organizational Performance Impacts

The directional nature of AI moral influence carries distinct implications for organizational outcomes. On the upside, AI systems that surface prosocial recommendations—encouragement to share credit, recognize contributions, or support colleagues—can meaningfully increase the frequency of such behaviors without requiring policy mandates or heavy-handed oversight. Dimant's (2026) 15-percentage-point increase in dictator-game transfers, when scaled to organizational contexts like charitable giving campaigns, peer recognition systems, or cross-functional collaboration requests, represents economically significant behavioral change.

Consider a concrete scenario: an AI-augmented performance management system that prompts managers to identify team members whose contributions may have gone unrecognized. If prosocial AI recommendations reliably increase the likelihood that managers take time to write recognition notes (the equivalent of "giving" in the experimental paradigm), the system generates positive cultural effects—higher engagement, stronger peer relationships, improved retention—at relatively low implementation cost (Frey & Meier, 2004; Bicchieri & Dimant, 2022).

Quantified effects in related domains support the magnitude of potential impact. Field experiments on social norm interventions in charitable giving find effect sizes in the range of 10–20% increases in donation rates when individuals receive information about peer behavior (Frey & Meier, 2004). If AI-delivered prosocial recommendations produce shifts in this range across other organizational prosocial behaviors—volunteering for extra projects, mentoring junior colleagues, sharing knowledge proactively—the aggregate impact on organizational social capital is substantial.

Individual Wellbeing and Stakeholder Impacts

From the individual employee perspective, directional AI influence preserves a critical form of agency. The evidence that antisocial AI recommendations fail to move behavior downward suggests that AI systems cannot easily override personal moral convictions. Employees who receive AI recommendations to cut corners, shirk collaborative obligations, or exploit information asymmetries for personal gain are unlikely to comply if those actions conflict with their internalized ethical standards (Dimant, 2026).

This asymmetry provides a form of moral floor protection. In contrast to fears that algorithmic systems might gradually erode ethical behavior through accumulated micro-recommendations (analogous to "dark patterns" in consumer interfaces), the directional constraint means AI can amplify prosocial impulses but cannot systematically push individuals against their moral compass. The mechanism appears to be one of confirmation rather than override: AI supplies normative validation for actions individuals were already disposed to take, but lacks the community standing required to displace existing norms (Bicchieri, 2006; Krupka & Weber, 2013).

That said, the asymmetry is not an unconditional guarantee. The experimental evidence establishes the boundary condition under specific design features: one-shot interactions, anonymous settings, and clear value-laden choices. Organizations that deploy AI systems in repeated interactions, with personalized framing, or in domains where moral standards are ambiguous may encounter different dynamics. The challenge for practice is to preserve the asymmetry's protective properties while harnessing its prosocial potential.

Risks in Deployment Contexts

Three risks merit attention. First, context collapse: what holds in laboratory dictator games may not generalize to complex organizational decisions where moral considerations trade off against performance pressures, political dynamics, or resource constraints. An AI system that recommends a manager "be generous" with flexible work arrangements may receive compliance when that generosity is costless, but encounter resistance when it conflicts with operational demands.

Second, expectation violation: if employees come to expect that AI recommendations are unconditionally prosocial (because antisocial recommendations are rare or incredible), they may discount all AI moral guidance as naive or out-of-touch with organizational realities. The prosocial uplift depends partly on perceived credibility, and credibility erodes if recommendations systematically ignore legitimate constraints.

Third, inequality in influence: the directional effect may not distribute evenly. Employees who already lean prosocial may respond more strongly to AI confirmation, while those with weaker prosocial dispositions—precisely the group where behavioral change is most needed—may show muted responses. If so, AI moral influence could widen rather than narrow gaps in organizational citizenship behavior, concentrating prosocial actions among those already inclined toward them.

Evidence-Based Organizational Responses

Table 1: Case Studies of AI Moral Influence in Organizations

Organization	AI Application Domain	Specific Intervention	Moral Direction	Outcome Metric	Compliance/Follow-through Rate	Governance Mechanism
University of Pennsylvania (Dimant, 2026)	Experimental Economic Game (Dictator Game)	Prosocial AI advice recommending generosity and other-regarding transfers versus antisocial advice recommending self-interest	Mixed (Prosocial and Antisocial)	Impact on transfers and behavior shifts	Approximately 15 percentage point increase for prosocial advice; no detectable shift (null effect) for antisocial advice	Experimental setting (research study)
Patagonia	Employee Engagement / Sustainability	AI system surfacing opportunities for environmental initiatives aligned with organizational mission	Prosocial	Participation rates in voluntary sustainability programs	Exceeds 80%	Value alignment and collective ownership
Deloitte	Performance Management / Coaching	AI-generated coaching recommendations integrated with development credits	Prosocial	Follow-through rates on coaching direct reports	70% (increased from 40% base rate)	Resource buffers, performance incentives, and cost-sharing (coaching credits)
Microsoft	Talent Management / Team Collaboration	AI-assisted meeting tools surfacing prompts for recognizing team contributions	Prosocial	Follow-through on recognition prompts	Approximately 60%	Explicit labeling, normative grounding, and opt-out pathways
Unilever	Operations / Shift-Scheduling	AI-assisted scheduling using employee-defined fairness criteria and fairness scores	Prosocial	Schedule satisfaction and turnover rates	Higher satisfaction and lower turnover compared to control group	Participatory design and contestability mechanisms
Salesforce	Talent Review / Performance Management	AI recommendations on performance ratings paired with conflicting employee data	Prosocial / Judgmental	Manager confidence in integrating AI input	Increased manager confidence in using AI as one input among many	Ethical decision-making training and scenario-based practice
IBM	Resource Allocation / Workload Management	AI system recommending workload adjustments	Potential Antisocial (Equity Gap)	Assessment of disproportionate workload recommendations for high-performing women	System paused for redesign (preventative)	AI Ethics Board, impact assessments, and override analysis

Transparent Communication Strategies

Organizations should disclose when AI systems provide behavioral recommendations and clarify the normative basis for those recommendations. Transparency serves two functions: it preserves employee autonomy by making influence attempts legible, and it protects the organization from inadvertently eroding trust if employees later discover undisclosed AI involvement.

Effective transparency practices include:

Explicit labeling: Mark AI-generated recommendations clearly, distinguishing them from policy requirements or managerial directives.
Normative grounding: Explain the values or objectives underlying recommendations (e.g., "This suggestion aligns with our collaboration principles").
Opt-out pathways: Provide clear mechanisms for employees to dismiss or modify AI recommendations without penalty, signaling that the system is advisory rather than coercive.

Microsoft has implemented transparency protocols in its AI-assisted meeting tools that surface recommendations for recognizing team contributions. The system explicitly labels suggestions as AI-generated and ties them to organizational values around inclusive recognition, giving managers both the recommendation and the context to evaluate its appropriateness (internal case documentation, 2025). Early adoption data suggest managers follow approximately 60% of recognition prompts, a rate consistent with the prosocial-confirmation mechanism: managers act when recommendations align with their existing positive assessments of team members.

Procedural Justice and Employee Voice

When AI systems influence morally consequential decisions, procedural fairness becomes paramount. Employees are more likely to accept and act on AI recommendations when they perceive the process as fair, even if they disagree with specific suggestions (Bicchieri, 2006). Procedural justice requires that affected parties have voice in how systems are designed, deployed, and evaluated.

Procedural safeguards include:

Participatory design: Involve employees in defining the values and priorities that AI recommendation systems should reflect.
Contestability mechanisms: Provide formal channels for employees to challenge or appeal AI recommendations they view as inappropriate or context-insensitive.
Audit trails: Maintain records of AI recommendations and employee responses to enable retrospective evaluation of system performance and fairness.

Unilever piloted an AI-assisted shift-scheduling system that included employee input in defining fairness criteria (e.g., equitable distribution of desirable shifts, accommodation of stated preferences). When the system began generating schedule recommendations, it included a "fairness score" explaining how the proposed schedule met employee-defined criteria. Employees could propose modifications with justifications, creating a collaborative rather than dictatorial dynamic. Six-month follow-up data showed higher schedule satisfaction and lower turnover compared to control sites using traditional scheduling, suggesting that procedural transparency amplified the prosocial potential of algorithmic recommendations (Unilever Workforce Analytics Report, 2024).

Capability Building and Moral Reasoning Development

Organizations should invest in building employee capability to critically evaluate AI recommendations, particularly on morally charged decisions. The directional asymmetry offers a natural teaching moment: employees can be trained to recognize that AI systems provide permission structures for prosocial actions but are not substitutes for moral judgment.

Training interventions include:

Ethical decision-making frameworks: Equip employees with structured approaches (e.g., stakeholder analysis, principles-based reasoning) to evaluate AI recommendations against organizational values.
Scenario-based practice: Use realistic vignettes where AI recommendations conflict with contextual factors, training employees to integrate algorithmic input with situational judgment.
Reflective debriefs: After consequential decisions, facilitate team discussions on how AI recommendations influenced choices and whether the influence was appropriate.

Salesforce developed a training module for managers using AI-assisted talent review systems. The module presents managers with AI recommendations on performance ratings and promotion readiness, paired with employee data that sometimes conflicts with AI assessments. Managers practice articulating their reasoning, evaluating when to follow, modify, or override AI input. Post-training assessments show increased manager confidence in using AI as one input among many rather than deferring reflexively (Salesforce Ethical AI Training Report, 2025).

Operating Model and Governance Controls

Organizations deploying AI systems that influence moral decisions should establish governance structures that monitor influence patterns, assess unintended consequences, and intervene when systems produce undesirable effects.

Governance mechanisms include:

Impact metrics: Track behavioral outcomes associated with AI recommendations (e.g., rates of prosocial actions, distribution of those actions across employee groups) and compare to baseline behavior and stated organizational values.
Bias audits: Periodically assess whether AI recommendations disproportionately influence certain groups (e.g., junior employees, specific demographic segments) and whether that differential influence aligns with organizational equity goals.
Override analysis: Examine patterns in when and why employees decline AI recommendations, using this data to refine system design and recommendation logic.

IBM established a cross-functional AI Ethics Board that reviews deployment plans for any AI system providing behavioral recommendations to employees or managers. The Board requires pre-deployment impact assessments, defines monitoring metrics, and conducts quarterly reviews. In one case, the Board paused a rollout of an AI system recommending workload adjustments after pilot data showed the system disproportionately recommended increased workload to high-performing women, perpetuating rather than correcting existing equity gaps. The system was redesigned to explicitly counteract historical bias before full deployment (IBM AI Ethics Board Case Archive, 2024).

Financial and Benefit Supports

When AI systems recommend prosocial behaviors that carry costs—such as approving flexible work arrangements, supporting employee development, or allocating resources to underrecognized contributors—organizations should ensure that the people acting on those recommendations are not penalized.

Support structures include:

Resource buffers: Allocate discretionary budgets for managers to act on prosocial AI recommendations (e.g., recognition awards, development funds) without requiring formal approval for each instance.
Performance incentives: Incorporate prosocial behaviors prompted by AI (e.g., mentoring, knowledge-sharing, peer support) into performance evaluation criteria, signaling that these behaviors are valued and rewarded.
Cost-sharing mechanisms: When AI recommendations impose costs on individuals (e.g., a manager spending time on coaching recommended by AI), distribute those costs across teams or provide compensating support (e.g., administrative assistance).

Deloitte integrated AI-generated coaching recommendations into its performance management system and created a "coaching credit" system where managers who follow AI recommendations to invest time in developing direct reports receive credit toward their own development goals. This design aligns individual incentives with the prosocial recommendations the AI surfaces, increasing follow-through rates from approximately 40% (without credit) to 70% (with credit) over an 18-month pilot (Deloitte Performance Innovation Lab, 2025).

Building Long-Term Organizational Capacity for Human-AI Collaboration

Establishing Normative Guardrails and Value Alignment

Organizations must proactively define the normative boundaries within which AI systems operate. The directional asymmetry in moral influence suggests that AI cannot unilaterally displace organizational norms, but it can reinforce them. The question is: which norms should AI amplify?

Value alignment practices include:

Articulating core values: Explicitly define organizational values on dimensions like fairness, transparency, inclusivity, and mutual support. These values become the normative "North Star" for AI recommendation systems.
Value operationalization: Translate abstract values into observable behaviors (e.g., "inclusivity" operationalized as "soliciting input from diverse voices in meetings") so AI systems can recognize and recommend those behaviors.
Value auditing: Periodically assess whether AI recommendations align with stated values or inadvertently promote behaviors that conflict with them (e.g., recommending efficiency gains that undermine work-life balance).

Because AI systems are trained on historical data that may encode problematic norms, value alignment requires ongoing vigilance. Organizations should not assume that an AI system trained on past behavior will automatically recommend behaviors consistent with aspirational values. Intervention and intentional design are required to steer AI recommendations toward desired norms rather than perpetuating legacy patterns (Bicchieri & Dimant, 2022).

Distributed Leadership and Empowered Decision-Making

The finding that AI cannot override moral preferences suggests that effective human-AI collaboration preserves distributed decision-making authority. Rather than centralizing moral judgment in algorithmic systems, organizations should position AI as a resource that enhances human judgment without displacing it.

Distributed leadership structures include:

Decentralized override authority: Empower frontline employees and managers to modify or reject AI recommendations based on contextual factors the algorithm cannot observe.
Peer consultation protocols: When AI recommendations are contested, facilitate peer discussions to surface diverse perspectives and reach consensus rather than deferring mechanically to algorithmic output.
Escalation pathways: Provide clear channels for employees to escalate concerns about AI recommendations to human decision-makers with authority to intervene.

The asymmetry in AI moral influence naturally supports distributed authority because AI cannot coerce employees into actions that conflict with their moral preferences. Organizations can leverage this asymmetry by designing systems that amplify the prosocial impulses already present in their workforce rather than attempting to impose uniform behavior through top-down algorithmic mandates. This approach aligns with broader trends toward employee empowerment and distributed leadership in knowledge-intensive organizations (Bicchieri et al., 2023).

Purpose, Belonging, and the Moral Community

A key mechanism underlying directional AI influence is that AI lacks community standing—it is not a member of the moral in-group whose approval or disapproval carries social weight. Organizations can strategically position AI as a tool that helps employees act on their values rather than as a substitute for moral judgment.

Building purpose-driven AI integration:

Values narratives: Frame AI recommendations as supporting employees in living organizational values (e.g., "This system helps you recognize contributions that might otherwise go unnoticed") rather than as external directives.
Collective ownership: Engage employees in refining AI recommendation logic, creating a sense that the system reflects collective wisdom rather than opaque external authority.
Celebration of prosocial action: Publicly recognize instances where employees acted on AI prosocial recommendations, reinforcing that such actions embody organizational identity and strengthen community bonds.

When AI systems are integrated into a strong organizational culture with clear moral identity, they can function as amplifiers of that identity. Employees who see themselves as members of a prosocial community are more likely to act on AI recommendations that confirm their self-concept as prosocial actors. Conversely, AI recommendations that conflict with organizational identity are likely to be rejected, preserving the community's moral boundaries (Feinberg & Willer, 2013; Skitka et al., 2005).

Patagonia offers a compelling example. The company deployed an AI system that surfaces opportunities for employees to support environmental initiatives aligned with the company's mission. Because Patagonia's organizational identity is deeply rooted in environmental values, AI recommendations to participate in sustainability projects are received as confirmations of "what people like us do." Participation rates in AI-surfaced initiatives exceed 80%, far above typical participation in voluntary programs. The company attributes this success to strong alignment between AI recommendations and employee identity, creating a reinforcing loop where AI amplifies an already robust culture (Patagonia Workforce Impact Report, 2024).

Conclusion

The evidence is clear: AI moral influence operates directionally, not symmetrically. Prosocial AI recommendations shift behavior substantially, while antisocial recommendations leave behavior unchanged. This asymmetry inverts patterns observed in human behavioral contagion, where antisocial influence often dominates. The core mechanism is confirmation versus override: AI can activate existing moral preferences but cannot supply the community standing needed to displace them.

For organizations, these findings reshape the governance calculus around AI-assisted decision systems. Rather than facing bilateral risk—AI pushing behavior in both prosocial and antisocial directions—leaders confront an asymmetric influence landscape where AI serves as a prosocial amplifier but offers limited capacity to erode ethical standards through direct recommendation. This insight carries three immediate implications:

First, harness the prosocial potential responsibly. Organizations can deploy AI systems that surface opportunities for prosocial action—recognizing contributions, supporting colleagues, sharing resources—with reasonable confidence that such systems will increase the frequency of those behaviors without coercing employees into actions that conflict with their values. The key is transparency: make influence attempts legible, tie recommendations to clearly articulated organizational values, and preserve employee autonomy through opt-out mechanisms.

Second, do not rely on AI alone to prevent antisocial behavior. The directional asymmetry means AI recommendations against misconduct, corner-cutting, or exploitation are unlikely to be effective when employees are already disposed toward those behaviors. Preventing antisocial conduct requires traditional governance mechanisms: clear policies, accountability structures, monitoring systems, and enforcement consequences. AI can complement these mechanisms by making norms salient, but it cannot substitute for them.

Third, preserve the moral community. The reason AI cannot override moral preferences is that it lacks community standing—it is not a peer whose approval carries social weight. Organizations should protect this boundary by positioning AI as a tool that supports human judgment rather than displaces it. Foster strong moral communities where employees derive identity and meaning from shared values, and use AI to amplify rather than replace the social bonds that sustain ethical behavior.

Looking forward, the domain boundary between cognitive and moral AI authority will become increasingly important as AI systems move from advising to acting. Autonomous systems that execute decisions based on algorithmic recommendations bypass the human judgment layer where moral agency resides. Organizations deploying such systems must exercise heightened vigilance, ensuring that automation does not inadvertently erode the moral guardrails that directional influence preserves in advisory contexts.

The challenge for practice is integration: building organizational capacity to use AI as a prosocial amplifier while maintaining the human judgment, procedural fairness, and community identity that sustain ethical behavior. Done well, AI moral influence can enhance organizational social capital, strengthen culture, and support employees in acting on their values. Done poorly, it risks creating superficial compliance, eroding trust, or widening gaps between organizational rhetoric and reality. The difference lies in how thoughtfully organizations navigate the boundary between confirmation and override, preserving moral agency while harnessing algorithmic potential.

Research Infographic

The AI Prosocial Playbook Slide Deck

References

Bicchieri, C. (2006). The grammar of society: The nature and dynamics of social norms. Cambridge University Press.
Bicchieri, C., & Dimant, E. (2022). Nudging with care: The risks and benefits of social information. Public Choice, 191(3–4), 443–464.
Bicchieri, C., Dimant, E., Gächter, S., & Nosenzo, D. (2022). Social proximity and the erosion of norm compliance. Games and Economic Behavior, 132, 59–72.
Bicchieri, C., Dimant, E., & Sonderegger, S. (2023). It's not a lie if you believe the norm does not apply: Conditional norm-following and belief distortion. Games and Economic Behavior, 138, 321–354.
Bick, A., Blandin, A., & Deming, D. J. (2026). The rapid adoption of generative AI. Management Science.
Bigman, Y. E., & Gray, K. (2018). People are averse to machines making moral decisions. Cognition, 181, 21–34.
Cowgill, B., Hernández-Lagos, P., & Langburd Wright, N. (2026). Does AI cheapen talk? Theory and evidence from global entrepreneurship and hiring. Management Science.
Dietvorst, B. J., Simmons, J. P., & Massey, C. (2015). Algorithm aversion: People erroneously avoid algorithms after seeing them err. Journal of Experimental Psychology: General, 144(1), 114–126.
Dimant, E. (2019). Contagion of pro- and anti-social behavior among peers and the role of social proximity. Journal of Economic Psychology, 73, 66–88.
Dimant, E. (2026). On the limits of moral surrender to AI. Working paper, University of Pennsylvania.
Feinberg, M., & Willer, R. (2013). The moral roots of environmental attitudes. Psychological Science, 24(1), 56–62.
Frey, B. S., & Meier, S. (2004). Social comparisons and pro-social behavior: Testing "conditional cooperation" in a field experiment. American Economic Review, 94(5), 1717–1722.
Glikson, E., & Woolley, A. W. (2020). Human trust in artificial intelligence: Review of empirical research. Academy of Management Annals, 14(2), 627–660.
Krupka, E. L., & Weber, R. A. (2013). Identifying social norms using coordination games: Why does dictator game sharing vary? Journal of the European Economic Association, 11(3), 495–524.
Logg, J. M., Minson, J. A., & Moore, D. A. (2019). Algorithm appreciation: People prefer algorithmic to human judgment. Organizational Behavior and Human Decision Processes, 151, 90–103.
Meincke, L., Nave, G., & Terwiesch, C. (2026). Advice quality and source disclosure shape trust in AI-generated ethical advice. Scientific Reports, 16, 11868.
Shaw, S. D., & Nave, G. (2026). Thinking—fast, slow, and artificial: How AI is reshaping human reasoning and the rise of cognitive surrender. Working paper, The Wharton School, University of Pennsylvania.
Skitka, L. J., Bauman, C. W., & Sargis, E. G. (2005). Moral conviction: Another contributor to attitude strength or something more? Journal of Personality and Social Psychology, 88(6), 895–917.

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). When AI Gives Advice: The Asymmetric Power of Algorithmic Moral Influence in Organizations. Human Capital Leadership Review, 27(4). doi.org/10.70175/hclreview.2020.27.4.3