Generative AI and the future of authorship: modern and contemporary art as a method for preserving human subjectivity

Oblikovanje medijske budućnosti u digitalnom okruženju

The Digital Media Environment and the Deficit of Human Presence

By the mid-2020s, the digital media environment had entered a phase of qualitative transformation driven by the widespread adoption of generative artificial intelligence. The speed of producing texts, images, audiovisual materials, and hybrid forms of content increased dramatically, while the threshold for entering media production decreased substantially. As a result, the media landscape is increasingly defined not by a scarcity of information, but by an overabundance of rapidly produced, algorithmically predictable, and functionally interchangeable messages.

This shift is not only technological but also anthropological in nature. Whereas the value of the media producer was once largely determined by the capacity to create content, the very function of production is no longer scarce. A different question now comes to the forefront: what remains specifically human under conditions of near-infinite generation? In other words, if the production of a single unit of content becomes almost instantaneous, the scarce resource is no longer the text as such, but rather authorial subjectivity, intellectual position, interpretive depth, and the ability to construct durable structures of meaning.

Thus, the contemporary media environment requires not merely the adoption of new tools, but a reconsideration of the very model of authorship itself. The central question is no longer how to produce

more, but how to preserve interpretive and ethical subjectivity in a space where the algorithm tends toward standardization.

One of the most significant theoretical and practical challenges for analyzing the digital environment is the phenomenon of model collapse (Shumailov et al., 2023). This term refers to a situation in which generative systems begin to train on data substantially produced by previous generations of those same systems. As a result, the quality, diversity, and reliability of the generated material decline: statistically rare elements disappear, extreme and marginal values are compressed, and the informational environment becomes increasingly homogeneous.

For cultural and media analysis, this phenomenon is especially important because it reveals a structural limitation of algorithmic production. A generative system grounded in probabilistic logic tends toward standardization and the reproduction of the most frequent patterns. What has often served throughout the history of culture and art as a source of rupture—an unconventional gesture, risk, error, deviation, or visual or conceptual disruption—appears within the statistical model as noise to be smoothed out.

Seen in these terms, the algorithmic environment is fundamentally ill-equipped to engage with the rare, the non-obvious, the paradoxical, and the genuinely new. It does not so much invent as recombine what already exists within a given probabilistic field. This is especially evident in domains where value is determined not by functional correctness, but by the force of authorial expression, the capacity to disrupt expectation, and the ability to propose a new frame of perception.

For this reason, model collapse matters not only as a technical problem in machine learning, but also as a metaphor for a broader cultural risk: the transformation of the digital environment into a space of endless self-citation, where what disappears is not information as such, but the possibility of radically new experience.

Alongside the rise of algorithmic production in the digital environment, another process is becoming increasingly visible: a gradual decline in trust toward open platforms. As generative production expands, the digital sphere is also experiencing a broader crisis of trust in the information circulating within open online environments. Under conditions of informational overload, content saturation, and the spread of disinformation, users are increasingly confronted with the difficulty of distinguishing between reliable and unreliable messages. As a result, trust is shifting away from the abstract public platform and toward more selective, value-aligned sources of communication (Edelman, 2025).

In practice, this shift manifests itself in a movement away from the logic of mass reach toward the logic of intimate presence. Users are increasingly migrating into closed or semi-open formats: thematic communities, messenger groups, subscription-based clubs, local educational spaces, and author-led channels marked by a high degree of personal involvement. In these contexts, value is determined not by the number of views, but by the quality of interaction, the repeatability of contact, the recognizability of voice, and the possibility of genuine intellectual exchange.

For this reason, in my own practice I have consciously shifted my focus toward closed educational formats, small groups, and an author-led art club in which communication is structured around live presence, repeated contact, and deep intellectual engagement. For projects in the humanities and cultural field, this is especially significant: here, audiences respond not to the volume of publication, but to the degree of authorial involvement, the distinctiveness of perspective, and the quality of shared experience.

Under these conditions, one can speak of a fundamental shift: the digital media platform is increasingly becoming a space of communication, co-presence, and collaborative meaning-making, where the value of what is difficult to scale algorithmically begins to rise.

• 2.    Framing the Problem: The Algorithm and the Limits of the Creative Act

  2.1.    Algorithmic Deadlock and the Irrational Nature of Creativity

2.2.    Contemporary Art as an Anti-Algorithmic Environment

The central tension of the contemporary media environment lies in the attempt to represent the creative act as a sequence of reproducible operations. From the perspective of platform logic and generative systems, this appears rational: if content can be produced faster, more cheaply, and at greater scale, then the problem would seem to be solved. Yet this logic remains persuasive only so long as creativity is understood as the recombination of already known elements.

The methodological framework proposed here proceeds from a broader understanding of creative thought. Creativity is not merely the recombination of pre-existing units, but the subject ’ s capacity to produce meaning under conditions of uncertainty. It emerges through internal resistance, doubt, the negotiation of cognitive barriers, the ability to work with ambiguity, and the willingness to remain with what is not yet fully resolved. These dimensions are especially important in the age of generative systems, where fluency can easily be mistaken for authorship.

It is precisely here that the limit of the algorithmic approach becomes visible. By its very architecture, a generative system tends toward probabilistic normativity: toward the statistically most plausible continuation. Human creativity, by contrast, often begins precisely where that norm is disrupted. What appears to the model as error, noise, malfunction, or irrelevant deviation may, in human experience, become a point of bifurcation—a site at which a new form, a new image, or a new way of seeing emerges.

The history of contemporary art repeatedly demonstrates this principle. Radical artistic gestures— from the avant-garde and abstraction to conceptual art and performance—have often arisen not from adherence to established norms, but from the deliberate disruption of habitual systems of representation. In this sense, art does not merely illustrate the creative process; it offers a particularly lucid model of thought that operates against predictability.

For precisely this reason, contemporary art and contemporary artistic thinking occupy a central place within my methodological framework. Contemporary art—especially in its non-classical, abstract, conceptual, intermedial, and process-based forms—does not offer the viewer a single correct reading. On the contrary, it demands active interpretation, the capacity to tolerate uncertainty, to assemble meaning from fragments, and to work through ambiguity.

This is precisely what makes it fundamentally valuable for the media environment. If platform algorithms seek to keep the user within a regime of predictable repetition, continually offering “ more of the same,” art demands the opposite: an effort of decoding and an internal reconfiguration of perception. It pulls the subject out of passive consumption and restores an active interpretive position.

Such an understanding of art is especially important in an age of digital saturation. Today, art can no longer be understood solely as an object of cultural consumption or a marker of status; it may also be approached as an instrument of cognitive training. Engagement with visually complex, open-ended, and not fully resolvable artistic forms cultivates precisely those capacities that become critically important in the digital environment: cognitive flexibility, tolerance for uncertainty, the capacity for divergent thinking, critical interpretation, and the formation of one’s own voice.

In this sense, art is treated in the present article not as a decorative supplement to the question of AI, but as a method for resisting the logic of the template.

• 3.    Theoretical Framework: Art as a Method for Developing Subjectivity

  3.1.    The Author’s Methodology and Its Interdisciplinary Foundation

3.2.    Key Cognitive Mechanisms: From Divergent Thinking to Visual Interpretation

The present article builds on an interdisciplinary methodological framework situated at the intersection of art history, creativity studies, aesthetics, pedagogy, and cognitive theory. Its central premise is that modern and contemporary art can serve not only as an object of analysis, but also as a practical medium for cultivating imagination, creativity, critical thinking, and innovative capacity— capacities that become especially significant in algorithmically structured media environments.

This logic becomes especially important in the context of the media environment, where the author increasingly works under conditions of information overload, high speed, fragmented attention, and the pressure of algorithmic templates. In such an environment, art becomes neither a luxury nor an optional humanistic embellishment, but an instrument for preparing a subject capable of sustaining complexity.

The methodology draws on four theoretical foundations:

• -    Divergent thinking (Guilford, 1950): the capacity to generate multiple interpretations from the same material

• -    Aesthetic experience (Dewey, 1934): art as active transformation of the perceiving subject

• -    Cultural mediation (Vygotsky, 1971): art as a "social technique of feeling"

• -    Visual thinking (Arnheim, 1968; Goodman, 1976): perception as a form of thought prior to verbal articulation

Modern and contemporary art thus emerges as a training environment in which precisely those capacities are developed that remain least susceptible to algorithmic reproduction.

The second layer is linked to the philosophy of aesthetic experience. In the spirit of John Dewey ’ s (Dewey, 1934) thought, art may be understood as a form of active experience that does not merely communicate information, but transforms the perceiving subject. Aesthetic experience here is not passive: it demands participation, involvement, and the correlation of the external image with inner experience. This is especially important for the media producer, who must not only recognize cultural codes, but also be able to transform perception into a form of expression.

The third layer is connected to cultural-historical psychology and to Vygotsky ’ s (Vygotsky, 1971) understanding of art as a form of cultural mediation of experience. Within this logic, the work of art may be understood as a specific “ social technique of feeling”—a form through which the subject learns to structure emotional and meaning-making experience. This is precisely why art becomes significant not only in aesthetic terms, but also in cognitive and pedagogical ones: it functions as a space for the formation of perception, imagination, interpretation, and reflection.

Finally, a crucial role in the proposed approach is played by theories of visual thinking and by the analysis of art as a field structured through rhythm, tension, balance, compositional logic, and symbolic

systems. In line with Rudolf Arnheim ’ s ideas, visual perception is understood not as the passive registration of an image, but as a form of thought in which relations between elements are grasped prior to verbal articulation. Combined with Nelson Goodman ’ s approach, which treats art as a mode of working with complex symbolic systems, this makes it possible to understand visual analysis as a fully developed cognitive practice (Arnheim, 1968). For the contemporary media author working in an environment saturated with visual communication, this capacity becomes fundamentally important (Berger, 1972).

Modern and contemporary art thus emerges as a training environment in which precisely those capacities are developed that remain least susceptible to algorithmic reproduction.

3.3.    The Four Stages of the Author’s Methodology

In practical terms, this methodology unfolds through four interrelated stages:

• -    Overcoming psychological barriers: work on fear of error, dependence on external validation, and habitual patterns shaped by platform metrics

• -    Creative discovery: transition from literal perception to the capacity to work with uncertainty, fragmentation, and incompleteness

• -    Formation of idea and image: connection of visual analysis with personal experience, memory, and cultural context

• -    Critical thinking and communication: capacity to articulate a position, build an argument, and sustain ethical responsibility for meaning produced

• 4.    Research Methodology and Authorial Position

  4.1.    Methodological Approach

• 5.    Practical Dimension: AI as Infrastructure, Not a Source of Authorship

  5.1.    AI as a Tool for Organizing Research and Media Work

These four stages constitute the foundation of authorial subjectivity in the media environment.

The article is conceptual-analytical in orientation, grounded in theoretical review, critical reflection on the author's methodological framework, and analysis of professional experience in cultural education and author-led media formats. It does not claim quantitative empirical verification, but proposes a conceptual model for describing the transformation of authorship in the digital environment.

Given the chosen research perspective, the article does not claim quantitative empirical verification in the strict sociological sense. Its aim is different: to propose a conceptual model for describing the transformation of the author ’ s role in the digital environment, and to demonstrate how humanistic practices—particularly engagement with modern and contemporary art—can function as instruments for preserving subjectivity. It should be noted that the present model is grounded primarily in the author's professional experience within cultural and educational contexts. Future research might productively examine how this framework applies across different media genres, professional communities, and institutional settings.

In my own practice, I use AI not as a substitute for the creative act, but as a means of organizing complex intellectual processes. This is especially important for an independent researcher and author

who works simultaneously across multiple types of tasks: preparing lectures, analyzing sources, designing educational programs, maintaining independent media formats, developing methodological materials, and engaging in strategic planning.

Within this logic, AI performs infrastructural functions: navigation, structuring, knowledge base creation, editorial diagnostics, and automation of routine processes. This frees cognitive resources for comparison, interpretation, conceptualization, and argumentation.

A typical workflow: I draft conceptual notes, then use AI to identify structural gaps and suggest alternatives. The final interpretive structure and core argument remain decisively human-authored.

For the humanities researcher, this is especially important, because the freeing of time here does not imply the abandonment of intellectual labor; on the contrary, it makes it possible to concentrate effort on the most demanding tasks: comparison, interpretation, conceptualization, argumentation, the selection of what is significant, and the construction of a coherent authorial structure.

A typical workflow in my own practice illustrates this boundary clearly. When preparing a lecture or analytical text, I begin by drafting my own conceptual notes: key questions, tensions, associations, and provisional arguments. Only after this initial human stage do I use AI tools to identify structural gaps, detect repetitions, or suggest alternative organizational pathways. I may also use them to generate discussion prompts or compare possible framings. The final interpretive structure, the core argument, and the finished text remain decisively human-authored. In this sense, AI supports the architecture of the work, but not the authorship that gives it meaning.

5.2.    The Limit of Applicability: Why Authorial Writing Cannot Be Fully Delegated

For all the effectiveness of AI as a technical and organizational tool, there is a fundamental boundary beyond which delegation ceases to be productive and begins to weaken authorial practice itself.

A generative system can perform tasks requiring speed and templated logic, but full delegation weakens authorial practice. An authorial text is valuable not only for what it communicates, but for how it comes into being: interruptions, tensions, inner rhythm, and individual argumentation. These "imperfections" carry living thought.

In my practice, I draft the semantic core and tonal framework, then use AI for structure and diagnostics. In essayistic and personally saturated writing, I avoid algorithmic editing entirely.

At the same time, there are genres into which I deliberately do not allow algorithmic editing at all— above all, essayistic and personally saturated forms of writing, where not only the thought itself matters, but also the form of its emergence. In such genres, even careful algorithmic correction may prove excessive, since the smoothing of style often also means the weakening of the text ’ s subjective tonal register.

5.3.    The Architect of Meaning as a New Role Model

On the basis of the theoretical and practical material outlined above, I propose an analytical model for a new role within contemporary media production: the architect of meaning.

Unlike the classical figure of the “ content maker,” whose primary orientation is toward the regular production and distribution of materials, the architect of meaning operates according to a different logic. Their task is not the endless expansion of output, but the design of intellectual and communicative

environments in which durable relations of meaning emerge between knowledge, experience, visual culture, community, and technology (van Dijck et al., 2018).

The architect of meaning uses AI to automate routine and infrastructural processes, builds personal systems for the storage, selection, and verification of knowledge, works with reliable sources and consciously limits informational noise, develops a personal methodology of interpretation, creates not merely content, but a space of intellectual presence, sustains ethical and semantic responsibility for the utterance produced.

In this sense, the issue is not a struggle between human and machine, but a redistribution of functions. The more effectively the algorithm handles the technical and infrastructural layer, the more decisive becomes the human capacity for selection, interpretation, the linking of heterogeneous elements, the formation of a position, and the creation of forms of experience that cannot be reduced to a template (Mollick, 2024).

• 6.    Modern and Contemporary Art as an Instrument of Cognitive Flexibility in the Media Environment

  6.1.    Art as a Practice of Interrupting Algorithmic Repetition

6.2.    From Aesthetic Experience to Expanded Media Literacy

If generative AI tends to reinforce patterns of probabilistic repetition, modern and contemporary art trains the capacity to interrupt predictability. This is precisely why it becomes especially valuable not only for artists or art historians, but for everyone working with media, communication, education, visual language, and cultural production.

Modern and contemporary art—especially in its abstract, conceptual, interdisciplinary, and open-ended forms—demands active cognitive and interpretive work from the viewer (Dewey, 1934). The viewer is not given a ready-made answer; instead, they are compelled to construct connections, sustain ambiguity, shift attention between visual and semantic levels, relate what they see to their own experience, and thereby participate in the production of meaning.

This process is precisely what matters in the digital environment. Under conditions in which platform-based media tend to reinforce repetition, recognizability, and predictable response, art introduces rupture, tension, and the necessity of effort into the field of perception (Eisner, 2002). It restores the subject ’ s capacity not merely to consume signals, but to work actively with them.

In the context of the present article, it is important to emphasize that engagement with art is not limited to the development of cultural literacy. Its outcome may also be the formation of a range of cognitive, interpretive, and communicative competencies (Eisner, 2002) that prove productive in the contemporary media environment (Tishman, 2018).

These include: multilayered reading, recognition of symbolic structures, tolerance for uncertainty, divergent thinking, original composition, cultivation of attentiveness, and transformation of perception into authorial position.

For this reason, modern and contemporary art is understood in this model as a means of cultivating a deeper form of media literacy. What is at stake is not only technical literacy, and not only a critical attitude toward sources, but a deeper level altogether: the capacity to work with complexity, to

6.3.    From the Author-as-Producer to the Author-as-Interpreter

In the traditional logic of media production, the author is often understood primarily as a producer, someone who produces material within a given genre and for a specific task. In a digital environment saturated with AI tools, this model becomes insufficient. As the production layer becomes partially automated, the author ’ s value shifts ever more clearly toward interpretation, selection, framing, composition, and the construction of context (Mollick, 2024).

It is precisely at this point that modern and contemporary art, understood as method, acquires particular importance. It contributes to the formation of the author as interpreter capable of perceiving heterogeneous connections, sustaining multilayeredness, and creating utterances grounded in internal logic rather than external demand.

Such an author becomes what, within the framework of this article, I designate as the architect of meaning. Their value is determined not by the speed of production, but by the capacity to transform cultural, visual, and personal experience into a media utterance that is intellectually rich, ethically responsible, and able to sustain meaningful communicative presence.

• 7.    Results and Conceptual Conclusions

  7.1.    A Shift in Scarcity: From the Scarcity of Content to the Scarcity of Meaning

  The first key conclusion is that the digital environment of the 2020s is marked by a shift in the fundamental form of scarcity. Whereas the limiting resource was once the means of producing and distributing content, today the scarce resource is no longer the mere possibility of publication, but the capacity to create meaning structures that are genuinely significant, distinct, and irreducible to one another (van Dijck et al., 2018).

7.2.    AI Does Not Eliminate the Author, but Radically Transforms the Author’s Function 7.3.    Modern and Contemporary Art as a Strategic Resource for the Media of the Future

This means that the author who is oriented exclusively toward scale, speed, and regularity of output finds themselves in an increasingly unstable position: it is precisely in this zone that algorithmic systems are becoming ever more competitive (Mollick, 2024). By contrast, the more sustainable strategy is one grounded in depth, selection, contextualization, intellectual honesty, and a distinctive personal perspective.

The second conclusion is that generative AI does not eliminate the figure of the author, but radically redistributes its functions. Where a substantial portion of time was once devoted to technical assembly, routine navigation, initial structuring, and the organization of material, there is now an opportunity to delegate part of this work to algorithmic tools (Mollick, 2024).

This does not, however, mean an automatic increase in quality. The time and cognitive resources thus freed can be used either to deepen authorial work or to accelerate the reproduction of superficial informational noise (Bender et al., 2021). The key question, therefore, lies not in the mere presence of the technology, but in the author ’ s cultural and cognitive readiness to use it in the service of complication rather than simplification.

It is precisely here that the figure of the architect of meaning emerges: an author who delegates the technical to the machine, but does not relinquish the intellectual and ethical core of their work.

The third conclusion concerns a reassessment of the role of modern and contemporary art in the context of media education and the training of professionals working in digital environments. Within the model proposed here, modern and contemporary art should be understood not as an optional humanistic field, but as a strategic resource for cultivating those capacities that become especially important in the age of generative systems (Greene, 1995).

These include, above all:

• —    the development of cognitive flexibility;

• —    the capacity for divergent thinking;

• —    the ability to work with visual complexity;

• —    the formation of interpretive skills;

• —    the ability to sustain uncertainty;

• —    the capacity to form a personal position;

• —    the cultivation of attentiveness and depth of perception (Tishman, 2018).

8. Conclusion

In this context, art appears not as an external supplement to the technological environment, but as one of the most productive means of preserving the human dimension within it.

This article has proposed an answer grounded in an interdisciplinary framework that brings together art history, the psychology of creativity, aesthetics, media theory, and the practice of independent cultural production. That answer can be stated as follows: The author ’ s sustainable place within digital media no longer lies in the domain of speed and scalability, but in the domain of meaningmaking, interpretation, intellectual risk, ethical responsibility, and the capacity to construct distinctive structures of meaning.

The central question, then, is not whether AI will replace authors, but which forms of authorship will remain viable in an algorithmically saturated media environment. Those who compete with generative systems on speed, volume, and repeatability will inevitably lose. Those who cultivate interpretive depth, ethical clarity, and the capacity to create irreducible meaning will define the media ecosystems of the future. In this sense, the future of media is not only a technological problem—it is a cultural, cognitive, and profoundly human one.