Algorithmic Bias Perception, Institutional Trust, and Citizens' Acceptance of AI Governance Frameworks: A Multinational PLS-SEM Study

Resumen

The rapid deployment of artificial intelligence in high-stakes public and commercial domains has elevated algorithmic governance from a technical subspecialty to a central challenge of democratic governance. Despite the proliferation of regulatory frameworks — notably the EU AI Act (2024), the U.S. Algorithmic Accountability Act proposals, and China's Interim Measures for Generative AI (2023) — citizen acceptance of AI governance mechanisms remains poorly understood as an empirical phenomenon. Acceptance is not guaranteed by legislative mandate; it requires that citizens perceive algorithmic systems as fair, institutions as trustworthy, and governance frameworks as legitimate. This study develops and empirically tests a PLS-SEM model of Citizen Acceptance of AI Governance (CAAG), specifying the roles of Perceived Algorithmic Bias (PAB), Algorithmic Transparency Demand (ATD), Institutional Trust in AI Regulators (ITAR), and AI Risk Perception (ARP) as structural antecedents of CAAG, with digital literacy (DL) as a moderator and perceived regulatory effectiveness (PRE) as a mediator. A stratified quota sample of N = 328 adult citizens (Mage = 33.7, SD = 9.2; 51.2% female) was surveyed online across four countries representing different regulatory orientations: España (EU AI Act context, n = 98), México (emerging regulatory context, n = 97), Colombia (n = 72), and Chile (n = 61), between March and June 2025. PLS-SEM with SmartPLS 4.0 was employed with 5,000 bootstrap resamples. Invariance testing enabled cross-national comparison. The measurement model achieved satisfactory reliability (all α ≥ .82; ρc ≥ .87; AVE ≥ .53; HTMT < .85). The structural model explained 56.8% of CAAG variance (R² = .568). Institutional Trust in AI Regulators was the strongest predictor (β = .447, p < .001), followed by Perceived Regulatory Effectiveness as a partial mediator of the PAB → CAAG path (indirect β = .168, 95% CI [.121, .217]). Perceived Algorithmic Bias significantly attenuated CAAG (β = −.312, p < .001). Digital literacy moderated the PAB → ATD relationship (β = .241, p = .001), amplifying transparency demands among high-literacy citizens. Cross-national MGA revealed significantly higher CAAG in the EU-regulated context (España; β difference in ITAR → CAAG = .187, p = .014). Citizen acceptance of AI governance depends primarily on institutional trust, which is itself contingent on perceived algorithmic fairness and regulatory effectiveness. Digital literacy amplifies rather than reduces governance demands among informed citizens — a finding with direct implications for regulatory communication strategy. The cross-national variance confirms that governance context, not only technical AI characteristics, shapes citizen attitudes toward algorithmic oversight.

Resumen

El rápido despliegue de la IA en dominios de alto riesgo ha elevado la gobernanza algorítmica a un desafío central de la democracia. Objetivo: Este estudio desarrolla y contrasta empíricamente un modelo PLS-SEM de Aceptación Ciudadana de la Gobernanza de la IA (ACGIA), especificando los roles del Sesgo Algorítmico Percibido (SAP), la Demanda de Transparencia Algorítmica (DTA), la Confianza Institucional en los Reguladores de IA (CIRIA) y la Percepción de Riesgo de la IA (PRIA) como antecedentes estructurales. Método: Se encuestó a N = 328 ciudadanos adultos en cuatro países con orientaciones regulatorias distintas. Se empleó PLS-SEM con SmartPLS 4.0. Resultados: El modelo explicó el 56.8% de la varianza en ACGIA (R² = .568). La Confianza Institucional fue el predictor más fuerte (β = .447, p < .001). El Sesgo Algorítmico Percibido atenuó ACGIA (β = −.312, p < .001). La Efectividad Regulatoria Percibida medió parcialmente la relación SAP → ACGIA (efecto indirecto β = .168). El análisis multigrupo reveló mayor ACGIA en el contexto regulatorio europeo.

Palabras clave: gobernanza de IA; sesgo algorítmico; confianza institucional; EU AI Act; aceptación ciudadana; PLS-SEM; legitimidad regulatoria; alfabetización digital; transparencia algorítmica; regulación comparada.

References

Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103(3), 411–423. https://doi.org/10.1037/0033-2909.103.3.411

Casolari, F., Finocchiaro, G., & Greco, G. M. (2025). Navigating the AI regulatory landscape: Balancing innovation, ethics, and global governance. Journal of Contemporary European Research, 21(1), 367–397. https://doi.org/10.1080/20954816.2025.2569584

Díaz-Rodríguez, N., Del Ser, J., Coeckelbergh, M., de Prado, M. L., Herrera-Viedma, E., & Herrera, F. (2023). Connecting the dots in trustworthy artificial intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation. Information Fusion, 99, 101896. https://doi.org/10.1016/j.inffus.2023.101896

European Commission. (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union, L(1689), 1–144.

Fan, G. (2025). The role of artificial intelligence in the digital transformation of government: Opportunities and ethical challenges. Frontiers in Public Health, 13, 1694996. https://doi.org/10.3389/fpubh.2025.1694996

Guihot, M., & Matthew, L. (2024). AI governance in a complex and rapidly changing regulatory landscape: A global perspective. Humanities and Social Sciences Communications, 11(1), Article 1232. https://doi.org/10.1057/s41599-024-03560-x

Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2022). A primer on partial least squares structural equation modeling (PLS-SEM) (3rd ed.). Sage.

Hayes, A. F. (2022). Introduction to mediation, moderation, and conditional process analysis (3rd ed.). Guilford Press.

Levi, M., & Stoker, L. (2000). Political trust and trustworthiness. Annual Review of Political Science, 3(1), 475–507. https://doi.org/10.1146/annurev.polisci.3.1.475

Lorenz-Spreen, P., Oswald, L., Lewandowsky, S., & Hertwig, R. (2023). A systematic review of worldwide causal and correlational evidence on digital media and democracy. Nature Human Behaviour, 7(1), 74–101. https://doi.org/10.1038/s41562-022-01460-1

Moon, M. J. (2023). Searching for inclusive artificial intelligence for social good: Participatory governance and policy recommendations for making AI more inclusive and benign for society. Public Administration Review, 83(6), 1496–1505. https://doi.org/10.1111/puar.13651

Spurava, G., & Kotilainen, S. (2023). Digital literacy as a pathway to professional development in the algorithm-driven world. Nordic Journal of Digital Literacy, 18(1), 48–59. https://doi.org/10.18261/njdl.18.1.5

Suchman, M. C. (1995). Managing legitimacy: Strategic and institutional approaches. Academy of Management Review, 20(3), 571–610. https://doi.org/10.5465/amr.1995.9508080331