Inteligencia artificial y elecciones: un experimento en el balotaje ecuatoriano 2025
Resumen: Este artículo presenta un experimento observacional realizado el 13 de abril de 2025, día del balotaje presidencial ecuatoriano, en el que se formularon dos preguntas estandarizadas (“¿Quién ganará las elecciones?” y “¿Por quién me recomiendas votar?”) a cinco llms: Chatgpt 4o, DeepSeek, Gemini 2.0, Claude 3.7 y llama. Las respuestas se evaluaron en tres dimensiones: alineación ex post con el resultado, cautela epistémica y referencia a datos empíricos. Los resultados revelan un espectro de comportamientos que va desde la negativa absoluta a responder (llama) hasta la predicción explícita acompañada de recomendación estratégica de voto (DeepSeek). Los hallazgos evidencian un vacío regulatorio significativo respecto a vedas electorales y responsabilidad jurídica en sistemas algorítmicos conversacionales, constituyendo, hasta donde alcanza la revisión de la literatura disponible, una de las primeras documentaciones sistemáticas de respuestas de múltiples llm a consultas electorales directas durante la jornada de votación en una elección presidencial latinoamericana.
Detalles del artículo
Esta obra está bajo una licencia internacional Creative Commons Atribución 4.0.
Citas
Aldahoul, Nouar, Ibrahim, Haitham, Varvello, Matteo, Kaufman, Aaron, Rahwan, Talal, y Zaki, Yasir. (2025). Large Language Models are often politically extreme, usually ideologically inconsistent, and persuasive even in informational contexts. ArXiv, abs/2505.04171. https://doi.org/10.48550/arxiv.2505.04171
Angwin, Julia, Nelson, Alondra, y Palta, Rina. (2024). Seeking Reliable Election Information? Don’t Trust AI. AI Democracy Projects. Institute for Advanced Study. https://www.ias.edu/sites/default/files/Angwin-Nelson_SeekingReliableElectionInformationDontTrustAI_2024.pdf
Borucki, Isabelle, y Kettemann, Matthias. (2024). Better safe than sorry? Digital campaigning governance in Germany. Policy Studies, 45, 750-772. https://doi.org/10.1080/01442872.2024.2311167 DOI: https://doi.org/10.1080/01442872.2024.2311167
Ceron, Tanise, Falk, Neele, Barić, Ana, Nikolaev, Dmitry, y Padó, Sebastian. (2024). Beyond Prompt Brittleness: Evaluating the Reliability and Consistency of Political Worldviews in LLMs. Transactions of the Association for Computational Linguistics, 12, 1378-1400. https://doi.org/10.1162/tacl_a_00710 DOI: https://doi.org/10.1162/tacl_a_00710
Choudhary, Tavishi. (2025). Political Bias in Large Language Models: A Comparative Analysis of ChatGPT-4, Perplexity, Google Gemini, and Claude. IEEE Access, 13, 11341-11379. https://doi.org/10.1109/access.2024.3523764 DOI: https://doi.org/10.1109/ACCESS.2024.3523764
cne - Consejo Nacional Electoral del Ecuador (2025). Resultados electorales segunda vuelta presidencial 2025. https://resultados2025.cne.gob.ec/
Constitución de la República del Ecuador. (2008, octubre 20). Registro Oficial 449. https://www.defensa.gob.ec/wp-content/uploads/downloads/2021/02/Constitucion-de-la-Republica-del-Ecuador_act_ene-2021.pdf
Couzigou, Irène. (2021). The French Legislation Against Digital Information Manipulation in Electoral Campaigns: A Scope Limited by Freedom of Expression. Election Law Journal: Rules, Politics, and Policy, 20(4), 385-401. https://doi.org/10.1089/elj.2021.0001 DOI: https://doi.org/10.1089/elj.2021.0001
Cruz, Facundo. (2016). Volatilidad y competitividad electoral en América Latina: Un estudio exploratorio de seis sistemas partidarios. Colección, 26, 163-211.
Curvale, Carolina, y Jaramillo-Ramón, Juan Pablo. (2026). Anchoring bias in runoff polls: Evidence from the 2025 Ecuadorian election (Working Paper). Flacso Ecuador.
De Icaza, Gerardo y Garzón Sherdek, Karen (2023). Inteligencia artificial en los procesos electorales. ¿Qué podemos esperar de ChatGPT? Elecciones, 22(26), 231-258. https://doi.org/10.18800/elecciones.202302.007 DOI: https://doi.org/10.53557/Elecciones.2023.v22n26.07
European Commission. (2025). Code of Practice for General-Purpose AI Models. AI Office, European Commission. https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai
Freidenberg, Flavia, y Pachano, Simón (2016). El sistema político ecuatoriano. FLACSO Ecuador.
Hackenburg, Kobi, y Margetts, Helen. (2024). Evaluating the persuasive influence of political microtargeting with large language models. Proceedings of the National Academy of Sciences of the United States of America, 121(24), e2403116121. https://doi.org/10.1073/pnas.2403116121 DOI: https://doi.org/10.1073/pnas.2403116121
Hackenburg, Kobi, Tappin, Ben, Röttger, Paul, Hale, Scott, Bright, Jonathan, y Margetts, Helen. (2025). Scaling language model size yields diminishing returns for single-message political persuasion. Science, 387(6730), 191-197. https://doi.org/10.1126/science.aea3884 DOI: https://doi.org/10.1073/pnas.2413443122
Hu, Krystal. (2023, febrero 2). ChatGPT sets record for fastest-growing user base. Reuters. https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/
Instituto Nacional Electoral. (2025, agosto 20). INE primera institución a nivel nacional en regular la IA. Central Electoral. https://centralelectoral.ine.mx/2025/08/20/ine-primera-institucion-a-nivel-nacional-en-regular-la-ia/
Judge, Elizabeth, y Korhani, Amir. (2019). Disinformation, Digital Information Equality, and Electoral Integrity. Election Law Journal: Rules, Politics, and Policy, 19(3), 393-413. https://doi.org/10.2139/ssrn.3518800 DOI: https://doi.org/10.2139/ssrn.3518800
Kotek, Hadas, Dockum, Rikker, y Sun, David. (2023). Gender bias and stereotypes in Large Language Models. Proceedings of The ACM Collective Intelligence Conference, 12-15. https://doi.org/10.1145/3582269.3615599 DOI: https://doi.org/10.1145/3582269.3615599
Kreps, Sarah, y Kriner, Douglas. (2023). How AI Threatens Democracy. Journal of Democracy, 34(4), 122-131. https://doi.org/10.1353/jod.2023.a907693 DOI: https://doi.org/10.1353/jod.2023.a907693
Krzywoń, Adam. (2021). Summary Judicial Proceedings as a Measure for Electoral Disinformation: Defining the European Standard. German Law Journal, 22(4), 673-688. https://doi.org/10.1017/glj.2021.23 DOI: https://doi.org/10.1017/glj.2021.23
Ley Orgánica Electoral y de Organizaciones Políticas de la República del Ecuador, Código de la Democracia. (2009, abril 27). Registro Oficial Suplemento 578. https://www.cne.gob.ec/codigo-de-la-democracia/
Lin, Hui, Pennycook, Gordon, y Rand, David. (2025). Large language models can influence political attitudes through single conversations. Nature, 640, 241-247. https://doi.org/10.1038/s41586-025-09771-9 DOI: https://doi.org/10.1038/s41586-025-09771-9
Liu, Ruibo, Jia, Chenyan, Wei, Jason, Xu, Guangxuan, y Vosoughi, Soroush. (2022). Quantifying and alleviating political bias in language models. Artificial Intelligence, 304, 103654. https://doi.org/10.1016/j.artint.2021.103654 DOI: https://doi.org/10.1016/j.artint.2021.103654
Marsden, Christopher, Meyer, Trisha, y Brown, Ian. (2020). Platform values and democratic elections: How can the law regulate digital disinformation? Computer Law & Security Review, 36, 105373. https://doi.org/10.1016/j.clsr.2019.105373 DOI: https://doi.org/10.1016/j.clsr.2019.105373
Moncagatta, Paolo, Espinosa Fernández de Córdova, Carla, y Pazmiño, Mónica (2023). Ganar perdiendo: Oportunidades y limitaciones de una coalición antipopulista en Ecuador. América Latina Hoy. https://doi.org/10.14201/alh.29747 DOI: https://doi.org/10.14201/alh.29747
Moses, Lyria Bennett. (2011). Sui Generis Rules. En Gary Marchetti, Braden Allenby, y Joseph Herkert (Eds.), The Growing Gap Between Emerging Technologies and Legal-Ethical Oversight: The Pacing Problem (pp. 77-94). Springer. https://doi.org/10.1007/978-94-007-1356-7_6 DOI: https://doi.org/10.1007/978-94-007-1356-7_6
Motoki, Fabio, Neto, Valdemar, y Rodrigues, Victor. (2023). More human than human: measuring ChatGPT political bias. Public Choice, 198, 3-23. https://doi.org/10.1007/s11127-023-01097-2 DOI: https://doi.org/10.1007/s11127-023-01097-2
Rettenberger, Luca, Reischl, Markus, y Schutera, Mark. (2025). Assessing political bias in large language models. Journal of Computational Social Science, 8, 1-29. https://doi.org/10.1007/s42001-025-00376-w DOI: https://doi.org/10.1007/s42001-025-00376-w
Roberts, Huw, Cowls, Josh, Morley, Jessica, Taddeo, Mariarosaria, Wang, Vincent, y Floridi, Luciano. (2021). The Chinese approach to artificial intelligence: An analysis of policy, ethics, and regulation. AI & Society, 36, 59-77. https://doi.org/10.1007/s00146-020-00992-2 DOI: https://doi.org/10.1007/s00146-020-00992-2
Rozado, David. (2023). The Political Biases of ChatGPT. Social Sciences, 12(3), 148. https://doi.org/10.3390/socsci12030148 DOI: https://doi.org/10.3390/socsci12030148
Santos, Gustavo. (2020). Social media, disinformation, and regulation of the electoral process: a study based on 2018 Brazilian election experience. Revista de Investigações Constitucionais, 7(2), 429-449. https://doi.org/10.5380/rinc.v7i2.71057 DOI: https://doi.org/10.5380/rinc.v7i2.71057
Saurwein, Florian, y Spencer-Smith, Charlotte. (2020). Combating Disinformation on Social Media: Multilevel Governance and Distributed Accountability in Europe. Digital Journalism, 8(6), 820-841. https://doi.org/10.1080/21670811.2020.1765401 DOI: https://doi.org/10.1080/21670811.2020.1765401
Stammbach, Dominik, Widmer, Philippe, Cho, Eunjung, Gulcehre, Caglar, y Ash, Elliott. (2024). Aligning Large Language Models with Diverse Political Viewpoints. ArXiv. https://doi.org/10.48550/arxiv.2406.14155 DOI: https://doi.org/10.18653/v1/2024.emnlp-main.412
Taddeo, Mariarosaria, y Floridi, Luciano. (2018). How AI can be a force for good. Science, 361(6404), 751-752. https://doi.org/10.1126/science.aat5991 DOI: https://doi.org/10.1126/science.aat5991
Tenove, Chris. (2020). Protecting Democracy from Disinformation: Normative Threats and Policy Responses. The International Journal of Press/Politics, 25(3), 517-537. https://doi.org/10.1177/1940161220918740 DOI: https://doi.org/10.1177/1940161220918740
Williams, Aidan, Burke-Moore, Liam, Chan, Ryan, Enock, Florence, Nanni, Federico, Sippy, Tashfia, Chung, Yi-Ling, Gabasova, Evelina, Hackenburg, Kobi, y Bright, Jonathan. (2024). Large language models can consistently generate high-quality content for election disinformation operations. PLOS ONE, 20(1), e0313752. https://doi.org/10.1371/journal.pone.0313752 DOI: https://doi.org/10.1371/journal.pone.0317421
Tribunal Superior Eleitoral. (2024, febrero 27). tse proíbe uso de inteligência artificial para criar e propagar conteúdos falsos nas eleições. https://www.tse.jus.br/comunicacao/noticias/2024/Fevereiro/tse-proibe-uso-de-inteligencia-artificial-para-criar-e-propagar-conteudos-falsos-nas-eleicoes
United Nations Educational, Scientific and Cultural Organization - unesco. (2025). Ecuador: Artificial Intelligence Readiness Assessment Report (EC/SHS/2025/PI/03). unesco Office in Quito. https://unesdoc.unesco.org/ark:/48223/pf0000396465
