Telling the public an AI — not humans — picked a company’s CSR initiative makes the firm look less competent, fair and authentic; people’s general attitudes toward AI determine how strongly competence judgments fall while technical knowledge does not.

Main Finding

When a firm discloses that an AI system (rather than a human team) led selection and recommendation of a CSR initiative, the public perceives the organization as less competent, fair, and transparent; attributes it less public-serving motive; and views the CSR as less authentic. Perceptions of capability predict motive attributions, which in turn predict perceived CSR authenticity. Disclosing AI shifts the level of these judgments but does not change the underlying attributional process. Attitudes toward AI moderate the effect on capability perceptions; AI literacy does not.

Key Points

• The study applies attribution theory: the disclosed decision-maker (AI vs human) serves as a source cue affecting capability judgments, motive attributions, and perceived CSR authenticity.
• Experimental manipulation: identical CSR initiative described as originating from either an AI system or a human team.
• Main outcomes reduced by AI attribution: perceived organizational competence, fairness, transparency, public-serving motive, and CSR authenticity.
• Causal chain observed: perceived capability → public-serving motive attribution → perceived CSR authenticity.
• Structure of the attributional process is invariant across disclosures; only the level of judgments shifts when AI is disclosed.
• Moderation: participants' general attitudes toward AI changed how strongly the AI disclosure affected capability perceptions. AI literacy (knowledge) had no moderating effect.

Data & Methods

• Design: Online randomized experiment (between-subjects).
• Sample: 517 U.S. adults recruited via Dynata panel.
• Treatment: Corporate press release vignette describing the same CSR initiative attributed either to an algorithmic/AI system or to a human team; managers review algorithm recommendations in the vignette context.
• Measures: Perceived organizational capability (competence, fairness, transparency), attributions of public-serving motive, perceived CSR authenticity; measures of attitudes toward AI and AI literacy used as moderators.
• Analysis: Mediation analysis showing capability predicts motive attributions, which predict authenticity; tests of moderation and invariance of process structure across disclosure conditions.

Implications for AI Economics

• Adoption trade-offs: Firms face a trade-off between the potential efficiency gains from algorithmic CSR decision-making and losses in perceived authenticity and stakeholder trust that could lower CSR effectiveness or brand value.
• Signaling and disclosure strategy: How firms disclose AI involvement matters economically—explicit AI attribution can depress perceived CSR value. Firms may need communication strategies (e.g., emphasizing human oversight or explainability) to mitigate negative perceptions.
• Targeting and segmentation: Heterogeneous public attitudes toward AI imply adoption and disclosure strategies should be tailored—markets with more positive AI attitudes will be less punishing of algorithmic decision-making.
• Investment and governance: Organizations may prefer human-in-the-loop designs, transparency practices, or third-party audits to preserve perceptions of competence, fairness, and authenticity while leveraging AI efficiencies.
• Market outcomes and stakeholder behavior: Reduced perceived authenticity may affect consumer support, donations, employee engagement, and investor assessments, with potential downstream effects on firm performance and CSR returns.
• Policy and regulation: Findings support the relevance of disclosure rules, explainability standards, and governance for algorithmic decisions in public-serving domains to maintain public trust.
• Future research suggestions (economic questions): quantify market impacts of AI-attributed CSR on donations/sales/stock returns; explore industry-specific effects; measure long-run dynamics (learning, habituation); test mitigation strategies (explainability, human oversight labels) and their cost-benefit trade-offs.