Bakhtinian Polyphony as an Epistemological Paradigm for Analyzing Authorship in Generative Artificial Intelligence Texts

ABSTRACT

We examine the historical evolution of authorship and contemporary tensions surrounding textual production by artificial intelligence systems, arguing that polyphony, as conceived by Mikhail Bakhtin, offers a consistent interpretative key for understanding the absence of full authorship in such cases. The generation of texts by generative models does not entail the presence of independent discursive consciousnesses nor of axiologically oriented voices engaged in dialogic interaction, a condition essential to the configuration of authorship in the polyphonic sense. This analysis contrasts such a perspective with approaches like the Turing test and the Chinese room argument, emphasizing that these models operate within functionalist and simulation-based limits. In contrast, polyphony requires the coexistence of responsive subjectivities, articulated through shared and situated language, which AI systems are still incapable of achieving. In this regard, polyphony proves to be an effective analytical operator in delineating authorship.

KEYWORDS:
Authorship; Polyphony; Generative Artificial Intelligence; Consciousness; Intentionality

RESUMO

Examinaremos a evolução histórica da autoria e as tensões contemporâneas em torno da produção textual por sistemas de inteligência artificial, argumentando que a polifonia, conforme concebida por Mikhail Bakhtin, oferece uma chave interpretativa consistente para compreender a ausência de autoria plena nesses casos. A produção de textos por modelos generativos não implica a presença de consciências discursivas independentes nem de vozes axiologicamente orientadas em interação dialógica, condição essencial para a configuração da autoria no sentido polifônico. A análise contrasta essa perspectiva com abordagens como o teste de Turing e o argumento da sala chinesa, destacando que tais modelos operam dentro de limites funcionalistas e simulacionais. Em contraposição, a polifonia exige a coexistência de subjetividades responsivas, articuladas por meio de linguagem compartilhada e situada, o que os sistemas de IA ainda não são capazes de realizar. Nesse sentido, a polifonia revela-se como um operador analítico eficaz na delimitação da autoria.

PALAVRAS-CHAVE:
Autoralidade; Polifonia; Inteligência Artificial Generativa; Consciência; Intencionalidade

Introduction

Authorship has, historically, shifted from collective and oral matrices to individualized forms, strengthened by the literate tradition and the modern consecration of the author as a singular source of meaning. The diffusion of generative artificial intelligence (AI) systems reopens this debate by producing texts that imitate human styles without any attributable consciousness or intentionality. In much of the technical-philosophical discussion, functionalist approaches-such as the Turing¹ imitation game and the Chinese room argument-privilege behavioral criteria of simulation, relegating properly dialogical and axiological dimensions to the background. In the light of Mikhail Bakhtin’s thought, polyphony offers an analytical framework for conceiving the text as a multivocal utterance supported by responsive and relatively autonomous voices; such a requirement allows one to interrogate the sufficiency of algorithmic productions in configuring authorship and discursive responsibility.

This article aims to: (i) reconstruct, in broad strokes, the evolution of the concept of authorship; (ii) delineate the limits of functionalist models applied to AI texts; (iii) propose Bakhtinian polyphony as an epistemological tool for analyzing the presence-or absence-of independent voices in generated texts; and (iv) discuss consequences for ethical debates and editorial practices in digital humanities. The structure is as follows: section 1 presents a historical overview of conceptions of authorship in Western and Eastern traditions; section 2 discusses contributions from literary theory; section 3 contrasts functionalist tests with the notion of polyphony; the final considerations synthesize implications for authorship in AI contexts.

1 Historical Overview

In Antiquity, the production and transmission of texts occurred mainly through oral tradition and manual copying by scribes, so that the figure of the individual author was not seen as the exclusive source of creation. Knowledge was considered a collective legacy, and the works often formed part of a shared cultural heritage in which texts were constantly reinterpreted and adapted. The written fixation of oral performances changes the way of reading and attributing authorship. As Walter Jackson Ong observes in his work Orality and Literacy,² when truly oral enunciative practices are transcribed, they tend to be studied as relatively stable texts, which dampens the performative, situational and often collective dimension of these productions, favoring the impression of a unified authorial subject.

In the Middle Ages, the transmission of texts was oriented by the authority of religious teachings and collective tradition. Sacred and classical texts were copied and disseminated in communal contexts in which the value of the message outweighed the importance of the identity of its creator. In this environment, the role of the author was secondary in comparison with the function of the text as a vehicle of knowledge and faith, reflecting a culture in which collectivity prevailed over individuality.

With the advent of the Renaissance, a growing appreciation of human potential and individual thought arose, driven by humanist ideals that rediscovered and celebrated classical culture. In this period, the figure of the author came to be understood as the original and unique source of a work, bearing its own intentions, identity and vision. This new outlook on literary creation gained even more intensity in Romanticism, when the exaltation of the creative self, subjectivity and emotion became central in artistic production. In the warning of his Aesthetic as Science of Expression and General Linguistic, the Italian philosopher Benedetto Croce writes that “art is expression of impressions, not expression of expression” (Croce, 1922, p. 13).³ Throughout his work he defends that art is essentially the expression of the author’s personality, which reinforces the idea that each work carries a unique mark of its creator.

The Renaissance approach involved intense translation work, editing and critique of ancient texts. Jacob Burckhardt (2014)⁴ describes how this cultural movement was decisive in shaping the Renaissance identity, emphasizing the importance of recovering the classics to affirm a new humanity ideal.

In the course of the fifteenth century the works of antiquity were discovered and diffused with extraordinary rapidity. All the writings of the Greek philosophers which we ourselves possess were now, at least in the form of Latin translations, everybody’s hands. It is a curious fact that some of the most zealous apostles of this new culture were men of the strictest piety, or even ascetics [...]The result of all these tendencies was that the Platonic Academy at Florence deliberately chose for its object the reconciliation of the spirit of antiquity with that of Christianity (Burckhardt, 2014, p. 499).⁵

Figures such as Marsilio Ficino and Giovanni Pico della Mirandola, for example, dedicated themselves to reinterpreting Plato’s ideas and other thinkers, contributing to these authors being seen not just as relics of the past but as living sources of inspiration for the new era. This recomposition process implied an authorship revaluation: humanists came to consider the classics as expressions of individual genius, even though originally their works had been transmitted more collectively and less centered on the author figure.

Thus, the evolution of the authorship concept shows a historical transition in which knowledge, initially shared and collective, gradually turned into an expression of individual genius. This change altered the practices of text production and transmission and transformed perceptions of creativity, originality and identity, leaving a legacy that continues to influence contemporary discussions about authorship in the arts and literature.

In Eastern traditions, various canons reveal processes of composition and transmission in which authorship is distributed over time. The Analects of Confucius (Confucius, 1997)⁶ articulate magistral voice and scholastic sedimentation; the chronicles of Sima Qian (2009) integrate narratives of multiple provenances; the transmission of the Sutras supposes cumulative commentary; Indian epics like The Mahabharata (2005) and the Ramayana (Rajagopalachari, 2013) show strata and expansions; and The Tale of Genji (Murasaki Shikibu, 2018)⁷ reverberates in courtly intertextual dialogues. In these cases, the authorial identity emerges less as an exclusive origin and more as a knot in a network of voices, exegetical practices and historical interventions.

These experiences suggest that, in different regions of the East, the notion of a unique author often gives way to cultural reinterpretation and compilation, in which the work represents the confluence of lineages of thought, historical commentaries and ritual memorization practices. Just as in the West, where the transition from orality to writing redefined the foundations of authorship, in the East political, religious and literary transformations shaped the way people conceive the figure of the one who composes or transmits a text. The historicity of the author’s idea therefore is not limited to a linear movement from anonymity to individuality but to a network of influences and reinterpretations that converge in different cultural contexts.

2 Critique and the Authorship Problematics in Literary Theory

Literary work emerges as a space in which the author’s personal experience and numerous accumulated influences intertwine, producing a text that dialogues with tradition and at the same time proposes innovations. Harold Bloom argues that poets are influenced by their predecessors in a process he calls the ‘anxiety of influence,’ in which the new poet seeks to differentiate himself from the poetic father to establish his own voice.

The later poet opens himself to what he believes to be a power in the parent-poem that does not belong to the parent proper, but to a range of being just beyond that precursor. He does this, in his poem, by so stationing its relation to the parent-poem as to generalize away the uniqueness of the earlier work (Bloom, 1973, p. 15).⁸

For Bloom, this conflictual and transformative relationship is fundamental to the emergence of works that, while rooted in tradition, manage to produce something original. Furthermore, Roland Barthes reinforces the idea that texts are constructed through a web of intertextuality:

A text is made of multiple writings, drawn from many cultures and entering mutual relations of dialogue, parody, contestation, but there is one place where this multiplicity is focused and that place is the reader, not, as was hitherto said, the author. The reader is the space on which all the quotations that make up a writing are inscribed without any of them being lost; a text’s unity lies not in its origin but in its destination (Barthes, 1977, p. 148).⁹

According to Barthes, the meaning of a work is not fixed in the intention or biography of the author but is formed in the encounter and overlapping of multiple references, which subverts the notion of authorship as the exclusive origin of meaning (Barthes, 1977).¹⁰ This perspective points to a creation that is, in essence, a process of constant re-signification of the influences that surround it.

Mikhail Mikhailovich Bakhtin contributes the notion that literature is a space of multiple voices and dialogues, in which the interaction between diverse discourses enriches the field of meaning of the work. For Bakhtin, the presence of different perspectives within the same text demonstrates that authorship is not an isolated act; it is, however, a process profoundly influenced by cultural and literary traditions:

The novel orchestrates all its themes, the totality of the world of objects and ideas depicted and expressed in it, by means of the social diversity of speech types [raznorečie] and by the differing individual voices that flourish under such conditions. Authorial speech, the speeches of narrators, inserted genres, the speech of characters are merely those fundamental compositional unities with whose help heteroglossia [raznorečie] can enter the novel; each of them permits a multiplicity of social voices and a wide variety of their links and interrelationships (always more or less dialogized) (Bakhtin, 1981, p. 263).¹¹

In his study of Dostoevsky, Bakhtin explores how ideas (cognitive content), when incorporated into the art of the novel, are freed from a merely logical or argumentative function and gain a life of their own in artistic dialogue. He clarifies that, in Dostoevsky, the idea ceases to be an abstract concept and becomes a lived event in the contact between consciousnesses, seeking to be understood and responded to by diverse voices. This dialogic character of the idea in the novel highlights the non-instrumental nature of art - the idea does not appear to prove a univocal thesis, but to be freely explored:

The idea is a live event, played out at the point of dialogic meeting between two or several consciousnesses. In this sense the idea is similar to the word, with which it is dialogically united. Like the word, the idea wants to be heard, understood, and “answered” by other voices from other positions. Like the word, the idea is by nature dialogic (Bakhtin, 1984, p. 88).¹²

In this excerpt, Bakhtin illustrates the potential of art to liberate ideational content from predetermined purposes. Instead of serving a closed conclusion, the idea, upon entering the artwork, detaches itself from a single purpose and begins to exist on its own as a living dialogue between perspectives. Art here operates in a non-teleological way: Dostoevsky does not use the characters and their ideas to directly defend a prior truth or a univocal moral teaching but creates an arena in which multiple ideological voices interact openly. This lack of immediate practical purpose does not mean indifference - on the contrary, there is a profound engagement with conflicts of value and thought, but in dialogical, not dogmatic, terms. Disinterested in utilitarian purposes, the artistic form is interested in giving full resonance to each voice and worldview. As Renata Coelho Marchezan complements:

Therefore, we see that Bakhtin favors considerations about literary work, but he always situates the aesthetics in a more general domain of human culture, which also includes science/knowledge and ethics. Art is also not indifferent to these two domains; it welcomes them and thus provides concrete unity to culture. When doing it, art releases the content of knowledge and the ethical act from the links, determinations and interests that it has in its specific domains. It is exactly this isolation -a separation that could be quickly confused with the concept of defamiliarization-from the content of its domains which also composes the explanation for the relative autonomy enjoyed by art; it is this isolation that makes possible the non-teleological, non-instrumental, disinterested, but not indifferent activity of the author-creator (Marchezan, 2015, pp. 209-210).¹³

Bakhtin (1986)¹⁴ argues that the writer’s creative voice is necessarily a second voice, that is, a voice that sounds through another. He suggests that a completely univocal word would be naive and insufficient for true art. In contrast, the authentic literary word is bivocal: the author does not speak directly and literally, but lends his voice to other voices, speaking through them. Thus, the author’s voice inserts itself behind the voice of the character, the narrator, or a particular stylized language:

Perhaps any literal, single-voiced word is naive and unsuitable for authentic creativity. Any truly creative voice can only be the second voice in the discourse. Only the second voice - pure relationship - can be completely objectless and not cast a figural, substantive shadow. The writer is a person who is able to work in a language while standing outside language, who has the gift of indirect speaking [...]. In this case, our own discourse becomes an object and acquires a second - its own - voice (Bakhtin, 1986, p. 110).¹⁵

In other words, the skilled writer manages to distance himself from common language and direct speech, only to reinsert them into the work in an artistic way, making them speak with double meaning. This is the principle of bivocal discourse: in literary works, the speech of characters or narrators simultaneously carries two voices - the character’s own voice and the author’s voice refracted in that speech. For example, in an ironic narration, the character’s words (first voice) are intentionally shaped by the author so that the reader also perceives the author’s attitude (second voice) towards those words. Bakhtin says that only this “pure” second voice (that is, which expresses only the author’s relationship, without its own content) can be totally “objectless” - because it does not introduce a new object of discourse, it only refracts an already existing one. This confirms the presence of the phenomenon of free or doubly oriented indirect discourse in literature, in which “our own discourse becomes an object and acquires a second - its own - voice” (Bakhtin, 1986, p. 110)¹⁶ within the artistic text. Bakhtin, therefore, defines the writer/artist as one who speaks with an alien voice, or rather, who manages to make his intention resonate through the social and ideological voices of the characters - a second voice that enriches the dialogue of the work without ever presenting itself in a flat or univocal way.

Bakhtin’s notion of polyphony, developed from reading Dostoevsky’s novels as orchestrations of autonomous voices in responsive interaction, can be extended based on the thesis that language is constitutively dialogic and that the novel achieves maximum dialogism. As Gary Saul Morson and Caryl Emerson summarize, “the nineteenth-century novel, and especially the novels of Dostoevsky, are the prime exemplars of this kind of maximal dialogism” (Morson; Emerson, 1989, p 52). This basis allows us to shift the focus from the strictly literary to cultural spheres, where dialogism is understood as a principle that leads from intralinguistic to intracultural relationships. From this articulation it follows that recognizing the dialogic dimension of authorship does not imply demanding a polyphonic text in a strict sense; it is about evaluating the responsive interaction of axiologically plural voices, applicable to contemporary textual productions in diverse cultural contexts, in light of the dialogic framework outlined above.

In parallel, Roland Barthes proposes that fixing the meaning of a text on the author’s biography and intention restricts the multiple possible interpretations: “To assign an Author to a text is to impose a brake on it, to furnish it with a final signified, to close writing. [...] once the author is found, the text is ‘explained,’ the critic has won” (Barthes, 1986, p. 54).¹⁷ For Barthes, the work becomes dissociated from the singular authority of the creator, allowing the reader to reconstruct its meaning freely and in a plural way. Thus, the text becomes autonomous, breaking with the tradition that placed the author as the sole and definitive source of meaning.

To further this discussion, it is worth mentioning other contributions from Bakhtin, who highlights the polyphony and dialogical nature of texts. For Bakhtin, literary works are configured as spaces of interaction between multiple voices and consciousnesses, which diminishes the relevance of a singular authorship. In his analysis, meaning emerges from the confrontation and coexistence of diverse perspectives, evidencing that textual creation is a collective and dynamic phenomenon. Walter Benjamin also adds an important dimension to the debate by addressing the question of the aura and originality of works: “for the first time in world history, mechanical reproduction emancipates the work of art from its parasitical dependence on ritual. To an ever greater degree the work of art reproduced becomes the work of art designed for reproducibility” (Benjamin, 1969, pp. 223-4).¹⁸ It is noted that he questions the singularity and authenticity that are traditionally attributed to the author.

Thus, the problem of authorship reveals a constant tension between the exaltation of creative genius and the understanding that texts are configured as spaces open to a multiplicity of interpretations. This contemporary perspective shifts the focus from the author’s exclusive intention to the interaction between reader, text, and cultural context, offering a more complex view of the production of meaning. Such reflection allows us to understand that, although the figure of the author continues to play a relevant role, it is only one part of a broader process of constructing and circulating meaning.

Furthermore, authorship can be understood not as an expression of an immaculate individual essence, but as the result of a creative process permeated by external influences. Originality manifests itself in the author’s ability to rearticulate, transform, and engage with this vast field of references, creating works that are simultaneously part of a tradition and an innovation within it.

3 Singularity, Turing Test, Chinese Room and Bakhtinian Polyphony

Generative artificial intelligence (AI) is a branch of machine learning dedicated to creating new and original data by deeply understanding the patterns present in large volumes of information. This field has evolved significantly with the development of models capable of synthesizing images, texts, music, and other forms of content, allowing the reproduction of existing information and the autonomous creation of new data. Based on the central concept of neural networks, these models are trained on vast datasets to learn complex representations, which give them the ability to generalize and generate samples that resemble the original distributions of the training data (Goodfellow et al., 2020).

The advent of attention-based architectures and sequential models has further transformed the landscape of generative AI, enabling the development of systems that produce coherent and contextually relevant outputs. Such advances can be observed in the evolution of language models, which demonstrate a certain mastery in creating texts that reproduce literary styles and articulate arguments in a logical and integrated way. This progress is the result of combining sophisticated statistical techniques with optimization methods that adjust model parameters based on large volumes of data, contributing to a better understanding and representation of the structures underlying language and visual perception (Vaswani et al., 2017).

In this context, advances in generative artificial intelligence have fostered an increasingly intense debate in the field of epistemology regarding authorship in texts produced by artificial systems.

Ray Kurzweil, a futurist and computer scientist, is known for his optimistic predictions about technological progress, and particularly the evolution of artificial intelligence. According to him, we are heading towards a point of technological singularity by the mid-21st century, in which machines will be able to surpass human intelligence in several aspects and become capable of improving themselves. In this scenario, Kurzweil also argues that artificial intelligence will eventually achieve what we understand as human consciousness or, at least, something very close to it:

So how will we come to terms with the consciousness that will be claimed by nonbiological intelligence? From a practical perspective such claims will be accepted. For one thing, ‘they’ will be us, so there won’t be any clear distinctions between biological and nonbiological intelligence. Furthermore, these nonbiological entities will be extremely intelligent, so they’ll be able to convince other humans (biological, nonbiological, or somewhere in between) that they are conscious. They will have all the delicate emotional cues that convince us today that humans are conscious. They will be able to make other humans laugh and cry. And they will get mad if others do not accept their claims (Kurzweil, 2005, p. 378).¹⁹

Regarding human consciousness, psychoanalysis, as formulated by Freud between the late 19th and early 20th centuries, conceived of the mind as composed of three instances: conscious, preconscious, and unconscious (Freud, 2019). The conscious encompasses everything that is in the immediate field of perception; the preconscious includes accessible content but not present in consciousness at all times; and the unconscious involves repressed content that influences behavior without directly emerging to the conscious surface. Although psychoanalysis did not invent the concept of consciousness, it developed a broader description of the psychic apparatus, placing consciousness alongside other mental dimensions.

In general, consciousness can be understood as subjective or phenomenological experience (related to the sensation of feeling something, perceiving colors, sounds, and emotions), self-awareness (self-consciousness, or the ability to recognize oneself as an agent who observes, feels, and thinks), and intentional or reflective consciousness (the ability to turn attention to one’s own mental processes, thinking about one’s own thoughts).

Kurzweil does not directly engage with or systematically apply Freud’s tripartite model of the mind in The Singularity Is Near. The Austrian psychoanalyst is mentioned only in fictionalized conversational snippets, more as a cameo figure in Kurzweil’s dialogue scenes. These passages show Sigmund Freud interjecting questions or short observations about identity and the self, but Kurzweil never formally discusses or critiques Freud’s psychoanalytic structure of consciousness in the book. Instead of developing Freudian concepts, Kurzweil’s focus is on the nature of consciousness in relation to emerging technologies, pattern recognition, and the possibility of replicating the mind in non-biological media.

Kurzweil’s conception of consciousness is based on aspects such as high-level information processing; that is, when a system not only processes data but is aware of this process and can reflect on its own states. For him, the creation of complex internal models of the world and of oneself, with the capacity for adaptation, reflection, decision-making, and continuous learning, would bring AI closer to what we understand as conscious states. The absence of critical dialogue with the libidinal economy proposed by psychoanalysis is hardly explained solely by a supposed internal weakness in Kurzweil’s theory; rather, it reflects the epistemological distance between an engineering project concerned with mechanical replicability and a hermeneutic tradition focused on drive conflict.

In computer science, the Turing test,²⁰ initially proposed as an operational criterion to determine whether a machine can be considered intelligent or conscious (Turing, 1950), is limited to evaluating the imitative and performative capacity of an artificial intelligence in controlled linguistic interactions. Its fundamental limitation lies in the behaviorist conception implicit in the evaluation of artificial intelligence, focusing exclusively on the external appearance of linguistic comprehension and interaction without questioning the axiological and authorial dimension of the discourse produced. In this respect, even a sophisticated and convincing simulation could pass the test without any guarantee that an autonomous and genuinely authorial discursive consciousness was in operation.

Indeed, the Chinese Room argument,²¹ proposed by John Searle (1980), advances the critique of the reductionism of the Turing test by arguing that syntactic manipulations, even if complex and seemingly meaningful, do not necessarily result in semantic comprehension or linguistic awareness (Searle, 1980). The Chinese Room demonstrates that the machine can produce comprehensible texts without, however, subjectively experiencing the meanings it manipulates, highlighting the absence of true discursive awareness. However, this argument remains restricted to the critique of semantic comprehension, not directly encompassing the authorial or dialogic dimension of language, an essential aspect in determining the authorship of texts produced by generative systems.

On the other hand, functionalist currents in philosophy of mind defend the idea that if a system performs the same functions and processes as a conscious brain, it would be conscious regardless of its substrate. Kurzweil aligns with this view, believing that consciousness could be emulated when information processing and the complexity of interconnections reach a certain level. He dismisses the need for a mysterious component for consciousness to exist, believing that when AI is sufficiently advanced and able to perform self-updates, it will achieve the same essence as human consciousness.

The arguments I make in this book with regard to consciousness are for the purpose of illustrating this vexing and paradoxical (and, therefore, profound) nature of consciousness: how one set of assumptions (that is, that a copy of my mind file either shares or does not share my consciousness) leads ultimately to an opposite view, and vice versa (Kurzweil, 2005, p. 389).²²

The notion of intentionality is fundamental to understanding human experience and the subject’s relationship with the world and has been the object of study for philosophers since the 19th century. Franz Brentano formally introduced the concept by observing that the mind is characterized by always being directed towards something, that is, it possesses a property that makes it refer to a content or object:

Every mental phenomenon is characterized by what the Scholastics of the Middle Ages called the intentional (or mental) inexistence of an object, and what we might call, though not wholly unambiguously, reference to a content, direction toward an object (which is not to be understood here as meaning a thing), or immanent objectivity. Every mental phenomenon includes something as object within itself, although they do not all do so in the same way. In presentation something is presented, in judgement something is affirmed or denied, in love loved, in hate hated, in desire desired and so on (Brentano, 1995, p. 68).

Later, Edmund Husserl (2014) deepened this reflection in his phenomenology, expanding the discussion on how consciousness is directed towards objects and constitutes the meaning of these objects in experience. This phenomenological approach highlights the intentional structure as the core of conscious life, emphasizing the distinction between mental acts and that to which they refer.

Daniel Dennett proposes the so-called intentional stance to explain how we attribute intentions to agents, humans or otherwise, to predict and understand their behavior. In The Intentional Stance, he writes that “The decision to adopt the intentional stance is free, but the facts about the success or failure of the stance, were one to adopt it, are perfectly objective” (Dennett, 1989, p. 24). Although Dennett emphasizes the heuristic value of this stance, he questions whether such attribution of intentions effectively reveals genuine internal states or whether it functions as a useful interpretive strategy: “[...] it is an extraordinarily powerful tool in prediction-a fact largely concealed by our typical concentration on cases in which it yields dubious or unreliable results” (Dennett, 1989, p. 24).

At the heart of this discussion, intentionality involves the capacity of a conscious subject-endowed with mental states that refer to objects-to construct meanings and act upon the world. The question that permeates these debates is whether such direction toward the world necessarily requires a subjective experience or a first-person point of view. In short, the complexity of the intentional phenomenon suggests that it is not reduced to a simple processing of information but is intrinsically linked to lived experience and the conceptual framework typical of conscious agents. In this sense, machines, even if they operate based on sophisticated algorithms, do not demonstrate the core characteristic of intentionality discussed by Brentano, Husserl, and Dennett, because they lack the phenomenological quality and the meaningful relationship with the world that these authors identify as an indispensable condition for genuine intentional states to exist.

We maintain, at this point, that Bakhtinian polyphony reveals itself to be epistemologically more complete and assertive precisely because the concept transcends the superficial criteria of communicative appearance or syntactic-semantic manipulation to demand, as an essential condition for true authorship, the autonomous existence of independent and axiologically plural voices. In Problems of Dostoevsky’s Poetics, Bakhtin argues:

Dostoevsky is the creator of the polyphonic novel. He created an essentially new novelistic genre. [...] A character’s word about himself and his world is just as fully weighted as the author’s word usually is; it is not subordinated to the character’s objectified image as merely one of his characteristics, nor does it serve as a mouthpiece for the author’s voice. It possesses extraordinary independence in the structure of the work; it sounds, as it were, alongside the author’s words and in a special way combines both with it and with the full and equally valid voices of other characters (Bakhtin, 1984, p. 7).²³

In this context, Bakhtinian polyphony functions as an analytical operator that integrates dialogism as its essential realization, allowing its use in authorship assessments beyond the novel, such as in AI-generated texts. Studies that contextualize Bakhtin within interdisciplinary disciplines advocate for expanded uses of polyphony, positioning it as a paradigm for examining cultural interactions without restricting it to a specific literary genre (Mandelker, 1995). Bakhtin evolves this integration in the revised edition of his work, describing polyphony as a dialogic realization that emphasizes the coexistence of responsive perspectives, without demanding that every authorial text be polyphonic in the Dostoevskian model (Bakhtin, 1984).²⁴ Thus, the recognition of dialogism in authorship does not imply the obligation of a polyphonic text but highlights the absence of independent consciousnesses in artificial simulations.

Polyphony, in the Bakhtinian conception, depends directly on the autonomy of the discursive consciousnesses that present themselves without subordination to a monologic author, establishing authentic dialogic relations with each other. Even in its dialogical plurality, literary work remains mediated by the active consciousness of the one who organizes, selects and mobilizes these voices. The absence of this element makes it impossible for artificial intelligences, whose functional structure is oriented by predefined parameters and conditioned by prior training data, to fully satisfy this necessary condition for effective and true authorship in the polyphonic sense.

In the contemporary context, this mediating and conscious figure is precisely the human who interacts with artificial intelligence. The human user, as a conscious agent who selects and organizes the elements and commands supplied to AI, occupies a position similar to that which Bakhtin describes as that of the polyphonic author: the one who articulates voices and contexts, making those voices acquire meaning in their coexistence.

In Speech Genres and Other Late Essays, Bakhtin clarifies that verbal creation is never neutral; consequently, it is not generated simply by a technical mechanism or by mere chance. On the contrary, it is intrinsically linked to the living and conscious human context: “When we select words in the process of constructing an utterance, we by no means always take them from the system of language in their neutral, dictionary form. We usually take them from other utterance […]” (Bakhtin, 1986, p. 88).²⁵

Applying this reflection to textual production via AI, it becomes evident that artificial intelligence, when generating texts, is always responding to a conscious and intentional human impulse. The human, as the active and conscious addressee of the generated production, significantly guides the content, style and purpose of the text. AI, in this sense, acts as a technological instrument that amplifies the reach of this dialogical-and polyphonic-capacity of the human consciousness, while keeping the authorship of the latter intact.

But these voices have not yet become fully independent real voices, they are not yet three autonomous consciousnesses. This occurs only in Dostoevsky’s longer novels. In The Double there is no monologic discourse gravitating solely toward itself and its referential object. Each word is dismantled dialogically, each word contains an interruption of voices, but there is not yet an authentic dialogue of unmergedconsciousnesses such as will later appear in the novels. Already the rudiments of counterpoint are here: it is implied in the very structure of the discourse (Bakhtin, 1984, pp. 220-1).²⁶

The initial statement of this excerpt reinforces the understanding that true polyphony is not reduced to a syntactic or stylistic juxtaposition of distinct speeches but depends essentially on independent and axiologically integral discursive consciousnesses.

Moreover, the application of polyphony as a criterion for evaluating authorship in the context of artificial intelligences involves a qualitative analysis of the internal discursive relations in the produced text, allowing the precise identification of the absence or presence of true discursive, dialogical and axiological alterity, elements indispensable to genuine authorship.

The ego ideal is, then, immensely powerful, and would seem to work much as the otherness of society does in Bakhtin. But Freud stops short of directly relating the ego ideal to social force [...]. By emphasizing leaders, Freud is psychologizing impersonal social force. Freud makes a significant contribution to understanding the way in which individual people are interpenetrated by group ideals. But what is lacking in his account of relations between individual and group is any concept of the means by which such relations are mediated: the scenario for interaction is over-personalized [...] By what means do the internalized phantoms communicate with the real actors? In what language does the internalized family romance play itself out in the theater of our minds?

Dialogism’s answer to this question is that it is the same language as we use in the theater of society. The group as a collective and the individual members who comprise it all share the same language [...]. The essential clash between what Freud calls ego and ego ideal is not an abstract tension between the former’s libidinal spontaneity and society’s introjected demands in the latter. This description is adequate so far as it goes, but it does not go far enough; what needs to be added is that this opposition works itself out in highly specific (and specifiable) circumstances as a conflict between inner and outer speech. From this assumption flows Bakhtin’s argument for a necessary link between the study of psychology and the study of language (Holquist, 2002, p. 52).

This fragment illustrates how, in Bakhtin, one can only speak of genuine authorship if there is room for an effective alterity. In the critique of Freud, it is observed that the internalized social force (the ‘ego ideal’) needs a means of expression -the languagethat is socially shared, showing that subjectivity is not formed in isolation but in real interaction with the other and with the repertoire of collective discourses. Thus, when this alterity (the dialogical dimension in the Bakhtinian sense) is absent or purely fictional, authorship does not legitimately and responsively occur.

In other words, both the Turing test and the Chinese room argument remain anchored in a functionalist or representational perspective, in which the question of intelligence is restricted to behavioral simulation or syntactic processing devoid of intentionality. Although ambitious, Kurzweil’s conjectures also operate within a tiered and ontologically diffuse logic, centered on consciousness, bypassing the classic Freudian conception. In contrast, Bakhtin’s conception of polyphony intervenes decisively, shifting the focus to the dialogical constitution of the discursive subject and to the ethical-enunciative conditions that underpin its authorial agency.

Thus, the distinction established is between the act of speaking or responding like a human and the actual ability to understand what is being said. What is at stake is the difference between appearance and essence: even if the machine presents impeccable linguistic performance, it does not necessarily mean that there is any degree of subjective experience or intentionality behind its responses. It becomes evident that cognition is not reduced to an automatic reaction to external stimuli, but it consists of a complex process of integration in which experiences, context, and intentionality intertwine dynamically.

The absence of an internal living experience-a capacity to feel or even to attribute meaning to the information processed-makes the production of AI qualitatively different from human authorship. At most, one can recognize an external functionality that simulates our capacities, but it is deprived of inner experience.

Final Considerations

The trajectory of authorship reveals a process intrinsically linked to historical transformations in the modes of subject constitution and to the practices of appropriation, circulation, and interpretation of texts. In this context, Bakhtin’s notion of polyphony, understood in its expanded sense as a dialogic paradigm, allows us to problematize the limits of generative artificial intelligence. Although Bakhtin related polyphony, strictly speaking, to the novel as a space of autonomous voices, later developments configured it as an epistemological instrument that encompasses dialogism in its totality, applicable to cultural contexts that do not necessarily presuppose the presence of polyphonic literary texts. In the revised version of his poetics, Bakhtin articulates dialogism with polyphony as a constitutive realization, which enables its extension to non-novelistic domains. From periods when knowledge was shared as collective patrimony to the modern consolidation of the authorial figure as a singular instance of creativity and meaning-making, the idea of authorship oscillates between regimes of anonymity, collaborative practices, and the emergence of artistic subjectivity as the core of individualized expression, highlighting the absence of an effective dialogical dynamic in artificial productions.

In Eastern traditions, the permeability between singular authorship and communal production demonstrates that literary and philosophical creation is often anchored in a cumulative, ritualized, and reiterated transmission over time. Works such as the Analects of Confucius, the Buddhist Sutras, or the Hindu epics express an authorial dynamic deeply linked to shared cultural memory. The presence of the creative subject blends with the interpretative continuity of generations, indicating that intentionality and authorship are not absent, but operate in forms distinct from those consolidated in the modern Western paradigm. The figure of the author is not annulled, but rather made more complex by the logic of rewriting, commentary, and communal updating that produces a type of diffuse, yet still responsive, authorship.

Contemporary literary theory has deepened this inflection by destabilizing the centrality of authorial intention as the meaning exclusive source. Harold Bloom introduced the notion of a structural anxiety in the face of the legacy of predecessors, while Roland Barthes shifted the origin of meaning from the author to the reader, dissolving the authorial figure as a univocal instance of semantic authority.

It is also necessary to consider that the discussion regarding the death of the author acquires greater conceptual scope when Michel Foucault’s reflection is included, for whom the author does not correspond to a fixed empirical figure but rather to a function arising from specific discursive regimes. In delivering the lecture, “What is an Author?,” Foucault stated: “Nevertheless, as a privileged moment of individualization in the history of ideas, knowledge, and literature, or in the history of philosophy and science, the question of the author demands a more direct response” (Foucault, 1998, p. 205).²⁷

The author thus emerges as a historical operator who individualizes discourses, establishing cuts amid knowledge practices that were previously more diffuse or collective. This conception alters the perspective broadened by Barthes, for whom “writing is the destruction of every voice, of every point of origin. Writing is that neutral, composite, oblique space where our subject slips away, the negative where all identity is lost, starting with the very identity of the body writing” (Barthes, 1977, p. 142).²⁸

Barthes suspends biographical centrality; Foucault details the institutional device that replaces such centrality, indicating that the author’s name demarcates “a certain number of discourses endowed with the ‘author-function’ while others are deprived of it” (Foucault, 1998, p. 211).²⁹ Therefore, the signature functions as a classifying principle that enables texts to circulate in circuits of legitimation, establishing boundaries between ordinary production and production subject to a specialized hermeneutical regime.

The relevance of this notion becomes even more evident when Foucault distinguishes the traditional author from the “founders of discursivity” (Foucault, 1998, p. 217).³⁰ According to him, such thinkers “are unique in that they are not just the authors of their own works. They have produced something else: the possibilities and the rules for the formation of other texts” (Foucault, 1998, p. 217).³¹ This differentiation is close to the dilemma formulated by Harold Bloom regarding the anxiety of influence, since it indicates that certain writers - seen as founders - establish textual matrices that reshape the horizon of subsequent creation. Bloom’s dialectic of the precursor and the epigone thus find a Foucauldian counterpart: the author-function evolves into a founding function when it creates rules of production capable of engendering entire critical communities.

In the debate about generative artificial intelligences, Foucault’s author-function offers a criterion distinct from that provided by mere linguistic performance evaluated by the Turing test. Lacking an institutional position in which its name produces cuts, responsibilities, and reading protocols, the language model remains external to the Foucauldian logic of the author. Even if the machine text appears coherent, it lacks the social inscription conferred by the name that establishes corpus limits, regulates attributions, and, above all, links discourse and responsibility.

Mikhail Bakhtin, in proposing the concept of polyphony, goes beyond the paradigm of passive intertextuality and affirms the existence of multiple autonomous voices that coexist in the same text, each carrying a unique axiological perspective. Originality, within this model, does not reside in the rupture with the past, but in the capacity to articulate a dialogical constellation of voices and consciousnesses, in which the author-creator acts as an orchestrator of alterities, and not as a sovereign emitter of meanings.

This conception implies overcoming models centered on the monologic author, since the work reveals itself, in its aesthetic and epistemic complexity, as an arena of confrontation between worldviews, between social languages and ideological voices in dispute. Polyphony, as formulated by Bakhtin, requires that the text should contain multiple speeches that represent effectively independent discursive consciousnesses, not subordinated to the authorial point of view.

The emergence of generative artificial intelligence systems, capable of producing texts with a high degree of coherence and stylization, repositions the discussion on new ground, but not immune to the categories inherited from philosophy and literary theory. Contemporary computational models, by passing tests such as Turing’s with ease, can imitate the modes of discursive organization proper to human language. However, philosophical criticism, as represented by John Searle in his Chinese Room argument, reveals that the ability to manipulate signs according to syntactic rules is not equivalent to possessing semantic understanding or the presence of a conscious subject.

This distinction relates to the limitations of computational functionalism, which, by reducing intelligence to the capacity to process information, ignores the experiential and phenomenological dimension of consciousness. The absence of a first-person point of view, of an interiority that relates to the world in a valued and situated way, compromises the possibility of attributing any degree of authentic intentionality to AI systems. Although machine-generated statements may exhibit structural complexity and contextual adaptation, they lack the ethical-enunciative component that characterizes human discourse: there is no “I” that responds, that assumes responsibility, that inscribes its otherness in the fabric of language.

It is precisely at this point that Bakhtin’s theory of polyphony offers the most demanding and epistemologically refined criterion for evaluating authorship. The textual production of an AI, even if well-operationalized and capable of simulating voices, does not achieve the polyphonic condition in the full sense of the term, because the voices it presents are not endowed with axiological autonomy. There is no plurality of consciousnesses in responsive interaction within the model’s structure. The voices simulated by these systems do not possess their own horizon of meaning, nor do they emerge from specific ethical-discursive positions. Rather, it is a statistical reproduction of patterns, lacking the genuine dialogic that underlies authorship according to Bakhtin.

Generative artificial intelligence systems, especially those based on autoregressive language models, such as GPT (Generative Pre-trained Transformer),³² fundamentally operate by applying probabilistic processes to large sets of previously accumulated linguistic data. This statistical operation necessarily implies the capture and reproduction of dominant discursive patterns in these sets, leading to the emergence of a text representative not of the authentic multiplicity of voices, but rather of a kind of average or normalization of discourse. Consequently, contrary to Bakhtin’s ideal of polyphony, generative systems end up, unintentionally but structurally inevitably, promoting the silencing of marginal, unusual, or less frequent voices in their training databases. Thus, while superficially they seem to express textual and discursive diversity, such systems reproduce what can be described as a statistical majority. In discursive terms, such regularities are equivalent to the most recurrent and conventional forms of linguistic usage, which frequently correspond to the culturally hegemonic, socially majority, and ideologically dominant perspectives in the textual base used in the training.

Literary or philosophical creation, engendered by a human subject, is inseparable from an existential hexis that projects the text onto the historical horizon of values and responsibility. The author appears as an enunciative consciousness capable of removing each sign from lexical neutrality, investing it with axiological intentionality that inscribes it in a hermeneutic field saturated with meaning. This responsive gesture establishes authorship as agency, never as mere combinatorial concatenation. Even if the work incorporates stylistic polyphonies or mimics heteroglossies, it is the dialogical dynamic established between these voices that brings together such multiplicity into a coherent whole, guided by a shared ethical imperative, without resorting to the centrality of a univocal authorial subject. Generative algorithms, although operating with formidable statistical complexity, remain detached from any experiential economy: they record without experiencing, reiterate without responding and compose without assuming.

It follows that the category of authorship preserves an unbreakable link with the phenomenological experience of the subject and with the ability to establish effective dialogical alterity. The machine configuration does not host such alterity in the Bakhtinian sense, since it lacks a consciousness capable of occupying a responsible discursive place before the other. Texts produced by artificial intelligence qualify, at best, as simulational artifacts of language, devoid of the ontological foundations essential to the authorial voice. The absence of experiential pathos converts these statements into sophisticated imitations, revealing the insurmountable limit between logo technical engineering and authorship understood as an ethical-dialogical event.

• 1
  The game, currently known as the “Turing test,” was named in honor of the British mathematician Alan Mathison Turing.
• 2
  ONG, Walter Jackson. Orality and Literacy: The Technologizing of the Word. 2. ed. London: Routledge, 2002.
• 3
  CROCE, Benedetto. Aesthetic as Science of Expression and General Linguistic. Translated by Douglas Ainslie. 2nd ed. London: Macmillan, 1922.
• 4
  BURCKHARDT, Jacob. The Civilization of the Renaissance in Italy. Translated by S. G. C. Middlemore. London: George Allen & Unwin Ltd.; New York: The Macmillan Company, 2014.
• 5
  For reference, see footnote 4.
• 6
  CONFUCIUS. The Analects of Confucius. Translated by Simon Leys. New York: W. W. Norton & Company, 1997.
• 7
  MURASAKI SHIKIBU. The Tale of Genji: The Authentic First Translation of the World's Earliest Novel. Translated by Kencho Suematsu. Preface by Michael Emmerich. [S. l.]: Tuttle Publishing, 2018.
• 8
  BLOOM, Harold. The Anxiety of Influence: A Theory of Poetry. New York: Oxford University Press, 1973. p. 15
• 9
  BARTHES, Roland. The Death of the Author. In: BARTHES, Roland. Image-Music-Text. Essays selected and translated by Stephen Heath. London: Fontana Press, 1977. pp. 142-148.
• 10
  For reference, see footnote 9.
• 11
  BAKHTIN, Mikhail. Discourse in the Novel. In: BAKHTIN, Mikhail Mikhailovich. The Dialogic Imagination: Four Essays. Edited by Michael Holquist. Translated by Caryl Emerson and Michael Holquist. Austin: University of Texas Press, 1981. pp. 259-422.
• 12
  BAKHTIN, Mikhail. Problems of Dostoevsky’s Poetics. Edited and translated by Caryl Emerson. Minneapolis: University of Minnesota Press, 1984. p. 88.
• 13
  MARCHEZAN, Renata Coelho. The Notion of Author in and from M. Bakhtin’s Work. Bakhtiniana: Revista de Estudos do Discurso, São Paulo, v. 10, n. 3, pp. 205-224, Sep./Dec. 2015. Available at: https://revistas.pucsp.br/index.php/bakhtiniana/article/view/22365/17706 . Accessed on: July 23, 2026.
• 14
  BAKHTIN, Mikhail. The Problem of the Text in Linguistics, Philology, and the Human Sciences: An Experiment in Philosophical Analysis. In: BAKHTIN, Mikhail M. Speech Genres and Other Late Essays. Edited by Caryl Emerson; Michael Holquist. Translated by Vern W. McGee. Austin: University of Texas Press, 1986. pp. 103-131.
• 15
  For reference, see footnote 14.
• 16
  For reference, see footnote 14.
• 17
  BARTHES, Roland. The Rustle of Language. Translated by Richard Howard. New York: Hill and Wang, 1986. p. 54.
• 18
  BENJAMIN, Walter. The Work of Art in the Age of Mechanical Reproduction. In: BENJAMIN, Walter. Illuminations. Edited by Hannah Arendt. Translated by Harry Zohn. New York: Schocken Books, 1969. pp. 223-224.
• 19
  KURZWEIL, Ray. The Singularity Is Near: When Humans Transcend Biology. New York: Viking, 2005.
• 20
  A human evaluator interacts, through a text-based interface, with two hidden interlocutors: a human and a machine. If the evaluator cannot reliably determine which of the two is the machine, the machine is considered to have passed the test. Although it is subject to criticism, the test remains a central reference in debates on artificial intelligence and the philosophy of mind, because it shifts the focus from the essence of intelligence to its observable manifestation.
• 21
  A person inside a room, who does not speak Chinese, follows instructions from a rulebook to manipulate characters in the language so as to answer questions written in Chinese correctly. Although they may appear to understand the language to someone on the outside, the individual is, in reality, merely correlating input symbols with output symbols, without any semantic understanding whatsoever. In the analogy with computers, the system would be capable of processing information syntactically (according to rules), but it would have no consciousness or understanding of the content it manipulates. In this way, indistinguishability in linguistic behavior can be regarded as only a superficial facet of intelligence.
• 22
  For reference, see footnote 19.
• 23
  For reference, see footnote 12.
• 24
  For reference, see footnote 12.
• 25
  BAKHTIN, Mikhail M. The Problem of Speech Genres. In: BAKHTIN, Mikhail M. Speech Genres and Other Late Essays. Edited by Caryl Emerson; Michael Holquist. Translated by Vern W. McGee. Austin: University of Texas Press, 1986. pp. 60-102.
• 26
  For reference, see footnote 8.
• 27
  FOUCAULT, Michel. What Is an Author? In: FOUCAULT, Michel. Aesthetics, Method, and Epistemology. Edited by James D. Faubion. Translated by Robert Hurley and others. New York: The New Press, 1998. pp. 205-222.
• 28
  For reference, see footnote 9.
• 29
  For reference, see footnote 27.
• 30
  For reference, see footnote 27.
• 31
  For reference, see footnote 27.
• 32
  The probabilistic approach underlying generative artificial intelligence consists in the statistical modeling of text, grounded in calculations of conditional probability, whereby the generation of each new token (word, symbol, or minimal linguistic unit) depends directly on probabilities derived from vast quantities of previously assimilated data. Models based on Transformer architectures, for example, construct their textual outputs by maximizing conditional probabilities, with the direct consequence that uncommon or divergent linguistic expressions - whether linguistically or culturally minoritarian - are assigned comparatively lower probabilities and thus become underrepresented or entirely excluded from the generated discourse.
• Translated by Lígia Razera Gallo - [e-mail]
• Reviews
  Due to the commitment assumed by Bakhtiniana. Revista de Estudos do Discurso [Bakhtiniana. Journal of Discourse Studies] to Open Science, this journal only publishes reviews that have been authorized by all involved.

Research Data and Other Materials Availability

The contents underlying the research text are included in the manuscript.

References

• BAKHTIN, Mikhail. O discurso no romance. In: BAKHTIN, Mikhail. Questões de literatura e de estética: a teoria do romance. Tradução de Aurora F. Bernardini, José Pereira Júnior, Augusto Góes Júnior, Helena S. Nazário, Homero F. de Andrade. 3. ed. São Paulo: Editora UNESP; Hucitec, 1993. pp. 71-210.
• BAKHTIN, Mikhail. Problemas da poética de Dostoiévski 3. ed. Tradução direta do russo, notas e prefácio de Paulo Bezerra. Rio de Janeiro: Forense Universitária, 2002.
• BAKHTIN, Mikhail. Os gêneros do discurso. In: BAKHTIN, Mikhail. Estética da criação verbal Introdução e tradução do russo de Paulo Bezerra. 4. ed. São Paulo: Martins Fontes, 2003. pp. 262-306.
• BAKHTIN, Mikhail. O problema do texto na linguística, na filologia e em outras ciências humanas: uma experiência de análise filosófica. In: BAKHTIN, Mikhail. Estética da criação verbal Introdução e tradução do russo de Paulo Bezerra. 4. ed. São Paulo: Martins Fontes, 2003. pp. 307-335.
• BARTHES, Roland. A morte do autor. In: BARTHES, Roland. O rumor da língua Tradução de Mário Laranjeira. São Paulo: Brasiliense, 1988. pp. 65-70.
• BENJAMIN, Walter. A obra de arte na era de sua reprodutibilidade técnica. In: BENJAMIN, Walter. Magia e técnica, arte e política: ensaios sobre literatura e história da cultura. Tradução de Sérgio Paulo Rouanet. Prefácio de Jeanne Marie Gagnebin. 3. ed. São Paulo: Brasiliense, 1987. pp. 165-196.
• BLOOM, Harold. A angústia da influência: uma teoria da poesia. Tradução de Marcos Santarrita. 2. ed. Rio de Janeiro: Imago, 2002.
• BRENTANO, Franz. Psychology from an Empirical Standpoint Edição de Oskar Kraus e Linda L. McAlister. Trans. Antos C. Rancurello, D. B. Terrell & Linda L. McAlister. 2. ed. London: Routledge, 1995.
• BURCKHARDT, Jacob. A cultura do Renascimento na Itália: um ensaio. Tradução de Vera Lúcia de Oliveira Sarmento e Fernando de Azevedo Corrêa. Brasília: Editora Universidade de Brasília, 1991.
• CONFÚCIO. Os Analectos Tradução do inglês Caroline Chang; tradução do chinês, introdução e notas D. C. Lau. Porto Alegre: L&PM, 2023.
• CROCE, Benedetto. Estética como ciência da expressão e linguística geral: teoria e história. Organização de Giuseppe Galasso. Tradução de Omayr José de Moraes Júnior. São Paulo: É Realizações, 2016.
• DENNETT, Daniel Clement. The Intentional Stance Cambridge, MA: MIT Press, 1989.
• FOUCAULT, Michel. O que é um autor? In: FOUCAULT, Michel. Ditos & escritos III: estética: literatura e pintura, música e cinema. Organização e seleção de textos de Manoel Barros da Motta. Tradução de Inês Autran Dourado Barbosa. 2. ed. Rio de Janeiro: Forense Universitária, 2009. pp. 264-298.
• FREUD, Sigmund. Obras completas, volume 4: A interpretação dos sonhos (1900). Tradução de Paulo César de Souza. 1. ed. São Paulo: Companhia das Letras, 2019.
• GOODFELLOW, Ian J.; POUGET-ABADIE, Jean; MIRZA, Mehdi; XU, Bing; WARDE-FARLEY, David; OZAIR, Sherjil; COURVILLE, Aaron C.; BENGIO, Yoshua. Generative adversarial networks. Communications of the ACM, [S. l.], v. 63, n. 11, pp. 139-144, nov. 2020. DOI: 10.1145/3422622 Disponível em: https://doi.org/10.1145/3422622 Acesso em: 23 jul. 2026.
  » https://doi.org/10.1145/3422622» https://doi.org/10.1145/3422622
• HOLQUIST, Michael. Dialogism: Bakhtin and His World. 2. ed. New York: Routledge, 2002.
• HUSSERL, Edmund. Investigações lógicas: prolegômenos à lógica pura. Tradução de Diogo Ferrer. Rio de Janeiro: Forense, 2014. v. 1.
• KURZWEIL, Ray. A singularidade está próxima: quando os humanos transcendem a biologia. Tradução de Ana Goldberger. São Paulo: Iluminuras, 2018.
• MANDELKER, Amy (org.). Bakhtin in Contexts: Across the Disciplines. Introduction Caryl Emerson. Evanston: Northwestern University Press, 1995.
• MARCHEZAN, Renata Coelho. A noção de autor na obra de M. Bakhtin e a partir dela. Bakhtiniana: Revista de Estudos do Discurso, São Paulo, v. 10, n. 3, pp. 186-204, set./dez. 2015. DOI: 10.1590/2176-457322365 Disponível em: https://doi.org/10.1590/2176-457322365 Acesso em: 23 jul. 2026.
  » https://doi.org/10.1590/2176-457322365» https://doi.org/10.1590/2176-457322365
• MORSON, Gary Saul; EMERSON, Caryl. Introduction: Rethinking Bakhtin. In: MORSON, Gary Saul; EMERSON, Caryl. Rethinking Bakhtin: Extensions and Challenges. Evanston: Northwestern University Press, 1989. pp. 1-60.
• ONG, Walter Jackson. Oralidade e cultura escrita: a tecnologização da palavra. Tradução de Enid Abreu Dobránszky. Campinas, SP: Papirus, 1998.
• RAJAGOPALACHARI, Chakravarti. Mahabharata 46th ed. Mumbai: Bharatiya Vidya Bhavan, 2005.
• RAJAGOPALACHARI, Chakravarti. Ramayana 53rd ed. Mumbai: Bharatiya Vidya Bhavan, 2013.
• SIMA, Qian. The First Emperor: Selections from the Historical Records. Trans. Raymond Dawson. Edição e introdução de K. E. Brashier. Oxford: Oxford University Press, 2009.
• SEARLE, John Rogers. Minds, Brains, and Programs. Behavioral and Brain Sciences, [S. l.], v. 3, n. 3, pp. 417-424, set. 1980. DOI: 10.1017/S0140525X00005756 Disponível em: https://doi.org/10.1017/S0140525X00005756 Acesso em: 23 jul. 2026.
  » https://doi.org/10.1017/S0140525X00005756» https://doi.org/10.1017/S0140525X00005756
• SHIKIBU, Murasaki. O conto de Genji Tradução de Autri Books. [S. l.]: Autri Books, 2026. E-book.
• TURING, Alan Mathison. Computing Machinery and Intelligence. Mind, [S. l.], v. 59, n. 236, pp. 433-460, out. 1950. DOI: 10.1093/mind/LIX.236.433 Disponível em: https://doi.org/10.1093/mind/LIX.236.433 Acesso em: 23 jul. 2026.
  » https://doi.org/10.1093/mind/LIX.236.433» https://doi.org/10.1093/mind/LIX.236.433
• VASWANI, Ashish; SHAZEER, Noam; PARMAR, Niki; USZKOREIT, Jakob; JONES, Llion; GOMEZ, Aidan N.; KAISER, Łukasz; POLOSUKHIN, Illia. Attention Is All You Need. In: CONFERENCE ON NEURAL INFORMATION PROCESSING SYSTEMS, 31., 2017, Long Beach. Advances in Neural Information Processing Systems 30 Red Hook: Curran Associates, 2017. pp. 5998-6008. Disponível em: https://proceedings.neurips.cc/paper_files/paper/2017/hash/3f5ee243547dee91fbd053c1c4a845aa-Abstract.html Acesso em: 23 jul. 2026.
  » https://proceedings.neurips.cc/paper_files/paper/2017/hash/3f5ee243547dee91fbd053c1c4a845aa-Abstract.html

• Editors in Charge
  Luciana Salazar Salgado (Universidade Federal de São Carlos - UFSCar)
  Lucia Santaella (Pontifícia Universidade Católica de São Paulo - PUC-SP)
  Tony Berber Sardinha (Pontifícia Universidade Católica de São Paulo - PUC-SP)

Review I

About the reviewerSCIMAGO INSTITUTIONS RANKINGS

Review I

The article offers a dense, original, and theoretically significant reflection on the limits of authorship in texts produced by artificial intelligence, grounded in a Bakhtinian epistemology of polyphony. The title is precise and accurately reflects the article’s content, clearly identifying both its central theoretical framework (Bakhtin and polyphony) and its object of analysis (authorship in AI-generated texts). Its aim is clearly stated: to examine the extent to which textual production by AI systems may or may not amount to full authorship, on the basis of the concept of polyphony. The article pursues this objective with considerable coherence, articulating theoretical arguments, historical references, and pertinent critiques of classical approaches, such as the Turing Test and the Chinese Room argument. The bibliography is broad, current, and highly relevant, including contemporary contributions from computer science and philosophy of mind. The discussion carefully brings together authors from different fields without sacrificing conceptual depth. The author proposes a theoretical shift from a technical-functional analysis of AI to an ethical-discursive criterion of authorship, which constitutes a significant contribution to the interdisciplinary debate. The proposal to apply Bakhtinian polyphony as a criterion for delimiting authorship in AI-generated texts is innovative, and the argument is developed with scholarly rigor, especially within the humanities. The article displays both argumentative density and stylistic sophistication, without compromising intelligibility for specialized readers. It is recommended only that certain lengthy passages be revised and reorganized in order to improve the flow of reading, without any loss of substance. The article might also usefully address, in its concluding remarks, the ethical and political implications of the substitution or simulation of authorship in educational and editorial contexts. ACCEPTED WITH SUGGESTIONS [Revised]

• peer review recommendation: accept

History

• Peer review received
  29 Apr 2025

Review II

About the reviewerSCIMAGO INSTITUTIONS RANKINGS

Review II

Statement of the article’s objective and coherence in its development throughout the text. With the aim of reflecting on the thought-provoking phenomenon of textual production by contemporary artificial intelligence systems-which would readily pass tests of imitative and controlled linguistic performance, such as the so-called Turing Test developed by the computer scientist Alan Mathison Turing-the article proposes, so to speak, a test of authorship. It frames the issue within the domain of the notion of the author, which, it should be recalled, also encompasses a corresponding notion of text. To this end, the article traces the historical constitution of the notion of authorship, pointing to its various political, religious, literary, philosophical, and technological transformations. With particular emphasis on the philosophical and literary domains, it highlights M. Bakhtin’s reflections on the author and his concepts of dialogism and polyphony. The article details the dialogic and axiological dimensions of this notion of authorship and argues that an analysis grounded in this perspective would reveal the absence of authorship in texts generated by artificial intelligence. The article thus addresses an engaging topic and presents a strong line of argumentation; for this reason, we recommend its publication. We nevertheless offer two recommendations. First, the article would benefit from the inclusion of a brief introductory passage incorporating into the main body of the text its objective, which is currently stated only in the abstract. Second, a more substantial adjustment is needed-possibly even a change in the article’s title-namely, a refinement of the notion of polyphony, so that acknowledging the dialogical dimension of authorship does not also imply requiring the production of a polyphonic text. It should be recalled that, in Bakhtin, polyphony is a defining feature of the novel as a genre, especially in the case of Dostoevsky’s novels. ACCEPTED WITH RESTRICTIONS [Revised]

• peer review recommendation: accept

History

• Peer review received
  08 Aug 2025

Review III

About the reviewerSCIMAGO INSTITUTIONS RANKINGS

Review III

In the initial review, we recommended publication subject to two revisions. The final version satisfactorily addresses the first recommendation through the inclusion of two introductory paragraphs and, in its own way, also responds to the second by circumventing the issue previously raised and defending-through an argument that now incorporates Mandelker and Morson-an expanded notion of polyphony. We reaffirm both the relevance of the article and the merit of its publication. ACCEPTED

• peer review recommendation: accept

History

• Peer review received
  11 Oct 2025

Publication Dates

• Publication in this collection
  21 Aug 2026
• Date of issue
  2026

History

• Received
  30 Mar 2025
• Accepted
  06 Aug 2026