At a glance
- What it is: Reimagining AI in Latin America: situated narratives of users, developers, and decision-makers on understanding and governing AI
- Who: Teresa Correa (Universidad Diego Portales) along with Francisca Luco, Mónica Humeres, Dusan Cotoras, Alexandra Davidoff, Yelena Hernández-Estrada, Iñaki Oyarzún-Merino, and Claudia López. The team comes from the FAIR Millennium Nucleus (Futures of Artificial Intelligence Research), with affiliations at Universidad Diego Portales, the University of Chile, and Federico Santa María Technical University.
- Where: Communication and Change (Springer), 2025, open-access article. doi.org/10.1007/s44382-025-00012-1
- Type: qualitative study. It combines three projects carried out between 2023 and 2024: interviews and focus groups with users of public services, interviews with developers, and a digital ethnography of the sessions of Congress on AI regulation.
First reading: what it does and what it finds
It helps to start with the type of work, because it frames the rest. It is qualitative and comparative. The team does not measure whether AI works or evaluate any model: it brings together the narratives of three actors who are almost never studied at the same table. On one side, users of three algorithm-mediated public services (the School Admission System, FONASA’s virtual medical-leave analyst, and the transportation planning system): 27 interviews and six focus groups with 48 people. On another, 18 interviews with developers. And third, a digital ethnography of 51 sessions of the Chilean Congress where how to regulate AI was discussed. The question is how each group understands what AI is and what it demands of its governance.
The first finding, and the one the authors flag as the most striking, is that all three groups share the same geopolitical image: Chile as an experimental subject. Users describe themselves as guinea pigs, exposed to technologies without anyone explaining anything to them. Developers talk about a sandbox country, where an idea is tested but cannot be scaled and ends up being exported abroad to be fully carried out. And in Congress, industry voices describe the region as a playground, a place of consumption and infrastructure while the models are developed in the rich world. The authors read this as a power asymmetry between the Global North and the Global South, and connect it to a very Chilean local memory: that of the country as a neoliberal laboratory after the dictatorship. It is not a technical fact; it is the lens through which people interpret their place on the AI map.
The second finding is subtler and, I think, the most useful. What each actor demands of AI changes depending on whether they talk about AI in the abstract or about a specific system. Users, faced with the general idea of AI, deploy what the literature calls folk theories (AI as a giant encyclopedia, or as an all-seeing puppeteer) and show functional trust: they trust it as long as it works, and ask for little more. But when the focus shifts to a specific system that makes decisions about their health or their children’s school, skepticism rises and concrete demands appear: simple, visual explanations, knowing why a medical leave was rejected, and above all human mediation. One participant who had spent more than a year filing medical leave claims for postpartum depression sums up the point: she does not understand how a machine, which in her view lacks judgment, can reject a medical leave.
The other two actors complete the picture. Developers seek to demystify: they prefer to explain AI as a machine or a computing system and avoid anthropomorphic metaphors, because they consider them misleading. And decision-makers get stuck on how to define AI in order to regulate it, between broad definitions that cover even a calculator and narrow definitions that age quickly. There the authors record a familiar move: industry voices use the hammer metaphor, a neutral tool that cannot be blamed, to argue that current laws are enough and that no new regulation is needed.
Second reading: from Latin America
The first thing worth emphasizing is that this is evidence from inside the region, with fieldwork in Spanish, in real public services and in Congress itself. The AI governance literature almost always speaks from the Global North, so the contribution lies in looking closely at something rarely examined: how AI is understood and challenged when you are not the one building it.
Where you look from also matters. Chile tops several regional AI readiness indexes and has had a National AI Policy since 2021, but with low institutional trust. That combination, high adoption and low trust, is exactly what produces the functional trust the paper describes: people accept the technology as inevitable, without demanding much of it, until it touches them directly. If this happens in one of the best-positioned countries in the region, it is reasonable to think the pattern becomes more pronounced where the state is weaker. That last point is my own reading: the study was done only in Chile, and the authors themselves call for broadening the comparison before generalizing.
Where the work connects with the present is in its distinction between the abstract and the concrete. The fieldwork is from 2023 and 2024, and it mostly portrays public-service algorithms and legislative debates, not the era of mass-use chatbots. But that finding does not age; it becomes more relevant: today anyone converses with a model in the abstract without asking anything of it, and doubts appear only when that same model decides something that matters. The policy lesson the authors do draw is clear: anchor transparency, accountability, and participation in specific systems and their uses, not in the diffuse idea of AI. As my own extrapolation, that recipe fits what is coming, where it is increasingly easy to plug a model into an administrative procedure without anyone noticing the change.
The question it leaves applies to any country in the Global South. If people only demand transparency when they see the specific system that affects them, the governance task is to make those systems visible, not to stop at general principles about AI. As long as the conversation stays in the abstract, the study suggests, the concerns of those who suffer the consequences most are left out.
The fine print
- It is a qualitative study: its value lies in explaining how and why these actors understand and challenge AI, not in measuring how many Chileans think each thing. The narratives illuminate phenomena, not frequencies.
- All the fieldwork is in Chile. The authors are explicit that comparison with other countries is needed before reading these patterns as regional.
- The work with decision-makers is an ethnography of congressional sessions from 2023 and 2024, before chatbots went mainstream. Institutional matters, admittedly, age slowly.
Paper keywords: Artificial Intelligence, Situated narratives, AI governance, AI understanding, Global South, Chile
Automated reading. This text was generated by Claude, an Anthropic model, from the original source, without line-by-line human review. It may contain errors or debatable interpretations; to check any point, see the original source.