This paper argues that the attribution of agency and intentionality to large language models (LLMs) and contemporary chatbots should not be explained merely as the effect of anthropomorphic cues that lead users to mistakenly ascribe human-like mental properties to machines, as in the famous case of Weizenbaum’s “ELIZA effect” from the 1960s. Rather, it arises from the way conversational technologies activate and organise human conversational behaviour. These systems do not primarily induce beliefs about machine minds; instead, they draw users into familiar language games, calling on them to perform established practices of address, response, clarification, agreement, and follow-up. Agency attribution, on this view, is not inferred from perceived features but enacted through participation in conversational practices. 
If users report a sense of being addressed by an intentional voice, simultaneously they often deny that the system possesses subjective understanding. The reason, it will be argued, is that agency is not “detected” in LLMs but assigned through familiar interpretive schemas that persist even in the face of explicit knowledge of the systems’ non-semantic, non-subjective nature. This tension is not a contradiction once one recognizes that the operative level is that of practice rather than belief. When a system produces turns that match familiar conversational expectations (such as requests for clarification, acknowledgements, or reformulations), users respond in kind, adopting the normative postures associated with dialogue. The interaction thus acquires a form of self-normativity: once the user accepts the system as a conversational partner, the exchange begins to generate its own local standards of adequacy, relevance, cooperation, and trust, even when the user knows that no genuine understanding is involved. Agency attribution, if at all, operates through habituated modes of response activated by rhetorics of coherence, signals of reciprocal recognition, affective cues, and communicative presence, independently of any attribution of inner states. 
Contemporary conversational technologies institutionalize and privatize this conversational engagement. What distinguishes these systems, therefore, is not primarily the simulation of anthropomorphic appearance, but the harnessing and mobilization of intersubjective habits of interaction. Anthropomorphism is not an illusion or a side effect, but an infrastructural condition of access to computational systems. Anthropomorphism, in other words, makes the interface extractive toward human practices of interaction: from trust-signals, politeness, repetition and corrections to dialogical acts of negotiation, exchange of emotional cues and so on. 
Conclusively, the paper then situates this performative structure within the political economy of language automation. Interfaces built around dialogue, politeness, and reciprocal recognition do more than improve usability: they stabilise participation in language games and make that participation economically useful. The attribution of an authorial or intentional voice is actively cultivated, as it supports engagement, delegation, and forms of reliance that strengthen users’ attachment to these systems. 
Ultimately, by shifting the focus from anthropomorphic illusion to performative participation, the paper revises contemporary debates on anthropomorphism and machine agency: users actively participate in the performative construction of an interlocutor by engaging in dialogical language games. What appears as confusion about the true nature of artificial authorship is better understood as the managed activation of human conversational practices within automated systems. Agency, on this view, is neither discovered in machines nor falsely projected onto them, but enacted through participation in language games that conversational technologies are designed to elicit and sustain.

Emanuele Arielli is Professor of Aesthetics at the IUAV University in Venice, Italy. His research focuses on the intersections between aesthetics, philosophy of technology, epistemology and philosophy of language, with particular attention to the impact of artificial intelligence on media and art. He has taught at several institutions in Italy and Germany. Arielli has authored several books and numerous academic papers, collaborating with international institutions. His last book Artificial Aesthetics, 2024 (written with media-theorist Lev Manovich) has been translated into Spanish, Italian, Chinese and Japanese. Publications under https://iuav.academia.edu/EmanueleArielli. Orcid: https://orcid.org/0000-0003-2039-5196

AI futures are often presented as settled, but they rest on three assumptions that seldom get questioned. Utility legitimizes access to technology for peripheral users only when it serves productive ends. Authorship places creative value in individual creators and novel outputs. Scale assumes innovation spreads by working the same way everywhere. Together they cast the Majority World as an endpoint of diffusion, a place where technologies designed elsewhere are simply adopted. Drawing on the past six years of research on technological adoption and agency in Latin America, this keynote unsettles each assumption in turn. It argues that a more nuanced examination of technology appropriation in the Majority World is itself a way of making futures. 

The first case moves beyond utility. Research on the digital leisure divide among forcibly displaced Venezuelans in Brazil shows that people value connectivity mostly for play, intimacy, entertainment and content creation. Humanitarian and development agendas treat these uses as trivial, and that neglect reveals a moral economy that decides whose digital lives count as legitimate. The second moves beyond authorship. An Antropofagia approach, rooted in the Latin American practice of symbolically cannibalizing colonial influences, puts collective, process-oriented ideas of creative value in place of the individualized, product-centered models of ownership found in Western AI debates. The third moves beyond scale. Through the concept of artificial mediations, creative practitioners across four Latin American countries show how place-based, labor and aesthetic frictions disrupt AI’s promise of seamless expansion and open room to negotiate platform power. Moving beyond utility, authorship and scale means treating Latin America as a source of theory, not a recipient of technology. Friction, play and appropriation then become ways to analyze and practice more plural, situated and hopeful technological futures. 

Keywords: Agency; Creativity; Decoloniality; Majority Word; Generative AI 

Daniela Jaramillo-Dent is an SNSF Ambizione Fellow and Senior Research and Teaching Associate at the University of Zurich. She studies creators and creatives in dominant and peripheral spaces: examining how they make sense of, adapt to, and push back against the technologies that shape their work. Her research spans creators in Europe and creative communities in Latin America and the Majority World, asking how AI and platforms redistribute visibility, value, and creative agency between these spaces. Her current work traces the frictions generative AI creates in Latin American creative sectors. Her award-winning work on minority creators has appeared in journals such as International Journal of Cultural Studies, New Media & Society and Social Media + Society. She is Latin America Research Lead at the Inclusive AI Lab (Utrecht University), Key Regional Leader of the TikTok Cultures Research Network (Curtin University), and inaugural Associate Editor of Creator and Influencer Studies (Springer Nature).

AI futures are often presented as settled, but they rest on three assumptions that seldom get questioned. Utility legitimizes access to technology for peripheral users only when it serves productive ends. Authorship places creative value in individual creators and novel outputs. Scale assumes innovation spreads by working the same way everywhere. Together they cast the Majority World as an endpoint of diffusion, a place where technologies designed elsewhere are simply adopted. Drawing on the past six years of research on technological adoption and agency in Latin America, this keynote unsettles each assumption in turn. It argues that a more nuanced examination of technology appropriation in the Majority World is itself a way of making futures. 

The first case moves beyond utility. Research on the digital leisure divide among forcibly displaced Venezuelans in Brazil shows that people value connectivity mostly for play, intimacy, entertainment and content creation. Humanitarian and development agendas treat these uses as trivial, and that neglect reveals a moral economy that decides whose digital lives count as legitimate. The second moves beyond authorship. An Antropofagia approach, rooted in the Latin American practice of symbolically cannibalizing colonial influences, puts collective, process-oriented ideas of creative value in place of the individualized, product-centered models of ownership found in Western AI debates. The third moves beyond scale. Through the concept of artificial mediations, creative practitioners across four Latin American countries show how place-based, labor and aesthetic frictions disrupt AI’s promise of seamless expansion and open room to negotiate platform power. Moving beyond utility, authorship and scale means treating Latin America as a source of theory, not a recipient of technology. Friction, play and appropriation then become ways to analyze and practice more plural, situated and hopeful technological futures. 

Keywords: Agency; Creativity; Decoloniality; Majority Word; Generative AI 

Andrew Prahl is an Associate Professor at Nanyang Technological University, Singapore, whose research focuses on human–machine communication, human–AI trust, and the social consequences of automation. His work examines how people and organizations interpret machine authority, failure, and replacement across settings including healthcare, robotics, and Hybrid AI. He has held elected leadership roles in the International Communication Association’s Human–Machine Communication community and previously spent nearly a decade in advertising and public relations.