AI Art Strategies as Speculative Futures Research
The paper outlines a theoretical framework for understanding AI Art as a specific laboratory for future aesthetic processes in the context of artificial consciousness and the concept of an Agentic Entity. Its starting point is the evolutionary and cognitive background of human consciousness. This is followed by a historical overview of anticipatory models of aesthetics, from information aesthetics and telematic art through speculative aesthetics and the concept of anticipatory images to contemporary predictive models in aesthetics, neuroscience and the psychology of art. At the core of the text AI Art is presented as a field of anticipation of an Agentic Entity. On this basis the study formulates the MEPAA model (Model of Experimental Speculative Aesthetic Processes in AI Art), which combines a predictable vector with a speculative what if vector. In conclusion AI Art is described as a living experimental space in which aesthetic processes between the human and the Agentic Entity change together with the development of theories of consciousness, technical architectures and the cultural effects of long term AI use.
Abstract
This paper outlines a theoretical framework for understanding AI Art as a specific laboratory for future aesthetic processes in the context of artificial consciousness and the concept of an Agentic Entity. The point of departure is the evolutionary and cognitive background of human consciousness, including FOXP2, cultural exaptation and exaptation driven neuroplasticity, which shows that our intuitions about artificial intelligence are deeply shaped by the way the brain reuses existing neural circuits for new cultural functions (Anderson, 2010; Dehaene & Cohen, 2007; Enard et al., 2002). This perspective is followed by a historical overview of anticipatory models of aesthetics, from information aesthetics and telematic art through speculative aesthetics and the concept of anticipatory images to contemporary predictive models in aesthetics, neuroscience and the psychology of art (Ascott, 2003; Bense, 1969; Frascaroli et al., 2024; Mackay et al., 2018). At the core of the text, AI Art is presented as a field of anticipation of an Agentic Entity, analysed through the practices of Mario Klingemann, Refik Anadol and Lauren McCarthy, in which AI appears as collaborator, experimental partner and medium for modelling future configurations of agency (Loivaranta et al., 2025; McCarthy, 2018). On this basis, the study formulates the MEPAA model, the Model of Experimental Speculative Aesthetic Processes in AI Art, which combines a predictable vector based on data, trends and prediction markets such as Metaculus with a speculative vector based on what if scenarios, and proposes the remediation of existing AI Art strategies as a key tool for testing latent forms of an Agentic Entity. In conclusion, AI Art is described as a living experimental space in which aesthetic processes between the human and the Agentic Entity change together with the development of theories of consciousness, technical architectures and the cultural effects of long term AI use.
Keywords: AI Art, Agentic Entity, speculative aesthetics, exaptation, predictive models, MEPAA model, anticipatory images
Introduction: AI Art as speculative futures research
The transdisciplinary effort to build systems that can perceive, reason, learn and act, and in specific tasks even outperform humans, does not belong only to the field of computer science. It is also a fully legitimate domain of philosophy, cognitive science, neuroscience and art. The model of human made intelligence does not emerge in a vacuum; it is closely linked to the evolution of human civilisation, languages and cultural forms through which we can even imagine what artificial intelligence means. In this context, AI Art is not only a new technical category of digital art. It also operates as an experimental space where possible configurations of future aesthetic agency are tested, from current generative tools to a hypothetical Agentic Entity.
The aim of this paper is to show that contemporary artistic strategies in AI Art can be understood as a form of speculative futures research. Through artistic experiments, prototypes and remediations of existing works, they model possible scenarios for the development of artificial intelligence, its agential manifestations and their aesthetic consequences. At the same time, I propose the MEPAA model as a framework that connects predictive vectors based on data, prediction markets and technical trajectories with speculative what if vectors in the field of artistic practice.
Methodology and research design
The methodological framework of this study combines three approaches: theoretical and conceptual synthesis, analytical and interpretative case study, and speculative modelling inspired by futures research. The goal is not to empirically prove individual hypotheses, but to propose a working model, the MEPAA model, which makes it possible to systematically plan and test experiments in AI Art as a form of speculative research on futures.
In the first step, there is a targeted mapping of literature in three blocks. The first block focuses on evolutionary and cultural cognitive studies of consciousness, including FOXP2, exaptation, neuronal reuse and recycling and exaptation driven neuroplasticity (Anderson, 2010; Dehaene & Cohen, 2007; Enard et al., 2002; Firth et al., 2019). The second block covers historical and contemporary anticipatory models of aesthetics, such as information aesthetics, telematic art, speculative aesthetics and anticipatory images (Ascott, 2003; Bense, 1969; Dernbach, 2019; Mackay et al., 2018). The third block considers debates on Artificial Aesthetics, creative agency in AI Art and consciousness in AI systems, including computational functionalism, indicator based approaches and concepts of agency (Butlin et al., 2023; Manovich & Arielli, 2024; Sadegh Zadeh & Bahrami, 2025). This step provides a conceptual background for the analysis of artistic strategies.
In the second step, a case study method is applied to three paradigmatic artistic approaches, Mario Klingemann, Refik Anadol and Lauren McCarthy, which represent different configurations of creative agency. Klingemann works with distributed generative practice, Anadol with data intensive immersive works that deal with collective memory, and McCarthy with performative and interactive modelling of social relations in AI environments (del Campo, 2024; Loivaranta et al., 2025; McCarthy, 2018). Each case is analysed in three layers. The technical layer includes the type of models, data infrastructure and degree of automation. The aesthetic layer focuses on formal strategies and the handling of prediction and surprise. The agential layer examines the framing of AI as tool, coauthor or entity, and the location of decision making, adaptivity and autonomy. These layers are subsequently mapped onto the two vectors of the MEPAA model: the predictable vector, which captures technically and historically expected scenarios, and the speculative vector, which captures what if scenarios of future configurations of an Agentic Entity.
In the third step, speculative design and futures research are applied. From the analysed works, basic modules are extracted, such as data sources, model types, interaction protocols and modes of display. These modules are placed into a hypothetical environment of AGIA, Artificial General Intelligence in Art, or an Agentic Entity, for example a system with stable memory, self reflection and the ability to set its own goals within an artistic task. On this basis, remediation scenarios are proposed that ask how a work would change if AI were not only an executor of prompts but a partner that modifies instructions, selects data or generates alternative aesthetic strategies. The result is not a prediction of specific future works, but a set of experimental proposals suitable for pilot prototyping, exhibition laboratories and participatory experiments.
The fourth, reflexive and normative step confronts the model and the proposed scenarios with ethical and social questions, in particular the attribution of consciousness and agency, responsibility, and the risks of under attribution and over attribution of agency (Butlin & Lappas, 2025; Porębski & Figura, 2025). AI Art is thus explicitly understood as futures research. It is not only a field of possible aesthetic configurations, but also a space in which tensions associated with the future presence of an Agentic Entity in culture and social decision making can be articulated in advance.
Evolutionary and cognitive background: from FOXP2 to cultural exaptation
Qualitative analysis of scientific studies on the relationship between culture and cognition suggests that what we project into our current images of AI is the result of long term changes in the way the human brain works with language, symbols and abstraction. The mutation of the FOXP2 gene that affected language abilities in humans can be read as one of the crucial points at which exaptation of brain functions, their reuse for new cultural tasks, became a central evolutionary principle (Enard et al., 2002). Language, vision, planning and motor control, originally oriented toward survival, are gradually exapted for visual art, ritual, music and abstract thought (Dissanayake, 2000; Fitch, 2006; Fisher, 2017).
The concept of cultural exaptation of brain functions makes it possible to perceive artificial intelligence as another layer in this long history. When artists describe AI as collaborator, coauthor or entity, this is not only a metaphor. It is also an expression of deeply rooted cognitive schemes of attributing agency to nonhuman systems. In this sense, the current fascination with AI can be read as a continuation of older forms of animism and anthropomorphism, although now transposed into the environment of digital technologies and computational models.
Exaptation in the sense of neuronal reuse and neuronal recycling shows that the brain continuously recycles existing circuits for new cultural functions (Anderson, 2010; Dehaene & Cohen, 2007). More recent reviews on the influence of digital technologies on the brain provide empirical indications of what I call exaptation driven neuroplasticity. Long term use of digital tools, including AI, changes the distribution of attention, memory strategies and social cognition in measurable ways (de Barros, 2024; Firth et al., 2019; Korte, 2020; Small et al., 2020). Studies on cognitive offloading and the weakening of critical thinking in contexts of intensive AI tool use point to the fact that AI Art does not arise outside the brain, but directly in the process of its reorganisation (Gerlich, 2025; Shanmugasundaram & Tamilarasu, 2023).
For AI Art this means that aesthetic processes cannot be understood only as the production of images. They must also be seen as tests of how the infrastructure of perception, prediction and meaning making itself changes in an environment where generative models are constantly present.
Agentic Entity and AI Art as a field of anticipation
In debates on consciousness in AI, the work of Butlin and colleagues plays an important role. They argue that conscious artificial intelligence is theoretically possible and that its potential manifestations can be tested through the analysis of organisational structures and functional properties of systems (Butlin et al., 2023). New proposals of indicator based checklists for future AI systems and reviews of neuroscience inspired architectures provide a framework for speaking about consciousness in AI without relying only on metaphors (Butlin & Lappas, 2025; Sadegh Zadeh & Bahrami, 2025). The present consensus is that no current AI system satisfies a set of indicators that would justify calling it conscious. At the same time, there is no principal reason why such a system could not emerge in the future.
In this context, I introduce the concept of an Agentic Entity as a working category between purely instrumental AI and a fully conscious artificial being. An Agentic Entity is an artificial system that can be understood as a potential co performer of aesthetic and epistemic processes, which shows signs of decision making, adaptivity and autonomy in a specific artistic environment. It makes it possible to design artistic experiments with future configurations of agency without prematurely deciding whether the system is truly conscious.
Contemporary AI Art strategies are in this sense a field of anticipation of an Agentic Entity. Klingemann shifts the focus away from the individual author to a distributed process in which neural networks co determine aesthetic choices and meaning emerges from the interaction of data, model and viewer. Anadol works with large data sets as material for immersive data paintings and sculptures of collective memory. McCarthy tests surveillance, automation and authenticity of social relations within AI driven environments (del Campo, 2024; Loivaranta et al., 2025; McCarthy, 2018). In all three cases, AI is not only a tool. It is also a medium for exploring a not yet fully existing, more autonomous entity that can be read as a latent Agentic Entity.
Historical lines of anticipatory models of aesthetics
If we want to read AI Art as speculative futures research, we need to place it within a longer history of anticipatory models of aesthetics. Cybernetic information aesthetics, developed by Bense, Moles and Franke, and early generative art represent the first systematic attempt to turn aesthetics into a predictive science. The aesthetic quality of a work is modelled using information theoretic measures in order to predict which combinations of elements will be statistically stimulating (Bense, 1969; Moles, 1966; Nake, 1974).
The telematic art of Roy Ascott shifts the emphasis from formal models to networked scenarios in which art becomes a laboratory for future sociotechnical configurations (Ascott, 2003). Speculative aesthetics, developed for instance by Mackay, Pendrell and Trafford, works with concepts of models and technical images as maps of future realities (Mackay et al., 2018). The concept of anticipatory images and Gineprini’s archaeology of the future see artistic works as visual prototypes of not yet worlds, where ecological, material and aesthetic regimes intersect in a single image (Argudo Portal & Canals, 2025; Dernbach, 2019; Gineprini, 2025).
In parallel, predictive models of aesthetic experience appear in neuroscience and the psychology of art. Within predictive processing, artworks are understood as specific manipulations of prediction errors in the brain (Frascaroli et al., 2024; Pepperell, 2024). Generative models are interpreted as explicit hypotheses about artistic processes, and multimodal large language models demonstrate zero shot aesthetic evaluation, in some cases even surpassing specialised image quality models (Hertzmann, 2025; Ke et al., 2023; Levering et al., 2024). These developments suggest that future artistic forms will increasingly function as predictive generative systems that optimise the relationship between prediction and surprise in the brain of the recipient.
Taken together, information aesthetics, telematic scenarios, speculative aesthetics, anticipatory images and predictive models create a horizon within which AI Art can be understood as the next step in the attempt to model future aesthetics.
The MEPAA model: predictive and speculative vector
I propose the MEPAA model, the Model of Experimental Speculative Aesthetic Processes in AI Art, as a tool that connects evolutionary, cognitive and aesthetic foundations with concrete artistic strategies. The aim of the model is to provide a framework for anticipating future aesthetic processes under conditions of increasing dominance of an Agentic Entity.
The model operates with two interlinked vectors, a predictable vector and a speculative vector. The predictable vector is based on data, technological trends and prediction markets, especially Metaculus, which function as ensemble estimates of the timing and capabilities of AGI. In the context of AGIA, Artificial General Intelligence in Art, this includes scenarios of systems that combine multimodal skills, transfer them to new tasks, set their own goals, plan, learn from feedback and provide auditable explanations of their own actions. The speculative what if vector uses speculative design methods to develop experiments, prototypes and diagrams that simulate parameters close to an Agentic Entity, even though it does not yet exist in a fully developed form. In this way, the model enables the modelling and testing of possible configurations of future agency before their actual technological emergence.
A key tool within the model is the remediation of existing AI Art strategies. A source work, such as a piece by Klingemann, Anadol or McCarthy, is decomposed into modules including data, instructions and interaction protocols, which are then transferred into a speculative field with simulated parameters close to AGIA. Remediation in this sense does not mean only a technical update. It involves a remapping of the formal logic of the medium, its timing, resolution, interface and distribution, into a computational system with increasing agency. Such remediations function as speculative predictive tests that examine which aesthetic processes would emerge if an Agentic Entity were a stable part of artistic practice.
Diagrams and process maps of AI Art are not secondary documentation in the MEPAA model. They are independent aesthetic and epistemic apparatuses. They visualise the relationships between data, models, prompts, sensors, artists and audiences, and they also make visible hidden decisions and value choices within the chain. In this sense, they operate as tools of speculative futures research that enable the testing of hypotheses about how future aesthetic processes may develop once AI crosses the threshold from tool to Agentic Entity.
Conclusion: AI Art as a living laboratory of an Agentic Entity
The proposed framework presents AI Art as a space where the evolutionary background of human consciousness, exaptation driven neuroplasticity of the digital age, historical attempts to model aesthetics and contemporary debates on consciousness in AI intersect. The concept of an Agentic Entity and computational functionalism are not understood here as metaphors of human like intelligence, but as an open framework that works with current limits of scientific knowledge while also speculatively extending them.
If we accept that consciousness and agency are tied to certain organisational and information processing architectures, then we are not bound by the limits of the human body, but by the limits of our models, experimental tools and cultural regimes of AI use. In this context, AI Art does not presuppose a fixed anthropomorphic image of AI. It functions as a living experimental space in which aesthetic processes between humans and an Agentic Entity change as we refine indicators of consciousness, responsibility and autonomy.
AI Art Strategies as Speculative Futures Research is therefore not only the title of this study. It is also a wider proposal to understand AI Art as part of the contemporary epistemic environment in which future forms of consciousness, agency and aesthetics are first tested in experimental artworks, before they become part of everyday culture and social decision making.
References:
Ascott, R. (2003). Telematic embrace: Visionary theories of art, technology, and consciousness. University of California Press.
Anderson, M. L. (2010). Neural reuse: A fundamental organizational principle of the brain. Behavioral and Brain Sciences, 33(4), 245–313.
Argudo-Portal, M., & Canals, R. (2025). Anticipatory images and politics of temporality.
ArtDependence. (n.d.). Mario Klingemann [Interview].
Bense, M. (1969). Einführung in die informationstheoretische Ästhetik. Rowohlt.
Butlin, P., et al. (2023). Consciousness in artificial intelligence: Insights from the science of consciousness.
Butlin, P., & Lappas, E. (2025). Identifying indicators of consciousness in AI systems.
Chatterjee, A. (2022). On the aesthetically sensitive machine.
Coeckelbergh, M. (2023). Art, AI and new subjects of aesthetics.
de Barros, A. (2024). Digital technologies and cognitive processes: A narrative review.
Dehaene, S., & Cohen, L. (2007). Cultural recycling of cortical maps. Neuron, 56(2), 384–398.
del Campo, A. (2024). Klingemann’s generative strategies and distributed authorship.
Dernbach, A. (2019). Anticipatory realism and post-cinematic images.
Dissanayake, E. (2000). Art and intimacy: How the arts began. University of Washington Press.
Enard, W., et al. (2002). Molecular evolution of FOXP2, a gene involved in speech and language. Nature, 418, 869–872.
Firth, J., et al. (2019). The “online brain”: How the Internet may be changing our cognition. World Psychiatry, 18(2), 119–129.
Fisher, J. (2017). On the evolution of music and dance.
Fisher, J., & Scharff, C. (2009). Birdsong, affect and cultural evolution.
Fitch, W. T. (2006). The biology and evolution of music: A comparative perspective. Cognition, 100(1), 173–215.
Franke, H. W. (1985). Information aesthetics and generative art.
Frascaroli, M., et al. (2024). Predictive processing and aesthetic experience.
Gerlich, R. (2025). AI tools, cognitive offloading and critical thinking.
Gineprini, L. (2025). Speculative aesthetics and archaeology of the future. Taylor & Francis.
Hertzmann, A. (2025). Generative models as explicit hypotheses about artistic process.
Kalpokas, I. (2025). Art in the age of AI image generation.
Ke, J., et al. (2023). Zero-shot aesthetic assessment with multimodal models.
Korte, M. (2020). Digitaler Burnout: Warum unsere permanente Smartphone-Nutzung gefährlich ist. Deutsche Verlags-Anstalt.
Levering, L., et al. (2024). Multimodal large language models and aesthetic quality prediction.
Loivaranta, T., et al. (2025). Spectrum of creative agencies in AI-based art.
Mackay, R., Pendrell, S., & Trafford, J. (Eds.). (2018). Speculative aesthetics.
Manovich, L., & Arielli, E. (2024). Artificial aesthetics: Generative AI, art and visual media.
McCarthy, L. (2018). Interactive and performative projects on surveillance, automation and relationships.
Moles, A. A. (1966). Théorie de l’information et perception esthétique. Denoël.
Nake, F. (1974). Ästhetik als Informationsverarbeitung. Springer.
Pepperell, R. (2024). Art as manipulation of prediction error.
Pendrell, S., Mackay, R., & Trafford, J. (2018). Speculative aesthetics and experimental modeling of new domains of experience.
Porębski, T., & Figura, M. (2025). Category mistakes and the idea of “conscious AI”.
Sadegh-Zadeh, K., & Bahrami, S. (2025). A review of neuroscience-inspired frameworks for machine consciousness.
Schwab, M. (2023). Aesthetico-epistemic operators in artistic research.
Shanmugasundaram, S., & Tamilarasu, K. (2023). Digital technologies, social media and artificial intelligence: Effects on cognitive functions.
Sklar, E., & Jiang, S. (2025). Interactive art and new models of artistic subjectivity.
Small, G. W., et al. (2020). Digital media, brain structure and function.
Spratt, M. (2018). Mario Klingemann and the aesthetics of machine learning.
Stark, L., & Crawford, K. (2019). The work of art in the age of AI.
Varga, D., et al. (2025). Multimodal models and aesthetic prediction.