TOWARD A SPECULATIVE PREDICTIVE FRAMEWORK FORAESTHETIC PROCESSES IN AI ART
The paper presents the proposed MEPAA model (Model of Experimental-Speculative Aesthetic Processes in AI Art), designed for the analysis and simulation of future forms of artificial creativity. It builds on an interdisciplinary integration of artistic research, computational neuroscience, analytical and experimental aesthetics, and philosophy of mind. AI Art is approached as an intermediary and speculative field in which artistic strategies are used to test concepts of agency, autonomy, and consciousness in AI systems. Within this framework, MEPAA functions as a tool for anticipatory modeling of future trajectories of artistic agency, whose validation requires further experimental research.
Abstract
The paper presents the proposed MEPAA model (Model of Experimental-Speculative Aesthetic Processes in AI Art), designed for the analysis and simulation of future forms of artificial creativity. It builds on an interdisciplinary integration of artistic research, computational neuroscience, analytical and experimental aesthetics, and philosophy of mind. AI Art is approached here as an intermediary and speculative field, where artistic strategies are used to test concepts of agency, autonomy, and consciousness in AI systems. MEPAA functions as a tool for anticipatory modeling of future trajectories of artistic agency, the validation of which requires further experimental research.
Introduction
This paper outlines a theoretical perspective for identifying the parameters of the MEPAA model (Model of Experimental-Speculative Aesthetic Processes in AI Art), developed within the research project Aesthetics of AI Art. The project aims to analyze and model aesthetic processes in AI Art that assume the presence or potential emergence of autonomous, consciousness-like operations within artificial intelligence systems. Research is grounded in an interdisciplinary synthesis of experimental aesthetics, computational cognitive neuroscience, and computer science with the objective of understanding and simulating new modalities of creativity.
The theoretical framework is based on the concept of artificial intelligence as a potential Agentic Entity within artistic and aesthetic processes. This assumption is grounded in the premise that AI systems, particularly those capable of initiating or influencing decision-making operations, are relevant to a speculative-predictive model of aesthetic processes in AI Art. Within this model, AI is no longer considered solely as a tool but instead functions as an operative actor sharing aspects of creative agency, necessitating new ontological and epistemological approaches. The principal aim is to provide a comprehensive analysis of aesthetic processes in AI Art and explore the viability of its future forms, in which the interaction between human creators and AI systems exhibiting properties of consciousness or agency plays a central role.
The paper introduces theoretical concepts that define AI Art as a specific developmental phase of digital art, characterized by the latency of the Agentic Entity. This concept informs the design of research strategies in AI Art that possess anticipatory potential, offering a speculative view of future aesthetic processes. The Agentic Entity is here understood as a system capable of initiating, modifying, or directing causal effects within its environment, based on a degree of autonomy, intentionality, or decision-making functionality. In the context of artificial intelligence, this can include systems that exhibit decision-making capacity, adaptive behavior, and the ability to engage in interactions with human actors in ways that affect the outcome of the artistic process [1, 2].
This principle of agency is further linked to the concept of distributed agency, which does not locate the agency in a single subject but distributes it across heterogeneous actors, systems, technologies, and environments interacting within dynamic networks. This conception rejects traditional binary distinctions between an active subject and a passive object. These ideas are framed within a predictive model of aesthetic processes in AI Art defined by two vectors: a predictable trajectory grounded in technological development and a speculative trajectory that imagines aesthetic processes dominated by the Agentic Entity. To propose and test this model, methods from speculative design are applied.
The MEPAA model offers a framework that integrates artistic research, analytical and experimental aesthetics, computational neuroscience, and philosophy of mind within a single transdisciplinary perspective. Within this framework, AI Art functions as an intermediary field that not only explores the aesthetic operationalization of technologies, but also critically reflects on the boundaries and possibilities of knowledge itself.
From the proposed theoretical concepts and analyses, it follows that AI Art is not limited to the application of generative technologies. Rather, it constitutes a speculative domain for probing new forms of artificial creativity, subjectivity, and awareness. Artistic research strategies thus operate as epistemic instruments, transforming our understanding of aesthetics as an open, intersubjective process. The MEPAA model can therefore be seen as a flexible structure for predictive modeling of future aesthetic agency in AI systems, while also requiring further experimental validation through applied research and critical artistic practice.
Research Methods
The methodological framework of this research is structured around two primary components. The first is focused on defining procedures for addressing the theoretical and analytical position embedded in the discourse of AI Art aesthetics and the study of contemporary artistic practices. This position investigates artistic research strategies that transpose scientific methods into artistic inquiry and contributes to models of interaction between human creators and AI systems that contain features of agentic decision-making.
The second component constitutes a critical and verification-oriented perspective, aimed at testing theoretical assumptions through speculative research strategies. These strategies help construct a model of the future of AI Art, wherein the relationship between human and AI systems possessing elements of consciousness or autonomous agency plays a central role.
This methodological position is further divided into two key layers. First, discursive-analytical, employs tools of critical discourse analysis to examine how concepts such as agency, consciousness, autonomy, artistic strategies, or AGI are articulated across disciplinary and artistic contexts. The goal is to trace emerging semantic constellations and map the dynamic transformations of these terms in an expanded interdisciplinary field. The second, comparative conceptual analysis, focuses on critically evaluating divergent models of creativity, subjectivity, and aesthetic value within the interaction between human and machine intelligence. This is achieved through a transversal analytical approach, identifying shifts in aesthetic thinking and artistic practice considering AI’s emergence as a partially creative agent.
The theoretical-analytical methodology draws on qualitative research methods and content analysis applied to aesthetics of AI Art, current artistic practices, cognitive science, computer science, and relevant philosophical and cosmological perspectives. The objective is to identify dominant or potentially significant research components within an interdisciplinary matrix, with particular attention to forms of partnership between the human subject and AI systems exhibiting features of agentic decision-making.
A comprehensive literature review was conducted and organized into thematic clusters. Each cluster focuses on a distinct aspect of the examined phenomena, enabling a more targeted analysis of relevant theories and their interrelations. These include:
- recent AI Art production, with an emphasis on documented artistic projects, curatorial texts, and critical studies
- philosophical-aesthetic frameworks that define AI Art as a new phase in the evolution of digital art
- artistic research as a process of epistemic operation and conceptual strategy formation
- futurological and cosmological perspectives on AGI development
- the problem of consciousness in artificial systems and the interconnection between computer science, cognitive science, and computational neuroscience
- methodological approaches to forecasting AGI development through speculative design and prediction markets
The first part of this mapping focused on speculative design tools for conceptualizing possible future scenarios across artistic and scientific domains. The second examined theoretical and practical infrastructures of predictive platforms such as Metaculus and Polymarket, which generate collective forecasts about AGI through aggregated expert and non-expert predictions.
Based on these theoretical analyses and interdisciplinary comparative insights, three core conceptual pillars were formulated to guide the MEPAA model (Model of Experimental-Speculative Aesthetic Processes in AI Art). These include: (1) AI Art as a specific developmental phase marked by the latency of the Agentic Entity, (2) artistic research strategies with anticipatory potential, and (3) a speculative-predictive model of aesthetic processes. Together, these pillars provide a theoretical and methodological scaffold for designing, modeling, and testing possible trajectories of AI Art in relation to the emergence of artificial agency. In this sense, methodology is not solely a tool for analyzing current conditions but also a framework for formulating and testing conceptual scenarios through speculative design as a research strategy.
Concept I: AI Art as a Distinct Developmental Stage of Digital Art Characterized by the Latency of an Agentic Entity
AI Art can be understood as a specific developmental stage in the evolution of digital art, distinguished by artistic strategies that emerge from a convergence of technical, aesthetic, and epistemic processes. Within this framework, artificial intelligence is not merely employed as a tool but is increasingly acknowledged as a partial, and potentially future, Agentic Entity exhibiting decision-making, adaptivity, and autonomy. The core thesis of this concept is grounded in a qualitative analysis of artistic strategies in AI Art, focusing on their relationship to the concept of the Agentic Entity as a potential creative agent.
The foundational hypothesis asserts that AI Art constitutes a distinct phase in digital art evolution that displays both shared and unique features and that the latent agency of artificial intelligence significantly shapes the nature of artistic strategies. This hypothesis is constructed through an analytical synthesis of theoretical work in aesthetics, computer science, and cognitive neuroscience. As justification, this concept engages in a critical discourse with key theoretical frameworks defining AI Art, particularly through the contributions of Lev Manovich, Emmanuel Arielli, and Joanna Zylinska.
Manovich’s typology of AI Art emphasizes a shift from imitation to innovation, proposing that the future of AI Art lies in the ability of artificial intelligence to generate new aesthetic structures that transcend traditional human patterns. Both Arielli and Manovich highlight the current limitations of generative models to autonomously initiate artistic movements, underscoring their continued role as instruments of human authorship. Zylinska’s concept of a perception machine extends this discourse, offering a critical reflection on the perceptual, social, and epistemic consequences of machine vision within contemporary media environments.
Following this, two complementary dimensions of AI Art are proposed. The first encompasses the creative deployment of AI systems to expand the expressive vocabulary of AI Art. The second focuses on the speculative potential of the Agentic Entity as a source of novel aesthetic phenomena. This conceptualization is supported by a content analysis of artistic strategies employed by Mario Klingemann, Refik Anadol, Lauren Lee McCarthy, Pierre Huyghe, Sofia Crespo, and Michael Sedbon. Their practices share a commitment to partially transferring aesthetic decision-making to AI systems. These approaches anticipate attributes of artificial agency and consciousness, thereby shaping the cultural imaginary of Artificial General Intelligence even before its technical realization.
In Defining AI Arts (2019), Manovich outlines three approaches to defining AI Art, each reflecting different epistemic and aesthetic dimensions [3]. The first two are centered on art production using AI tools that mimic existing styles. The third defines AI Art as a potential instrument for surpassing traditional human aesthetic patterns. Here, Manovich suggests that the future of AI Art lies not in simulating past styles but in enabling artificial intelligence to generate radically new visual and sonic structures. This requires a transition from imitation to innovation, where AI redefines the boundaries of creativity by producing works that are not explicitly learned but emerge as unprecedented aesthetic phenomena.
Lev Manovich and Emmanuel Arielli, in their study Artificial Aesthetics (2024), offer an analytical perspective on current trends in AI Art by examining how generative models transform the process of artistic creation [4]. They emphasize critical issues such as authenticity, originality, and the definition of creativity within the context of algorithmically generated works. Their analysis indicates that contemporary forms of artificial intelligence do not possess the capacity to autonomously initiate or redefine new artistic movements. Instead, AI currently functions as a tool used by human creators to expand their expressive and aesthetic repertoire.
This perspective is complemented by Eugen Nikitin’s study Characteristics of AI Art in the Context of Lev Manovich’s Ideas (2024), which revisits and adapts Manovich’s principles of new media in the framework of big data and cultural analytics [5]. These domains are interpreted as latent environments that enable the emergence of creative processes involving artificial intelligence systems. Key conceptual foundations such as transcoding, modularity, and variability are presented as reference points for AI Art. Transcoding is understood as the translation of content into different cultural and technical formats and serves as the basis for what Manovich terms “deep remixability,” introduced in After Effects or Velvet Revolution (2006) [6]. Deep remixability denotes the ability to manipulate stylistic, historical, medial, and generic elements in a distinct compositional way.
In his 2024 writing, Manovich reinforces his argument that a specific aesthetic of artificial intelligence does not currently exist [7]. This view is extended to the discussion of generative AI where he expresses skepticism regarding the ability of contemporary artists to create exceptional works solely using AI. Rather, he identifies the potential of AI Art in the practices of cultural industry professionals such as illustrators, designers, and photographers who already employ AI tools to produce visually innovative works. In his view, it is the presence of technical skill that enables the generation of aesthetically relevant outcomes, not the existence of generative models alone.
The discourse on AI Art is shaped not only by technological innovation, but also by critical reflections on the cultural, perceptual, and epistemological shifts provoked by these technologies. One of the key figures in this discourse is Joanna Zylinska, who began outlining its contours in her publications AI Art: Machine Visions and Warped Dreams (2020) and The Future of Media (2022). In her more recent work, The Perception Machine: Our Photographic Future Between the Eye and AI (2023), she explores the transformation of photography and perception in the age of artificial intelligence (Zylinska 2020; 2023). Zylinska’s approach merges media theory with neuroscience, positioning the concept of perception machine as both a metaphor for visual and cognitive closure and a horizon of sociopolitical and affective opening.
At the core of her analysis lies the question of existence within a landscape shaped by image flows and machine vision. She critically engages with Mark Zuckerberg’s vision of the metaverse, identifying it as a further evolution of the perception machine with implications for the future of human perception and interaction. Zylinska calls on designers, programmers, artists, philosophers, and activists to engage in dialogue and experimentation aimed at “hacking” the perception machine before it fully closes into algorithmic determinism. This intervention is framed as a strategy for revealing its infra-potential and opening new conditions for perception and artistic creation.
The theoretical frameworks developed by Lev Manovich and Joanna Zylinska provide key foundations for formulating the hypothesis that the core definition of AI Art unfolds across two complementary dimensions. The first dimension concerns the application and creative use of artificial intelligence systems that significantly expand the operational scope of artistic practice, not only in terms of representation but also in terms of generativity, variability, and the reconfiguration of aesthetic structures. The second dimension is characterized by the robustness of this potential, which requires the exploration of its operative capacities for moving from imitation to innovation. Within this framework, it is assumed that AI systems may transcend the boundaries of traditionally understood artistic production and generate outputs that are not explicitly learned but rather emerge as new aesthetic phenomena that operate independently of historical models of creation.
These preliminary conclusions suggest that artistic strategies that may represent a fundamental divergence from previous stages in the evolution of art must at minimum anticipate the potential of the Agentic Entity if we truly want from computers what humans are incapable of achieving [3]. An analysis of the current state of scientific knowledge indicates that although there is currently no experimentally verified or generally accepted procedure for realizing AGI, it is possible to identify key concepts and computational theories that provide frameworks and indicators for modeling its essential properties.
The broader context provided by studies such as Consciousness in Artificial Intelligence [8] and Principles for Responsible AI Consciousness Research [9] explores the conditions under which elements of artificial consciousness could be implemented in intelligent systems. According to this analysis, current artificial intelligence systems do not demonstrate consciousness in its full form. Although certain aspects of consciousness may be simulated or modeled, the complexity and depth of actual consciousness, as it is understood in the context of human cognition, remain beyond the capabilities of present AI technologies. At the same time, the findings suggest that there are no fundamental technical barriers preventing the development of advanced artificial intelligence systems that might fulfill these indicators and thus potentially exhibit attributes of consciousness.
Given the cultural paradigm in which visions and mental models of technologies often precede their technical feasibility, it can be assumed that artistic production in the field of new media and specifically in AI Art anticipates certain outlines, forms, and intentions related to the concept of AGI. The presence of decision making and autonomy in such works does not merely function as an aesthetic feature, but it operates as a cultural anticipation of general artificial intelligence, shaping its imaginary and potential significance before its actual technological realization [10].
Through content analysis supported by an extensive literature review, this study aimed to provide a relevant overview of current AI Art production and to identify artistic strategies that integrate or emerge from a combination of technical, aesthetic, and epistemic processes to employ or anticipate the model of the Agentic Entity. The Agentic Entity refers to a subject or system capable of initiating, modifying, or directing causal effects in its environment based on a certain degree of autonomy, intentionality, or functional components of decision making. In the context of artificial intelligence, an Agentic Entity may represent a system that displays signs of decision-making capacity, adaptive behavior, and the ability to engage with other entities including human actors in ways that influence the resulting artistic process.
The principle of the Agentic Entity is also closely related to the concept of distributed agency, which holds that agency is not concentrated within a single subject but rather distributed across multiple actors, technologies, systems, and environments interacting within specific networks or processes. This concept challenges the traditional binary distinction between an active subject and a passive object [1].
The research survey includes sources documenting artistic projects such as catalogs, curatorial texts, artist statements, and scholarly studies, selected based on the assumption of a positive identification within a time frame spanning from the first projects utilizing machine learning techniques to the present. Based on these criteria, it was possible to positively identify artists who deliberately apply approaches oriented towards the use of potential properties of the Agentic Entity. Their work tends toward experiments in which a certain degree of aesthetic decision making is delegated to the artificial intelligence system itself. Within this group, explicit interest can also be identified in the issues of artificial consciousness and epistemic shifts achieved through artistic research. Representative figures in this group include Mario Klingemann, Refik Anadol, Lauren Lee McCarthy, Sofia Crespo, Pierre Huyghe, and Michael Sedbon.
Concept II: Artistic Research Strategies in AI Art with Anticipatory Potential
Artistic research strategies in AI Art that exhibit anticipatory potential offer insights into future aesthetic processes. The artwork and its corresponding artistic research strategy may be understood as an epistemic vector. This thesis is grounded in the assumption that artistic research strategies with anticipatory potential enable the articulation of conceptual models of future forms of art, culture, and artificial intelligence even before their technological realization.
The methodological approach is based on a qualitative analysis of artistic research practices of selected authors who represent differentiated approaches to artificial intelligence as a creative agent, extracted from a group of artists whose work demonstrates an explicit interest in the question of artificial consciousness and in epistemic shifts enabled by artistic research. To more precisely define the notion of artistic research strategies, a parallel analytical evaluation of the relevant literature was performed. The key identification features of artistic research strategies were defined as components of epistemic thinking, primarily in the form of research operations with an anticipated epistemic objective. The artists included in this group are Mario Klingemann, Refik Anadol, Lauren Lee McCarthy, and Sofia Crespo. One of the main criteria for their selection was the presence of anticipatory potential, understood as the capacity for imaginative foresight.
The analysis of the literature review revealed a definitional ambiguity that is characteristic of the field of artistic research, shaped by the continuous interweaving of artistic and scientific methodologies. This claim is primarily supported by the analysis of studies such as The Routledge Companion to Research in the Arts [11] and Split and Splice: A Phenomenology of Experimentation [12]. Due to the need for greater definitional clarity, especially in relation to the formulated research hypotheses, an interpretative line based on the work of Michael Schwab was adopted. In his studies Contemporary Research [13] and Transpositionality and Artistic Research [14] Schwab identifies two fundamental epistemic shifts: a move away from traditional models of knowledge toward more flexible research paradigms and a transition from dominant forms of artistic practice to experimental and less formalized modes of creation. At the same time, he proposes to understand artistic research as a process of transposing scientific methodological approaches into artistic production. On this basis a working framework was proposed for the present research that defines artistic research strategies as processes of transposition of scientific procedures into the context of artistic creation.
The artistic strategies of Mario Klingemann between 2018 and 2024 are based on creative procedures in which one can identify methodological principles and epistemological implications within the dynamic relationship between human and artificial intelligence. His work articulates the concept of distributed agency where the act of creation is not solely attributed to the human subject, but arises as a synergy between human and non-human actors. Klingemann’s artworks reflect a shift from the traditional notion of authorship gesture to algorithmic co-authorship operating with neural networks as active agents in aesthetic decision making. These strategies also thematize the emergence of meaning as a process shaped through the interaction between data structure, output variability, and the interpretive framework of the viewer, as exemplified in works such as Circuit Training (2020) and Uncanny Mirror (2020). In his practice, the artwork becomes a dynamic environment in which the form self-regulates, opening new perspectives on aesthetic autonomy. In doing so, Klingemann contributes to the redefinition of artistic creativity as a distributed, iterative, and non-linear process [15–17].
Refik Anadol is one of the pioneers of AI Art known for operations with data representations ranging from everyday phenomena to inquiries involving nature and neuroscience. His work combines generative algorithms with large-scale datasets that transform latent computational processes into visual projections. Anadol participates in machine learning and neural networks to analyze large volumes of data, from meteorological archives to brain scans. His installations often inhabit a liminal space between the digital and the physical, relying on the concepts of “data painting” and “data sculpture.” Projects such as Machine Hallucinations, Melting Memories, and Unsupervised explore collective memory, cognitive processes, and the potential of artificial intelligence in creative domains. Although still grounded in specialized systems, his practice synthesizes heterogeneous data, visualizes it, and invests it with meaning in ways that resonate with perception, learning, and interpretive cognition, suggesting an anticipatory premise of AGI [18–20].
Lauren McCarthy is known for her work on interactive technologies, performative art, and artificial intelligence. Her artistic research strategies focus on the relationship between humans and technology, often engaging in issues of surveillance, automation, and social interaction in the digital age. A central aspect of her projects lies in exploring the intersection of artificial intelligence and human interaction while investigating how AI influences our social relationships and personal decision-making. McCarthy designs systems that mimic or replace human social behavior, testing the boundaries between technology and authentic human experience.
Contrary to mainstream technological optimism, McCarthy examines the problematic dimensions of technological advancement, particularly in terms of artificial surveillance, social isolation, and the transformation of interpersonal relationships. She is not a passive observer of technological change, but an active participant in reshaping it through art. Her strategies integrate performative practices, software engineering, and a strong commitment to philosophical inquiry, offering a critical lens through which to consider the social and artistic trajectory of artificial intelligence.
Although McCarthy does not engage in the technical development of AGI, her artistic research strategies address fundamental questions essential for understanding AGI in social and philosophical terms, thereby constituting an anticipatory approach. One of her core operational principles is the simulation of AGI through systems that interpret user input and adapt in real time or by using performance as a mode of learning. Representative projects include Unlearning Language (2021), in which participants are placed in situations that require them to develop a new language that diverges from the conventions of AI and the patterns of natural language [21]. Another key work is Voice In My Head (2023), which simulates the implantation of an external voice that acts as a guide in navigating everyday life scenarios [22].
Sofia Crespo, in her artistic research strategies, employs artificial intelligence systems to simulate and reconstruct biological forms, suggesting the possible integration of key elements of AGI into artistic production, such as the capacity for autonomous creation and innovation. Her strategies operate with generative models that not only reconstruct existing biological structures, but also synthesize entirely new species, thus approaching the concept of emergent AGI creativity. The use of deep neural networks and evolutionary algorithms implies that systems inspired by such methods could autonomously design innovative biological structures.
In the context of AGI, it is anticipated that systems capable of autonomous learning and adaptation will possess biomimicry mechanisms. Crespo’s work indicates that AI can participate in bioinspired processes not only for aesthetic purposes but also for functional purposes, which can prove essential for AGI capable of applying knowledge in multiple domains. She creates environments in which AI plays an active role in the experimental interpretation of reality. The emergence of ecosystemic intelligence, in which various AI components cooperate in a dynamic balance analogous to biological ecosystems, would in an AGI context imply the capacity of AI not only to simulate, but also to actively shape and optimize complex systems across research, science, and the arts.
An analysis of artistic research strategies in AI Art reveals a strong anticipatory potential that articulates conceptual models of future forms of artificial intelligence within aesthetic and epistemic processes that are not yet technologically achievable. Artists such as Mario Klingemann, Refik Anadol, Lauren McCarthy, and Sofia Crespo employ artistic research strategies that acknowledge artificial intelligence systems as creative agents. Klingemann systematically develops a model of distributed agency and algorithmic co-authorship, Anadol focuses on the visualization of large data structures as manifestations of collective cognition, McCarthy investigates the boundaries between authenticity and simulation in technologically mediated social interaction, and Crespo explores the emergence of bioinspired creativity within the framework of ecosystemic intelligence.
A common feature across these artistic strategies is the presence of conceptual and operational characteristics of agentic capacity. Although these traits are not yet technologically realized, they are modeled, simulated, and critically examined within artistic research. This ability of artistic practice to anticipate potential trajectories in the development of artificial intelligence supports the hypothesis that these approaches may serve as epistemic vectors for AGI. In conclusion, the examined artistic research strategies in AI Art do not merely reflect current technological trends, but actively participate in constructing epistemic models that reshape our understanding of autonomy, consciousness, and creativity within digital culture.
Concept III: Speculative Predictive Model of Aesthetic Processes in AI Art
The parameters of the predictive model of aesthetic processes in AI Art can be defined through a predictable vector of development, derived primarily from technological analysis, and a speculative vector, which envisions aesthetic processes dominated by an Agentic Entity. This thesis assumes that the parameters of a predictive model of aesthetic processes in AI Art can be established through two complementary vectors: the technologically predictable and the speculative. The former is based on analyses of the evolution of generative models and artificial intelligence within the domain of art, while the latter models future scenarios in which AI operates as an Agentic Entity characterized by autonomy, decision-making, and adaptivity. The objective of the research, which relies predominantly on qualitative and discursive analysis, is to identify how these vectors influence the nature of artistic strategies and construct a framework for aesthetic processes that transcend current technological capacities toward the applicability of an Agentic Entity.
Aesthetic processes can generally be defined as dynamic cognitive, perceptual, and emotional operations that occur during the creation, reception, and interpretation of artistic or intentional content. These processes constitute a multilayered system of interaction between subject and object, integrating sensory input, culturally conditioned interpretations, and individual affective responses. It can also be posited that the aesthetic process operates as a synchronization of multiple levels of consciousness, a topic of contemporary interest in analytic aesthetics and in several models within theories of consciousness [23].
A significant interpretive framework in this context is the Global Workspace Theory, which posits that conscious information processing emerges through the coordination of multiple specialized cognitive subsystems [24]. In the context of neuroaesthetics, this theory also provides a perspective on how aesthetic experience is constituted through interactions between affective and cognitive modules within conscious experience [25]. Referring to current scientific understanding, one of the most plausible frameworks for the realization of an Agentic Entity is computational functionalism, within which Global Workspace Theory occupies an important position [26, 9].
The formation of an understanding of aesthetic processes in which an Agentic Entity plays a dominant role is determined by the possible function of this entity and its application within artistic strategies as well as its relationship to the recipient. Thinking about possible aesthetic processes in AI Art therefore involves not only the analysis of its operational capabilities but also the exploration of the interactive dimension in which the recipient perceives agency as an integral part of the aesthetic experience.
The analysis of the development of predictability for a minimal conceptual model of Artificial General Intelligence is based on the assumption that the most reliable forecasting apparatus for anticipating key features of general artificial intelligence such as self-awareness, metacognition, the ability to interpret language, problem solving in unknown contexts, and physical interaction is likely represented by the Metaculus forecasting platform. This platform, which is based on the principle of collective intelligence, aggregates the conclusions of expert forecasts and currently constitutes one of the most comprehensive approaches to predictive modeling of AGI development. When combined with knowledge from other scientific domains, this method can be understood as a predictable vector of development, though it remains primarily focused on technical parameters.
The analysis of artistic research strategies in the domain of AI Art reveals that the studied approaches demonstrate significant anticipatory potential, which articulates conceptual models of future forms of artificial intelligence across different levels within aesthetic and epistemic processes. These models frequently transcend current technological limitations and propose speculative visions such as distributed agency, algorithmic co-authorship, or the emergence of ecosystemic intelligence. The decomposition of these strategies along with their remediation at the level of functions and structural components creates a conceptual foundation for speculative operations [27].
Manovich’s theoretical approaches to AI Art reveal fundamental epistemic and aesthetic tensions that shape the current discourse in the field. On the one hand, artificial intelligence expands the spectrum of creative possibilities. However, it still serves primarily as a tool for the reproduction of existing styles. Overcoming this conservative tendency and reaching a state in which AI can generate radically new aesthetic paradigms remains a key challenge. This also implies that the domain of art does not constitute an entirely distinct space. Consequently, identifying parameters within the interactional dimension of a model in which the recipient perceives agency as part of the aesthetic experience continues to be supported by the conceptual foundations of Actor-Network Theory. This theory significantly contributes to the understanding of the role of technologies and non-human actors in creative networks and serves as an appropriate extension that redefines the concept of agency to include the active participation of technology in creative processes [1, 28, 29].
The reciprocal relationship between predictive and speculative components enables the articulation of a complex framework in which artistic strategies within AI Art navigate between current technological capabilities and potential future scenarios. This framework also opens space for theoretical engagement with the Agentic Entity as part of both the aesthetic process and the subject of artistic creation.
The proposition that the parameters of a predictive model of aesthetic processes in AI Art can be defined through a foreseeable vector of development and its speculative extension is grounded in a dual structure that juxtaposes technological reality with conceptual outlook. On the one hand, technological analyses based on the development of generative models, computational capacity, and advances in cognitive architecture allow the identification of stable trajectories for the aesthetic operationalization of artificial intelligence. On the other hand, the speculative dimension extends these trajectories through the anticipation of forms of the Agentic Entity, which plays an active, creative, and potentially autonomous role in artistic processes.
The search for, development of, and validation of possible aesthetic processes in AI Art involving the interaction between human and Agentic Entity depends not only on scientific forecasting procedures but also on human imagination and speculative potential. Thus, the central research questions related to this concept are oriented around the problem of how to design an experimental model of possible aesthetic processes in AI Art that is also testable for verifying hypotheses proposed within the intentions of existing theoretical frameworks.
As a proposed method for both design and validation, speculative design (MSD) has emerged as a viable strategy [30, 31]. MSD facilitates the integration of speculative operations with scenario modeling, offering a framework for thought experiments aimed at exploring future aesthetic trajectories in the context of artificial general intelligence [32]. These speculative visions need not be strictly constrained by the technological feasibility of AGI but may serve as tools for conceptually mapping the distance between the present moment and possible frameworks for artificial intelligence.
The meta-theoretical integration of individual theoretical analyses, as outlined above, offers the possibility of constructing a speculative predictive framework for AI Art. This framework allows for the formulation of developmental scenarios that model preliminary interactions between humans and artificial intelligence systems. The formulation of hypotheses and conceptual models for Methodologies of Speculative Design (MSD) can thus be tested through an experimental methodological framework that operates as a speculative predictive research strategy. Within this strategy, possible epistemological and aesthetic consequences of the development of an Agentic Entity are explored through conceptual experimentation. Particular emphasis is placed on time-structured predictions of the development of artificial general intelligence, which serve as a framework for testing both artistic and epistemic models of artificial agencies.
The design methods of MSD can be understood as a decompositional approach to AI Art projects that demonstrate predictive potential and contain elements indicative of AGI characteristics. Such an approach enables their speculative composition and remediation through the integration of multiple inputs. These inputs include predictive data sourced from the Metaculus platform and analysis of the current state of agency and embodiment in artificial intelligence systems [33].
The resulting MEPAA model may serve as a thought experiment, allowing conceptual testing and aesthetic reflection on potential future configurations of artificial intelligence within an artistic context. This methodological framework includes the aggregation of predictive data with relevant AGI criteria such as the Turing test, the Winograd Schema Challenge, and standardized cognitive benchmarks. It also involves mapping the relationships between AGI signifiers and components of artistic research strategies in AI Art, with special attention to domains such as metacognition, linguistic interpretation, unknown problem solving, self-reflection, and physical interaction.
The proposed approach integrates qualitative discursive analysis of selected artworks as case studies along with scenario-based and visual modeling. These are supported by tools for structuring prediction models and visualizing the developmental trajectories of AI Art in relation to the evolving contours of agency. In this way, the proposed framework creates space for an extended reflection on artificial intelligence as an aesthetic, epistemic, and speculative phenomenon, with an emphasis on anticipating its future configurations within artistic practice.
This concept is not merely an analytical treatment of existing artistic research approaches, but also a proposal for and validation of methodological procedures capable of simulating future trajectories of aesthetic processes in AI Art. The approaches presented here are grounded in the interconnection of qualitative analysis, discursive conceptual reflection, and speculative design methodologies that form the structural backbone of the MEPAA model. This model functions as a dynamic framework for testing hypotheses concerning the potential configurations of artificial agency, its aesthetic operability, and its cultural epistemic significance. The MEPAA model is thus not only an analytical tool but also a thought experiment that enables the construction and validation of speculative scenarios for the future of artificial intelligence in the context of artistic creation.
Conclusions
By subsuming both the analyses and proposed conceptual frameworks, it can be concluded that AI Art does not merely represent the result of applying generative technologies to artistic production, but rather constitutes a speculative experimental field that enables the conceptual testing of emergent forms of artificial creativity, subjectivity, and consciousness. Within this framework, artistic research strategies function as epistemic vectors through which aesthetics is redefined as a dynamic, intersubjectively constituted, and ontologically open-ended process.
The proposed levels of the predictive model of aesthetic processes in AI Art do not serve solely as a descriptive framework reflecting the state of technological development but instead operate as a conceptual infrastructure that enables the articulation and testing of possible configurations of a future Agentic Entity. In this formulation, the interactive dimension between the recipient and the artificial intelligence system is transformed into an intersubjective exchange between entities with differing ontological statuses, thereby contributing to a reconceptualization of the very notion of aesthetics.
The MEPAA model (Model of Experimental-Speculative Aesthetic Processes in AI Art) thus establishes an operative space for the transdisciplinary integration of artistic research, analytical aesthetics, computational neuroscience, and philosophy of mind. In this context, AI Art operates as an intermediary field through which not only the aesthetic operationalization of technologies can be explored, but also the epistemological limits and potentialities of knowledge within the contemporary post-digital horizon.
Based on these concepts, the proposed MEPAA model demonstrates the potential to become a relevant methodological tool for anticipatory modeling of aesthetic trajectories of artificial intelligence in the context of artistic creation. However, its heuristic and applicative value requires experimental verification through specific research and creative implementations.
References
[1] Bruno Latour. Reassembling the Social: An Introduction to Actor-Network-Theory. Oxford University Press, Oxford, 2005.
[2] Karen Barad. Meeting the Universe Halfway: Quantum Physics and the Entanglement of Matter and Meaning. Duke University Press, Durham, NC, 2007.
[3] Lev Manovich. Defining AI Arts. Three Proposals, 2019.
[4] Lev Manovich and Emanuele Arielli. Artificial Aesthetics: Generative AI, Art and Visual Media. Self-published, 2024.
[5] Eugeny Nikitin. Characteristics of AI Art in the Context of Lev Manovich’s Idea, December 2024.
[6] Lev Manovich. AFTER EFFECTS, OR VELVET REVOLUTION. Artifact, 1(2):67–75, October 2007.
[7] Lev Manovich. “A summary of my recent post ‘There is no AI Aesthetics’ in a 7 slides…” LinkedIn post, May 2024.
[8] Patrick Butlin, Robert Long, Eric Elmoznino, Yoshua Bengio, Jonathan Birch, Axel Constant, George Deane, Stephen M. Fleming, Chris Frith, Xu Ji, Ryota Kanai, Colin Klein, Grace Lindsay, Matthias Michel, Liad Mudrik, Megan A. K. Peters, Eric Schwitzgebel, Jonathan Simon, and Rufin VanRullen. Consciousness in Artificial Intelligence: Insights from the Science of Consciousness. August 2023.
[9] Patrick Butlin and Theodorus Lappas. Principles for Responsible AI Consciousness Research. Journal of Artificial Intelligence Research, 82:1673–1690, March 2025.
[10] André Bazin. What is Cinema? Volume 1. University of California Press, Berkeley, 1967.
[11] Michael A. R. Biggs, Henrik Karlsson, Michael Biggs, and Stiftelsen Riksbankens Jubileumsfond. The Routledge Companion to Research in the Arts. Routledge, London, 2012.
[12] Hans-Jörg Rheinberger. Split and Splice: A Phenomenology of Experimentation. University of Chicago Press, Chicago, 2023.
[13] Michael Schwab. Research Exposition on Artistic Research and Transdisciplinarity. HUB – Journal of Research in Art, Design and Society, Special Issue 0, 2023.
[14] Michael Schwab. Transpositionality and Artistic Research. In Michael Schwab (ed.), Transpositions: Aesthetico-Epistemic Operators in Artistic Research, pp. 191–214. Leuven University Press, 2018.
[15] Matias del Campo. Art Beyond Mechanical Reproduction: In Conversation with AI Artist Mario Klingemann. Architectural Design, 94(3):64, May 2024.
[16] Jason Nelson. Midjourney’s AI Platform Blocks Images of Chinese President, Sparking Ethics Debate. Decrypt, 2023.
[17] Emily L. Spratt. Creation, Curation, and Classification: Mario Klingemann and Emily L. Spratt in Conversation. XRDS: Crossroads, The ACM Magazine for Students, 24(3):34–43, April 2018.
[18] Refik Anadol. Space in the Mind of a Machine: Immersive Narratives. Architectural Design, 92(3):28–37, May 2022.
[19] Refik Anadol. Synaesthetic Architecture: A Building Dreams. Architectural Design, 90(3):76–85, May 2020.
[20] Zach Winn. Pushing the frontiers of art and technology with generative AI. MIT CSAIL, 2023.
[21] Lauren Lee McCarthy. Unlearning Language, 2018–2021.
[22] Lauren McCarthy. Voice In My Head, 2023.
[23] Vlastimil Zuska. Estetický prožitek: Vrstvy, úrovně a fáze estetické zkušenosti. Karolinum, Praha, 2024.
[24] George A. Mashour, Pieter Roelfsema, Jean-Pierre Changeux, and Stanislas Dehaene. Conscious Processing and the Global Neuronal Workspace Hypothesis. Neuron, 105(5):776–798, March 2020.
[25] Semir Zeki, Yan Bao, and Ernst Pöppel. Neuroaesthetics: The Art, Science, and Brain Triptych. PsyCh Journal, 9(4):427–428, August 2020.
[26] Oron Shagrir. The Rise and Fall of Computational Functionalism. In Hilary Putnam, pp. 220–250. Cambridge University Press, 2005.
[27] Tomáš Marušiak. Umelecko-výskumné stratégie AIArt a vedomie. Journal of Interactive Media (JOINME), 6, 2025.
[28] Olga Amsterdamska. Surely You Are Joking, Monsieur Latour! Science, Technology, & Human Values, 15(4):495–504, 1990.
[29] Pierre Bourdieu. The Rules of Art. Genesis and Structure of the Literary Field. Stanford University Press, 1996.
[30] Anthony Dunne and Fiona Raby. Design, Fiction, and Social Dreaming. MIT Press, 2013.
[31] Jinghong Chen. The Role of AI: Speculative Design in Redefining Artistic Collaboration. Journal of Ecohumanism, 3(8), November 2024.
[32] Laura Barendregt and Nora S. Vaage. Speculative Design as Thought Experiment. She Ji: The Journal of Design, Economics, and Innovation, 7(3):374–402, September 2021.
[33] Metaculus. Metaculus: Forecasting AGI and AI Progress (online platform).