Headnote
Abstract: After the GPT-3 boom, there is an ongoing debate regarding the language heuristics in automatic text generation, as well as their usefulness for scientific inquiry. In this article, we conduct a computational study that engages in a discursive problematization of textual analysis methods, seeking to reconcile epistemological tensions between positivist and interpretive paradigms in Psychology, Literature, and Computer Science. For our "experiment", (a) a pseudo-random probabilistic language model is trained on a classical narrative works containing epic and heroic traits; (b) a stochastic text is generated using n-gram modeling in Python; (c) the output text is interpreted using Analytical Psychology and Literary Criticism; (d) an integrative discussion reconciles the previous steps, suggesting a new methodological approach for ideation and theorizing in scientific endeavors. We conclude the study optimistically, highlighting the benefits of this mode of inquiry for social psychology.
Keywords: Interdisciplinary research; Psychology; Literature, Computational heuristics; Methodology.
Resumo: Após o boom do GPT-3, há um debate em andamento sobre as heurísticas da linguagem na geração automática de texto, bem como sua utilidade para a investigação científica. Neste artigo, conduzimos um estudo computacional que se dedica a uma problematização discursiva de métodos de análise textual, buscando reconciliar tensões epistemológicas entre paradigmas positivistas e interpretativos em Psicologia, Literatura e Ciência da Computação. Para nosso "experimento", (a) um modelo de linguagem probabilística pseudoaleatória é treinado em obras narrativas clássicas contendo traços épicos e heróicos; (b) um texto estocástico é gerado usando modelagem x-gram em Python; (c) o texto de saída é interpretado usando Psicologia Analítica e Crítica Literária; (d) uma discussão integrativa reconcilia as etapas anteriores, sugerindo uma nova abordagem metodológica para ideação e teorização em empreendimentos científicos. Concluímos o estudo com otimismo, destacando os beneficios desse modo de investigação para a psicologia social.
Palavras-chave: Pesquisa interdisciplinar; Psicologia; Literatura; Heuristica computacional; Metodologia.
Resumen: Tras el auge de GPT-3, existe un debate en curso sobre la heuristica lingüística en la generación automática de textos, así como su utilidad para la investigación cientifica. En este artículo, realizamos un estudio computacional que aborda la problematización discursiva de los métodos de análisis textual, buscando conciliar las tensiones epistemológicas entre los paradigmas positivistas e interpretativos en Psicología, Literatura e Informática. Para nuestro experimento, (a) se entrena un modelo de lenguaje probabilístico pseudoaleatorio con obras narrativas clásicas que contienen características épicas y heroicas; (b) se genera un texto estocástico mediante el modelado de n-gramas en Python; (c) el texto resultante se interpreta mediante Psicología Analítica y Crítica Literaria; (d) una discusión integradora concilia los pasos anteriores, sugiriendo un nuevo enfoque metodológico para la idealización y la teorización en el quehacer científico. Concluimos el estudio con optimismo, destacando los beneficios de este modo de investigación para la psicología social.
Palabras clave: Investigación interdisciplinaria; Psicología; Literatura; Heurística informática; Metodología.|
l. Introduction
Despite the recent surge in generative AI- epitomized by ChatGPT's "i.7 billion visits in December 2024 (Duarte, 2025) machines that probabilistically recombme language are hardly new. In the 1970s scholars were already benchmarking proto-LLAIs for narrative cre-ativity, textual aesthetics, and semantic coherence (Ryan, I9S7). What is unprecedented today is the penetration of such models beyond Computer Science into the Humanities and Social Sciences, where they intervene in meaningmalang practices central to many branches of Psychology (Nelson, 2020; Sarker. 2021).
Within contemporary socialpsychological debate, algorithms are increasingly framed as coparticipants m the negotiation of identity, persuasion, and collective imagination (Brady & Crockett, 202·; Rahwan et al., 2019). Outputs once dismissed as "mere statistics" now circulate as cultural artifacts, subtly transforming how groups deliberate, remember, and assign value (Beer, 2017). This circulation reopens classic psychosocial questions-How are social representations formed? What counts as agency?-while demanding methodological fluency capable of moving among computational modeling, critical hermeneutics. and theoretical reflection. Unfortunately, the broader discussion still oscillates between alarm ("a threat to creativity") and technoeuphoria ("algorithmic salvation") (Tamkin et al., 2021), obscuring the nuanced ways in which people and machines already coproduce meaning (Jovchelo\it-ch, 2007).
The present article situates itself deliberately between those poles. Rather than claim to resolve current controversies, it introduces a procedural scaffold-a chained generationinter-pretation loop-designed to provide future studies with clearer analytic traction. More than half a century ago, Bachelard (1964) argued that philosophy's higher task is to unite poetry and science. We pursue that ambition pragmatically by staging a dialogue m which probabilistic metrics converse with symbolic interpretation. Each computational step yields a fragment of machine genera ted narrative that we read through archetypal lenses derived from Carl Jung (l9S0b) and Northrop Frye (2000): the model proposes, the analyst responds, theorizes, and feeds the dialogue forward. Such alternating turns echo Mikhail Bakhtin's (19S1) chain of answerability and Hubert Hermans' (2001) dialogical self, where knowledge crystallizes through negotiations among shifting Ipositions, interlocutors, and sociocultural artifacts-including algorithms treated as quasiagents (Dennett, 19S7).
Our methodological stance also challenges the territorial logic of academic "fiefdoms." If disciplines are enclosed domains (Weingart, 201S), lnterdisciplinarity provides the bridges that span them (Repko et al., 2011). Wicked problems and emergent technologies oblige us to cross those bridges (Peters, 2017; Smaldino & O'Connor, 2020), even as quantitative and qualitative camps guard their gates (Foucault, 2019; Ignatow Sz Mihalcea, 201S). The chained computationalterpretation strategy operationalizes this crossing in abductive theorizing (Nelson, 2020). Thus, technical and philosophical voices do not driftapart; they braid into a single argumentative line that keeps subjective meanwgmaking in view.
For our empirical illustration, we fmetuned a transformer on books with epic traits- Dante, Homer, Tolkien-whose crosstemporal appeal suggests a reservoir of resilient archetypes (Jung, 19S0b). Remixed under probabilistic constraints, this reservoir allows us to observe how cultural memory can be reorganized and how subjective connotations migrate across media. Our reading detects strong echoes of anuria and shadow; once re-inscribed by a machine, these figures illustrate how gendered anchoring and moral threat can surface as researchable hypotheses (Brady & Crockett, 2024).
We conclude by stating that even a socalled "stochastic parrot" (Bender et al., 2021) can catalyze disciplined ideation (Bjork et al., 2010) and pro%isional theorizing (Gregory & Hen-fridsson, 2021) when placed in a dialogical circuit attentive to discourse and subjectivity. What follows, therefore, is a worked example of methodological fluency-a provisional bridge for future crossings among Psychology, Computer Science, and Literary Analysis-rather than a finished theory of how humans and machines coconstruct social reality.
2. Literature Review
2.1. Archetypes
The analytical psychology of Carl Jung and some literary theorists, such as the critic Northrop Frye and the philosopher Gaston Bachelard, worked with the idea of the archetype. The study of the archetype "is derived from the repeated observation that ... the myths and fairytales of world literature contain definite motifs which crop up everywhere" (Jung, 197S, p. 449). These typical images that appear m different forms of fiction are archetypal ideas. While conceptually distinct, the recurrence of archetypal figures in myth and literature bears somehow a resemblance to what social-representation theory describes as "thematic kernels" - recurring symbolic cores that help communities structure shared meaning (Moloney, Hall & Walker, 2005). Jung broadly conceived the archetype idea, documenting countless examples and giving various definitions at different tunes in his work. In the present text, the archetype will be understood as a pattern
that constantly recurs in the course of history and appears wherever creative fantasy is freely expressed. Essentially, therefore, it is a mythological figure When we examine these images more closely, we find that they give form to countless typical experiences ... They present a picture of psychic life in the average, divided up and projected mto the manifold figures of the mythological pantheon.....In each of these images there is a little piece of human psychology and human fate, a remnant of the joys and sorrows that have been repeated countless times in our ancestral history, and on the average follow ever the same course. (Jung, 1971, p. Si)
As a manifestation of the psychic life of human beings, an archetypal image can appear in different forms of art. Literature is an interesting field for the study of archetypes because it reproduces, through tradition or even without the author's intention, some narrative patterns which can also be understood as the "readiness to produce over and over again the same or similar mythical ideas" (Jung, 1972b, p. 69). Although drawn from a different theoretical tradition, such recurring motifs also recall what social-identity theorists describe as "identity markers" - symbolic elements that help groups construct, maintain, or reinterpret shared narratives of their collective past (Psaltis, 2012). Northrop Frye studies the archetype as a "symbol, usually an image, which recurs often enough m literature to be recognizable as an element of one's literary experience as a whole" (Frye, 2000, p. 365). Following Frye's ideas. it is possible to analyze a typical literary figure as an image that connects one work to another For example, we can isolate a sentence from the computer's automatically generated text to understand the method of literary analysis based on the theory of archetypes. The sentence created by the algorithm is: "the keavtn i wide circling' line" Now. we can compare the generated text with other works. In Dante's Divine Comedy, Paradise encompasses circles of celestial bodies. The circle is a shape symbolically associated with God. Analyzing Christian symbolism, Jung notes that God is "an all-embracing totality, which, like the definition of Godhead, is expressed iconographically by the circle" (Jung 1975, p. 155). In the short story The Library" of Babel, Jorge Luis Borges writes about a "circular book with a continuous spine that goes completely around the walls.....That cyclical book is God" (Borges, 1999a, p. 113). Furthermore, both Jung (1975) and Borges (1999b) mention the idea that God is a circle whose center is everywhere and the circumference nowhere. According to Northrop Frye's theory, it is possible to select an archetype - the circle, m this case - and observe as it spreads throughout the literary universe, undertaking a critical exercise that enriches the systematic view of literature.
In this study, a "human-algorithmic dialogue" is established, taking as a starting point the stochastic sentences generated from a LM, through the interpretation of experts. We believe that this can lead to new research insights, new analytical categories, and, to some extent, a synthesis of some of the most recurrent archetypal images of the whole corpus When carefully framed, such a dialogue offers the field of social psychology a complementary perspective on how enduring symbolic structures are being reconfigured in the context of Al-driven media ecology, contributing to broader discussions on discourse, representation, and meaning-making in the digital age.
2.2. Stochastic Parrots
"Stochastic parrot" is a term used in machine learning to describe a Large Language Model (LLM) capable of generating fluent and persuasive language without any semantic understanding of what it produces (Bender et al„ 2021). Although rooted in computational linguistics, the way users interpret and engage with such outputs raises questions that partially overlap with concerns in social psychology - particularly regarding attribution processes and the anthropomorphizing of artificial agents (Gamez-Djokic, Waytz, & Kouchalu, 2025). As these models grow in scope and are increasingly applied in high-stakes domains such as healthcare, finance, and public policy, their limitations invite scrutiny (O'Neil, 2016). Among the most pressing issues are the reproduction of hegemonic worldviews - including classist, sexist, and racist assumptions - the amplification of bias through automated textual patterns, and the opacity surrounding responsibility for potentially harmful content. From a discursive-psychological perspective, these outputs may resonate, in form if not in origin, with what Margaret Wetherell (2021) identifies as dominant interpretative repertoires: culturally embedded discursive frames that shape what can be said, by whom, and with what authority.
A notable portion of the disadvantages associated with stochastic parrots, however, arises from the fact that LLMs, like GPT-3, undergo training using extensive corpora that encompass a significant volume of web-based information (Dale, 2021). Hence, they primarily depend on fine-tuning techniques to tailor their performance to particular applications. The relative lack of control over the type of text entering the model also leads to a lack of control over its output. Most of the aforementioned issues with stochastic parrots dissipate when we provide more localized context and meaculously select the training corpus (Bender et al., 2021). Due to the smaller volume of information used in such cases, the generated sentences tend to be less convincing on the surface. However, this trade-offin surface fluency leads to a crucial advantage: increased reliability of the core information, which is often prioritized in scientific dialogues.
This increased reliability derives from the controlled nature of the training corpus. By curating a smaller and more focused dataset, we raise the "signal-tc-noise ratio" (Plomp, 19S6), reducing the influence of irrelevant or misleading information often present in large-scale, web-based corpora. This selective approach enables the model to generate outputs that are more consistent and contextual!;· aligned with the target domain. Although this may come at the cost of surface-level fluency, the trade-offis intentional. In a broader perspective, such domain-specific fine-tuning may gradually shape the narrative contours of algorithmic output, raising questions about how collective imaginary is structured or refracted in these constrained settings - a line of inquiry that loosely converges with framing research (Scheu-fele, Jamie son, Sz Brossard, 2021. in Social Psychology, while still demanding methodological tools tailored to the dynamics of generative AI models.
3. Method
3.1. Corpus Composition
A corpus is "a set of real textual data collected in a given language that serve as material for linguistic analysis" (Sardinha, 2000). The first stage of our study was the corpus composition, which is detailed in Chart 1.
The corpus documents are classic western books with easily identifiable epic or heroic traits. They are canons of epic-heroic literature, or in other words, considered genuine, legit, and quite representative of the genre in academia. -Although the selected works were originally written in many languages, we chose to analyze the English version of them to use the more advanced language techniques in Python. The corpus material was downloaded from Project Gutenberg digital book repository, with the exception of Tolkien's books, which were bought in digital format due to copyrights policy. The text pre-processing involved excluding the content of the back cover, summary, author's notes, page numbers, etc.
This curation step framed the corpus not simply as a dataset, but more as a symbolic resource - a repository of shared cultural material to be revisited and re-signified across analytical cycles. This treatment somehow echoes insights from textual treatment in social-representation theory concerning the role of textual artifacts in the stabilization and circulation of shared meanings (Jovchelovitch, 2007).
3.2. Python Programming
The automatic text generation algorithm was created using the Python language version 3.6.9. Much of the resources for text processing were already implemented on the Natural Language Toolkit (NLTK) library. Following Iyad Rahwan et al. (2019) call to treat AI systems as "algorithmic interlocutors," we cast this model as the quantitative voice in an alternating dialogue with human interpretation.
3.2.1. Rending Files and Pre-Processing
The book files were read individually and pre-processed to eliminate non-alphabetic characters such as numbers, punctuation marks and special symbols. Because we work with shallow syntax (Fedenci et al., 1996), this step was essential to avoid inappropriate punctuation or meaningless values in our textual output. We also converted all the characters to lowercase so that the interpreter does not consider different instances of the same word written in different forms.
3.2.2. Model Training and Text Generation
Text generation considers the probability of words. The method is based on a predictive model known as n-grarn, inspired by Claude Shannon's concepts of Information Theory (Grant, 2016). This model has been used in a variety of applications, such as DNA protein sequencing, or Google autocomplete feature on its search engines iNaiar & Renau, 2012). In short, the n-gram model aims to answer the question: "given a known set of n words, what is the likelihood of other words occurring?" To calculate the probability distribution that decides the next word, the model needs a representative training corpus (Sardinha, 2000) with good diversity. This condition was satisfied in our "experiment", as we selected at least a complete book of each author to train a subcorpus (see Chart 1..
In the n-gram model, we call words " tokens We do that because the algorithm does not understand the meaning of the words, they are only a sequence of characters. The "n" value in the n-gram model is fixed, and the algorithm always evaluates "n-l" tokens to predict the "nth" token. In the algorithm, by default, "n" equals three, meaning that we are working with trigrams. To make the n-gram concept clearer we can use a Markov· Chain (see Figure l).
A Markov chain (Revui, _00S) is a sequence of states (represented by circles) linked by transitions (represented by arrows). Every state, that is not an "end state", leads to a next one. In the example, given the two words "I am" there are three different possibilities. Since the \farkov chain is enriched with probabilities, a sentence can be more or less likely than others. The sentence "I am not a beaver," for instance, is more likely to occur than "I am a monkey" or "1 am not you". The sentences "I am Thomas" or "1 am so beautiful" are not possible m this example, because they did not occur m the training corpus. It is important to notice that at this stage, the model functions as an "abductive probe" (Nelson, 2020): its statistical regularities are not conclusions but prompts that invite the next interpretive turn.
In short, what our algorithm basically does is: (a) read a document; (b) take a couple of random words from this document; (c) calculate the possible triplets of words throughout the document (trigrams) and store them in a frequency table; id) given the couple of words, calculate the probability of occurrence of every possible third word; (e) draw the next word taking into account the probabilities; if: repeat steps C, D and E until reaching a final state; (g) when reaching final state, draw the next pair of initial words, increasing the random seed in the inference function, (h) repeat the process until it reaches the word limit pre-established by the "length" argument of the function; The same steps are valid to generate text usmg all the corpus documents, just replace "this document" for "all documents" in the previous instructions. Seen through a psychosocial lens, each generated sequence is a provisional hypothesis about how cultural material can recombine under probabilistic constraints, a hypothesis the analyst will test in the subsequent analysis.
3.3. Analyzing Archetypes
We analyzed the texts formed by the algorithm detailed above, always seeking intertex-tuakty and common narrative structures. The textual content generated from each author as well as the excerpt that somehow "synthesizes" the entire corpus were used in the archetypal analysis. The analysts identified archetypal figures that could be associated with the generated texts and interpretively established parallels. This stage constitutes the qualitative counter-move that somehow "answers" the algorithmic probe, re-embeddmg statistical patterns in, for instance, narratives of identity, gender and morality that are very familiar to social-psychology debates. There is no straightforward way to make an archetypal analysis, since it is a recursive process that depends on exhaustive intertextual comparisons grounded on deep human experiences. For this reason, archetypal analysis is highly dependent on human interpretation and cannot be easily reproduced by a machine (just archetypal traits). Rather than a limitation, that asymmetry is what completes the dialogical loop envisioned in our introduction.
We start by asking if any well-known archetypes can appear in one of the phrases that the algorithm generated: "and gave the girdle tojuno his sister wife so queen", for example. After making this initial question, we should search in the literature other poetic works for similar patterns involving female figures. Assuming we have found the text The Golden Ass, we can now analyze the goddess Juno, wife and sister of the Great Jupiter (Apuleio, 2020). Consequently, we have a mythical and literary pattern that appears in two different texts: the sentence generated by the algorithm and the classic work The Golden Ass. This pattern is a woman who is a wife and also a sister. Our archetypal analysis method attempts to extend this comparative study to the other texts beyond our corpus, starting from the sentences that the AI algorithm created. In this way, we can start to think about the possible reproduction of well-known archetypal traits initiated by the algorithm.
4. Results
The analysis of the text generated by the AI algorithm showed patterns that can be related to the following archetypes: amma (Chart 02) and shadow (Chart 03). The Erst archetype found was the amma, which is the personification of the feminine that exists inside a man. "No man is so entirely masculine that he has nothing feminine m him. The fact is, rather, that very masculine men have - carefully guarded and hidden - a very softemotional life, often incorrectly described as 'feminine."' (Jung, 1972b, p. 189). This element described as something feminine is related to the amrna, which appears in myths and literary works wearing "the features of Aphrodite, Helen (Selene), Persephone, and Hecate" (Jung, 1979, p. 21). Furthermore, the amrna is an archetypal image that represents the collective image of women in men's psyche. In literature, it is represented by idealized characters who present exacerbated beauty or high magic power, like the witch Circe in Horner's Odyssey. Hesiod's Theogony exemplifies this typical pattern by singing about "he Heliconian Muses", who possess the great and holy mountain of Helicon and dance on their softfeet around the violet-dark fountain and the altar of Cronus' mighty son" (Hesiod, 2006, p. 3). A possible-though as yet untested-parallel can be drawn to what Sandra JovchelcAitch (2007) terms "gendered anchoring" in social representations. Whether such prototype figures play a comparable norm-shaping role m current social cognition remains still a relevant empirical question for scholars of today! Psalos, 2012).
We also found traits of what the Jungian call the 'shadow·, the lower and repressed personality of every individual. Whenever the reader identifies a character who is the opposite of the hero or protagonist - such as Dr. Jekyll and Mr. Hyde, or Dorian Gray and his picture - the shadow is present in the narrative. The shadow also represents the characteristics that people tend to hide: envy, inferiority, fear, jealousy, etc. In myths and literature, this dark personality manifests itself through the villain. For Jung,
The shadow coincides with the "personal" unconscious (which corresponds to Freud's conception of the unconscious). Again, like the anima, this figure has often been portrayed by poets and writers. I would mention the Faust-Mephistopheles relationship and E. T A. Hoffmann's tale The Devil's Elixir as two especially typical descriptions. The shadow personifies everything that the subject refuses to acknowledge about himself and yet is always thrusting itself upon him directly or indirectly-for instance, inferior traits of character and other incompatible tendencies (Jung, 19S0b. pp. 2S·-2S5}.
Using phrases and words from the chosen corpus, the algorithm generated texts whose content can be related to the archetypes. For example, we can explore the anima archetype using the excerpts in Chart S:
The first example [and gave the girdle tojuno his sister wife so queen) has an evident literary parallel. In Ovid's Metamorphoses, Juno is the "queen of all the world, Jove's sister, / His wife, indeed his sister" (Ovid, 1953. p. 72). The anima embraces what Joseph Campbell calls "the paragon of all paragons of beauty, the reply to all desire, the bliss-bestowing goal of every hero's earthly and unearthly quest. She is mother, sister, mistress, bride" (Campbell, 2004, p. 101). The manifestation of this archetype also relates to characters who have a relationship with the hero or protagonist. Marie-Louise von Franz argues that "it is typical of mythological relationships that the woman is always the mother, sister, wife, daughter of her husband, father, and so on" (Von Franz, 2017). Furthermore, the "anima and her condensation with the figures of the sister, wife, mother, and daughter, plus the associated incest motif, can be found in Goethe ("You were in times gone by my wife or sister·), as well as in the amrna figure of the regina orjemina alba in alchemy" (Jung;, 19S0b. p. 2S5). Another narrative pattern involving the bride is in the story of the beautiful Helena, the wife of King Menelaus. The character was kidnapped by Pans and involved in the Trojan War, as narrated in the Iliad. Orpheus' search for Eurydice also illustrates this typical facet of the amrna: the wife or maiden who must be rescued. This rescue script might e\"oke benevolent-sexism themes noted in social-representation research, that could be explored in a separate study (Jovchelovitch, 2007).
The second selected sentence (Island where dwells the great goddess) manifests the theme of the amma that lives in an unknown place. In the Odyssey, the witch Circe also lives on an island, where Ulysses eventually lands with his crew. Another example in mythology is the nymph Calypso, who lived on the island of Ogygia (Grimal, 1990). The myths show-preferences for the scenarios that Joseph Campbell (20O4. p. 72) called "regions of the unknown (desert, jungle, deep sea, alien land etc)", and Jung claims that the hero usually gets into "the region of danger (watery abyss, cavern, forest, island, castle etc.)" (Jung, 19S0a, p. 335). Therefore, it is not surprising that certain female characters are related to caves, forests, or isolated places. To a certain extent, this excerpt generated by the algorithm finds similarities in narratives
in which a hero has to make a perilous journey into a place of great danger where the heroine is held .... We should have enough training in metaphorical thinking by now to realize that the sea, the sea monster, and the foreign island he lands on are all the same place and mean the same thing. (Frye, 1990, pp. 190-191)
In our study, these regions of the unknown can be related to the anuria, the personification of the woman that exists within the man, being, therefore, something generally repressed and unrecognizable. In succession, the next sentence m the table {great and powerful goddest) recapitulates the word "goddess" found in the second example. The female deity can be associated with a mighty female character, like the divinities that inhabited Olympus: Athena, Hera, Artemis, Aphrodite, and others. This pattern can also be found in "the Mother of God, the Virgin, and Sophia" (Jung, 19S0b, p. Si), or even in characters that are generally better and more beautiful than humans are, like Beatrice in the Divine Comedy.
The fourth example (emilia come be asjull of love of a dungeon) shows a female character located in a dungeon, a variation of the unknown location, and brings up the figure of the loving woman. Calypso herself kept Ulysses "in her spacious cavern: she wanted to make him her husband" (Homer, 2002, p. S3). The archetypal narratives are full of weddings and love encounters. For example, an elven lady "full of love" is found m the plot of The Lord of the Rings. After the destruction of SauronJ Aragorn finally marries Arwen, as both were in love smce the beginning of the narrative. The union with the loving woman is also a constant in the hero's quest. Joseph Campbell claims that the hero's ultimate adventure, "when all the barriers and ogres have been overcome, is commonly represented as a mystical marriage (ir.po$ ydfiOQ) of the triumphant hero-soul with the Queen Goddess of the World" (2004, p. 100). The woman m love also appears in Ovid's Metamorphoses, specifically in the story of Thisbe, who had an inescapable love for Pyramus. "They came to know each other, as time passed/ Love flourished, and if their parents had/ Not come between them, then they would have shared/ A happy wedding bed" (Ovid, 195S). The tale of Pyramus and Thisbe is the "archetype" of Shakespeare's Romeo and Juliet. In this play, Juliet is the representation of the loving anuria.
The fifth sentence analyzed can be related to the woman who provides help (the companionate who succoured me). In Homer. Athena assists Telemachus on his journey in search of Ulysses. Ariadne is another example of a female character who comes to help the hero. "When Theseus arrived in Crete to do battle with the Minotaur Ariadne fell in love with him; to enable him to find his way in the Labyrinth where the \finotaur was confined she gave him a ball of thread, which he unwound to show him the way to return" (Grimal, 1990, p. 59). Here, we can see again that the anima tends to orbit hidden or chthomc places (Ariadne's thread was extended along the labyrinth', often appearing as the woman who offers help. We can also bring to discussion the character Galadnel, who helped Frodo during his journey through Middle-earth, presenting the hobbit with a magic flask that illuminated the underground paths of Mordor. Finally, the manifestations of the anima presented so far - the mistress, the helper, and the protector - can be found in the figure of the Valkyrie. Johnm Langer attributes four essential aspects to Valkyries: "attendants (serving in Valhalla), lovers /wives, fighters (choosing and protecting heroes and kings), prophetesses (m connection with destiny;" (2015, p. 539).
The sixth example taken from the text created by the algorithm [o maiden thou sorrow laden thy gracious countenance) represents the melancholic woman. The sad and unfortunate lady appears in the work of Edgar Allan Poe. In the short story The Oval Portrait, for example, the beautiful woman who serves as a model for the composition of a painting becomes increasingly discouraged and weak. Her death is the terrible consequence of the painter's obsession. On the other hand, Gaston Bachelard establishes a concept called The Ophelia Complex, which can be related to the image, also very present m literature, of the suicidal woman or the sufferer "who can only weep about her pain and whose eyes are easily 'drowned m tears'" (19S3, p. S2). In Hamlet, Ophelia says "And I of ladies most deject and wretched" (Shakespeare, 2003, p. 163), personifying the melancholic and tearful anima. She can be related to "an archetypal model figure of the overemotional woman" (Von Franz, 2017). Furthermore, the last excerpt created by the algorithm mentions the character Ophelia {the most beautified opkelia). This repetition can exemplify the recurrence of the archetypal figure, an image that connects one literary work with another. If "the name of Ophelia comes repeatedly to the lips under the most varied of circumstances, we conclude that this unity, her name, is the symbol for a great law of the imagination" (Bachelard, 19S3, p. S9). In the present study, this great law of human imagination is precisely the archerype of anima.
Consistent with our chained design, the preliminary insights on anima merely flag hypotheses that the next computational loop - and future empirical work - can test more formally (Nelson, 2020). Exploring now the shadow archetype, with Chart S text excerpts:
The first example extracted from the text generated already contains a word that refers to the shadow archetype {utterance of the eternal shade"). This eternal shadow can be related to the conception of darkness and obscurity m human thought since ancient times. In The Republic (Plato, 2003), a group of men lived in a cave, chained and facing the wall. A fire burned behind the prisoners. Between them and the flame, there was a wall over which various objects were carried by other men. For the prisoners, the "truth would be nothing more than shadows" (Plato, 2003, p. 22l). All the richness of the Platonic allegory is in the shades that exist inside a cave, creating the most fruitful reflections about the "real" and reality. We should note that the allegory of the cave was revisited by many authors. Therefore, it is not surprising that Gaston Bachelard, who studied the imagination of the earthen element, stated that a "light full of dreams prevails within the cave, and the shadows thrown on the cave walls can easily be compared to visions seen in dreams" (Bachelard, 2011, p. 149). The repetition of the cave and shadow themes shows how one text can be linked to another through a typical pattern frequently revisited by the authors. If this darkness imagery still influences m the same way contemporary moral-threat perception could be a matter for future investigation in the social psychology field ;Gamez-Djokic. Waytz, & Kouchala, 2025).
As an archetype, the shadow can connect one poetic work to another, originating what Northrop Frye described "as an element of one's literary experience as a whole" (Frye, 2000, p. 365). So, it is possible to continue tracking the archetypal figure in question as it spreads through literature. For example, the shadow is one of the central themes of John Donne's poem "A Lecture Upon the Shadow":
STAND still, and I will read to thee
A Lecture, love, in Loves philosophy.
These three houres that we have spent.
Walking here. Two shadowes went
Along with us, which we our selves produc'd;
(Donne, 1994, pp. 53-5+J
In the novel The Lord of the Rings, a more recent narrative, the villain Sauron is characterized as a fearful shadow. His main allies are the "Ringwraiths. shadows under his great Shadow" jTollaen, 2005, p. 51). All the mentioned examples show that the concept of shadow permeates several texts, representing a typical idea that always returns, as we will observe from the material produced by the algorithm. In the text generated by the algorithm, some excerpts can be related to the main manifestation of the shadow archetype, the villain (Frye, 2000, p. 304). In the book Morphology of Tfte Folktale, Vladimir Propp affirms that the role of the villainous character is "to cause some form of misfortune, damage, or harm. The vulain(s) may be a dragon, a deviL bandits, a witch, or a stepmother etc." (Propp, 2009, p. 27). In countless works of fiction, the hero has a dark counterpart that represents "the inferior and therefore hidden aspect of the personality, the weakness that goes with every strength, the night that follows every day, the evil m the good" (Jung, I9S5, p. 219). The evil opposite of the hero manifests itself through the perverse character, the dark enemy found in many works of fiction (Sauron, Lord Voldemort, the Devil of Christianity, \Iephistopheles, and others).
In the text generated by the algorithm, we can observe the presence of the villam/enemy m the sentences "and a villain at least i am", "into the house was an enemyjdr beyond the powers of all your efforts", "know of this hellish villain" e "And a villain iago", in examples two, three, four, and seven. As the algorithm reuses common phrases and words taken from a sample of several literary works, it is not surprising that the textual content generated presents patterns that can be related to the villain, an extremely recurring character m fiction. Viewed through a social-identity lens, the villain lines generated by the model act as archetypal out-group foils that help define heroic m-group virtue - although demonstrating this dynamic rigorously m algorithmic text will require dedicated follow-up analyses (Gamez-Djokic, Waytz, & Kouchalu, 2025).
Feelings that are generally denied and hidden through the persona (the mask that the individual wears to present himself to the outside reality!1 are also part of the shadow. Thus, examples six {.4nd turned their angry thoughts', and eight [sami began to feel gloomy or depressed in this country} can be related to this hidden personality. The shadow expresses emotions considered uncomfortable, often revealing a trait of character "which is strange even to the person concerned" ;Jung, 19S0b, p. 279). In other words, this archetype shows a shadowy face of the psyche. Consequently, we must discuss the fifth selected sentence of the generated text [the depths of a sombre brooding brain). These depths of a dark brain can be associated with the idea of the unconscious:
by the Age of the Enlightenment, some perceptive students of human nature had recognized the existence of unconscious mentation. .... Goethe and Schiller, whom Freud could quote by the hour, had sought the roots of poetic creation in the unconscious. The romantic poets, in England, France, and the German states alike, had paid tribute to what Coleridge called "the twilight realms of consciousness." In Freud's own lifetime, Henry James explicitly linked the unconscious with dreams; the narrator in his novella "The Aspern Papers" speaks of "the unconscious cerebration of sleep." Freud could discover very similar formulations in the memorable epigrams of Schopenhauer and Nietzsche. His particular contribution was to take a shadowy, as it were poetic, notion, lend it precision, and make it into the foundation of a psychology by specifying the origins and contents of the unconscious and its imperious ways of pressing toward expression.
(Gay, 2006, p. 24)
Therefore, the unconscious (the shadowy notion described by Freud or maybe what Coleridge called the twilight realms of consciousness) is a portion of human nature often discussed in literature, psychology, and philosophy. For Carl Jung, this concept encompasses
everything of which I know, but of which I arn not at the moment thinking; everything of which I was once conscious but have now forgotten; everything perceived by my senses, but not noted by my conscious mind; everything which, involuntarily and without paying attention to it, I feel, think, remember, want, and do; all the future things that are talong shape in me and will sometime come to consciousness: all this is the content of the unconscious.....To this marginal phenomenon, which is born of alternating shades of light and darkness, there also belong the Freudian findings we have already noted. (Jung, 1972a, p. 1S5)
Jung associated Freud's ideas with shades of light and darkness, something similar to the twilight realms of consciousness quoted above. Furthermore, E. M. Meletinsld (2015) writes that the shadow is the unconscious part of the personality, and Gaston Bachelard also shows that beneath "the tall psychic house, there is a labyrinth in us that leads to our hell" (2011, p. 170). The examples above describe the recurring notion of the dark part of the human being, an idea we can connect to the algorithm's sentence: the depths of a sombre brooding brain. Taken together, these preliminary antma and shadow traits illustrate how a lightweight generative model can surface narrative regularities that Analytical Psychology and Critical Studies of Literature recognizes, while also offering social-psychology researchers a compact, testable set of hypotheses about representation, gender and in-/out-group imagery that merit follow-up in dedicated empirical designs.
5. Discussion
After analyzing archetypal traits of the stochastic text produced by the n-gram model, our final goal in this article is to make epistemological and methodological connections among some academic fiefdoms that seldom engage m practice: Psychology, Literature, and Computer Sciences (Weingart, 201S).
These efforts also resonate, more tentatively, with current debates m social psychology When positioned as discursive artifacts, our algonthmically generated narratives suggest that even lightweight generative models - when trained on culturally resonant corpora - can recombine symbolic patterns m ways that reproduce familiar social imaginary. Specifically, the recurrent presence of amma-like figures may reflect entrenched affective scripts around femininity, inviting further investigation into how generative systems might stabilize gendered representations through textual form (Jovchelovitch, 2007; Psalas, 2012). Likewise, the prominence of the shadow archetype over 'light" motifs raises questions about algorithmic amplification of threat-oriented narratives, a theme with relevance to studies on moral salience and out-group framing (Brady & Crockett, 202·; Waytz, 2020). Rather than general claims about agency or identity, our contribution lies in outlining how specific symbolic configurations - historically shaped and computationally resurfaced - can form testable hypotheses for social-psychological inquiry into discourse and representation.
By establishing discursiveness (Foucault, 2019) between knowledge domains with distinct perspectives for text, we are harmonizing scientific endeavors often considered incompatible but potentially complementary (Tsoukas, 2009). When scholars from different knowledge domains collaborate, they might either: (a) become too enthusiastic and overlook the limitations of the other academic disciplines; or (b) too critical and doubtful about the other disciplines' actual contributions. We took some precautions to avoid incurring into such inaccuracies.
The main precaution taken was to internalize that interdisciplinary involves mutual understanding between different kinds of knowledge building [Repko et al., 2011). There must be "sense of touch" to combine theoretical constructs It is also necessary to understand the internal logic of research approaches to avoid confusing distinct phenomena in the process (Carlin, 2016). The lack of adherence to this principle is precisely why certain Computer Science works, such as 'Datacrysm" (Rudder, 2014). are criticized outside the hard sciences When the positivist perspective dominates (Burrell & \forgan, 2017), the researcher's tendency is to make careless generalizations about sensitive research topics long debated m humanities and social sciences. This kind of mistake happens for a simple reason: we are not supposed to unrestrictedly apply models from the natural sciences to the social held, because most of the computational models ignore human spontaneity, human adaptability to social contingencies, and ethical issues (Esfeld, 2022).
Some computer scientists even imply the "end of theory" as soon as we have more organized data to train better algorithms (Anderson, 200S). As Noam Chomsky & Michel Foucault (2015) agreed once, the end of theory is a Utopia, since it assumes that we can arrive at a final and complete understanding of the world. This is not possible given the complexity and unpredictability of nature and social structures.
Just as problematic as the excesses of "data positivism" (Jones, 201S) is the tendency, within Analytical Psychology and Literary Criticism, to maintain distance from computational methods and refrain from incorporating thern into their broader research ecology (Latour, 19SS). While this resistance often stems from a legitimate concern with preser\ing interpretive nuance, it can inadvertently limit the capacity to engage with higher-order patterns and synthetic representations of cultural production. When critically employed, such computational abstractions can help render visible recurrent symbolic structures across extensive textual corpora (Latour, 2005.. From a psychosocial perspective, this dynamic aligns with the concept of objectiftcation - the process by which shared meanings become stabilized through concrete symbolic forms (Jovchelovitch, 2007). In our study, the iterative movement between algorithmic generation and interpretive analysis operationalizes this process in a new technical key, offering a controlled environment in which symbolic configurations are not only recombined, but also recontextualized through human reading.
To reconcile these opposites we performed three analytical movements (Venturini, 2010). They were intentionally related to the guiding principle! of Jung's theory, in which opposites such as light and shadow, anima and animus, mother and father, are reconciled for the understanding of the human being (Jung, 19S0b',.
The first movement was the automatic text generation from a selected corpus. It involved a predominantly quantitative and empiricist approach, linked to an objective ontology and positmst epistemology (Hatch Sz Cunliffe, 2013). We transformed entire books into bags of words, determined the frequency of these words, and calculated the probabilities of drawing a new word given a pre%ious set of existing ones. This assured us that the algorithm's text was not freely written like the corpus' book but rather had a stochastic orientation. In other words, although the algorithmic output was not completely predictable, clear rules and constraints influenced all textual outcomes pointing to a "general trend' of the corpus. The movement is, therefore, of closing, narrowing, reducing complexity. We are digesting several books and generating sentences that are, at the same tune, creating unseen texts and replicating some of the most recurrent patterns and archetypal traits of the selected works. In a psychosocial perspective, such compression e%"okes the discursive process of anchoring - where new information is made intelligible through existing symbolic structures (Psaltis, 2012).
The second movement is the archetypal analysis of the automatically generated text. It involved a predominantly qualitative and interpretive engagement which is very common in humanities and social sciences. This movement seeks to draw parallels between what was reduced by the algorithm and the well-known archetypes of Literary Criticism and Analytical Psychology (Frye, 1951; Jung, 19S0b). The archetypal analysis is strongly grounded in che analyst's mindset; hence, the algorithm is not aware of the cognitive structures used m the analysis nor the sociocultural context that the analyst is in. As a stochastic parrot, the algorithm will probably never be aware of such things due to its way of processing information! Bender et al.. 2021'.. Remember that despite the fact that L\f texts are becoming increasingly convincing, algorithmic decision-malang remains exceedingly calculative, utilizing decision trees, probabilistic models, and so forth. The second movement is, therefore, interpretive, and it involves opening and expanding ideas, drawing parallels between narrative structures the analyst met m the past and what they are actually perceiving while reading the algorithm's output (Boje. 201 l). This interpretive moment foregrounds precisely what a discursive perspective rn psychology values: the subject's position in meaning production, and the plasticity of symbolic material in new communicative arrangements (Jovchelovitch, 2007; Psaltis, 2012).
The third movement, which is precisely happening now, tries to reconcile the two previous movement. Such reconciliation passes through the explanation of the contexts that the combination of these "closing-opening" movements are useful. In our case, we agreed that this is the processes of scientific ideation and theorization. Ideation pertains to the generating, developing, and communicating of a large number of new ideas, often involving fast-paced dialogical methods such as brainstorming (Bjork et al., 2010). Theorization, on the other hand, involves providing meaning and significance to what scientists observe, using e%idences to refine existing theory, or construct new theory (Gregory & Henfridsson, 2021). Ideation processes usually provide the creative spark for the theorization process, which in its turn organize and validate the ideas. We suggest that LLM-generated material, when processed through interpretive frameworks, may become a land of creative and reflective elaboration that underlies both ideation and theorization.
Our computational experiment allowed us to observe both ideation and theorization in motion. As we engaged with the model's outputs, we began to recognize elements that resembled the anitna and tkadozv archetypes, prompting us to explore their recurrence not only as literary residues, but as potential indicators of deeper symbolic structures. Although our analysis is grounded in Analytical Psychology, the emergence of these motifs in machine-generated discourse gestures toward another phenomenon: the persistence and transformation of cultural patterns across different media environments - a topic of growing relevance in social-psychological research on symbolic circulation and anchoring (Brady & Crockett, 2024; Psaltis, 2012).
As the anuria is the conception of the female figure in men's psyche and it contains exacerbated idealizations about the role of women in society, we hypothesized that its strong presence in the output of LM happens due to historical issues of gender inequality. We found initial confirmations of this hypothesis in literature (Puspawarru & Amelia, 2023), emphasizing that, although it is beginning to change now (Mills et al., 2010), the anuria archetype is stronger in epic-heroic literature than the animus, given the fact that the tradition of this literary genre is gender-biased. We conclude our theorization process, thinking that it would be interesting to compare these results with contemporary narrative genres that are more progressive in terms of gender equality, such as the Narratives of Change (Wittmayer et al., 2019). We propose that future work contrast corpora with different gender paradigms - such as contemporary Narratives of Change (Wittmayer et aL, 2019) - to explore whether large language models trained on such materials reproduce or attenuate gendered representational biases. This question aligns directly with current debates in the social psychology of gender, identity and media.
The prevalence trend of the shadow archetype in relation to the light archetype in the algorithm's output initially surprised us and drew our attention. An idea came up collectively Based on the evidence we've had, we theorized that the strong presence of the shadow archetype in the stochastic text happens because the hero's journey has more moments of challenges and confrontations, surrounded by shadow, than sparse moments of glory and light, reserved for narrative resolutions (Campbell, 200·). This narrative asymmecry mirrors patterns identified in social psychology, particularly in research on emotional salience and symbolic threat: domains where negativity and danger exert disproportionate influence on collective meaning-making (Gamez-Djokic, Waytz, & Kouchaki, 2025). Whether these algorithmically amplified cues reinforce or reconfigure readers' affective schemata remains an open empirical question, with potential implications for studies in social cognition and mediated discourse.
We cannot underestimate the impact that processes of ideation and theonzaaon can potentially have on science. If we just look back, we can see many examples of that in the history of science. For instance, the appropriation of the physical idea of the electromagnetic field (Cassirer, 2005). From this initial metaphor (Lakoff& Johnson, 200S). the Social Theory of Fields was developed, claiming that social structures behave and interact as fields, or relatio-nally in a less deterministic manner. This theory proved itself immensely useful m explaining contemporary social phenomena, becoming widelK· influential in management, social sciences, and applied social sciences academic fields.
We can also recall the debate between the philosopher David Chalmers and the neuros-cienast Chris tof Koch about "the neural correlates of consciousness." Koch asserted, based on the empirical efforts of neurosciennsts in 1990s, that within 25 years, the brain regions responsible for consciousness would have been discovered and mapped. On the other hand, Chalmers, based on the philosophical complexity of the mind-body problem, argued that this would not come to fruition due to a theoretical issue. In the recent meeting that these scholars had at New York University, Chalmers' victory over Koch was proclaimed, since we are not even close to determine the body parts that are responsible for human consciousness. This, to some extent, marked a rare acknowledgment of philosophy's influence over the natural sciences (Horgan, 2023). In a similar way, we hope that the speculative and etiological posture adopted in this article - grounded in symbolic analysis but attentive to interdisciplinary resonance - may help to reposition generative models not only as technical tools, but as participants in the ongoing social negotiation of meaning.
We conclude this article by saying that although 'stochastic parrots' have been widely criticized for their misuse in poor customer services, questionable business models, and inequality perpetuation m decision systems ; Bender et al., 2021; O'NeiL 2016), we may have found a case where such parrots' words have the harmless purpose of inspire. Instead of dictating unfair rules, the parrot's words may give to scholars new insights for better ideas, better theories, better analyses. From a discursive psychology standpoint, the interpretive interface between human analyst and algorithmic output may offer not only technical affordance, but also conditions for subjective positioning, dialogical reflection and symbolic reorganization - key concerns of this journal's field of inquiry (Jovchelovitch, 2007; Psaltis, 2012). The interplay between LLM text and expert interpretation, as staged here, thus offers a modest demonstration of methodological fluency capable of enriching dialogues across Computer Sciences, Psychology, and Literature.
Sidebar
References
References
Anderson. C. (2008). The end of theory: The data deluge makes the scientific method obsolete. Wired Magazine, 16(7), 7-16.
Apuleio. R. O. (2020). O Asno de Ouro [Thegolden ass]. Editora 34.
Bachelard, G. (1964). The Psychoanalysis of Fire. Routledge & Kegan Paul.
Bachelard, G. (1983). Water and Dreams: An Essay on the Imagination of Matter. The Pegasus Foundation.
Bachelard, G (2011). Earth and Reveries of Repose: An Essay on Images oflnteriority. The Dallas Institute Publications.
Bakhtin. M. M. (1981). Vie dialogic imagination: Four essays (M. Holquist. Ed.; C. Emerson & M. Holquist. Trans.). University of Texas Press, https://doi.org/10.7560/715271
Beer. D. (2017). The social power of algorithms. Information, Communication & Society, 20(1), 1-13. https://doi.org/10.1080/1369118X.2016.1216147
Bender. E. M-. Gebrn, X. McMillan-Major. A.. & Shmitchell. S. (2021). Sobre OS pe-rigos dos papagaios estocasticos: modelos de linguagem podem ser grandes demais? In Anais da Conferencia da ACM de 2021 sobre [ustica, Responsabilidade e Transparencia (FAccT'21) (pp. 610-623). Associacao para Maquinas de Computacao (New York). https://doi org/10.1145/3442188.3445922
Bjork. X, Boccardelli. P., & Magnusson. M. (2010). Ideation capabilities for continuous innovation. Creativity and Innovation Management, 19(4), 385-396.
Boje. D. (2011). Narrative Methods for Organizational & Communication Research. Sage Press.
Borges. J. L. (1999a). Collected Fictions. Penguin Books.
Borges. J. L. (1999b). Obras Completas de large Luis Borges [Complete Works by Jorge Luis Borges]. Globo.
Brady. W. J. & Crockett, M. J (2024). Norm psychology in the digital age: How social media shapes the cultural evolution of normativity. Perspectives on Psychological Science, 19(1), 62-64. https://doi.org/10.1177/17456916231187395
Burrell, G. & Morgan. G. (2017). Sociological paradigms and organisational analysis: Elements of the sociology of corporate life. Routledge.
Campbell, J. (2004). The hero with a thousand faces. Princeton University Press
Carlin. A. P. (2016). On some limits of interdisciplinarity. Social Epistemology, 30(5-6), 624-642.
Cassirer. E. (2005). From Substance and Function. Continental Philosophy of Science, 84-89.
Chomsky. N. & Foucault. M. (2015). The Chomsky-Foucault debate: on human nature. The New Press.
Dale. R. (2021). GPT-3: What's it good for? Natural Language Engineering, 27(1), 113-118.
Dennett, D. (1987). The Intentional Stance. MIT Press.
Donne, X (1994). The complete poetry and selected prose of John Donne. Modern Library.
Duarte. F. (2025). Number of ChatGPT Users (2025). Exploding Topics. https://explodingiopics. com/blog/chatgpt-usersgtop
Esfeld. M (2022). From the open society to the closed society: reconsidering Popper on natural and social science. Futures, 102920.
Federici. $.. Montemagni, $.. &: Pirrelli. V. (1996). Shallow parsing and text chunking: a view on underspecification in syntax. Cognitive Science Research Paper-University of Sussex CSRP, 35-44.
Foucault. M. (2019). Discourse and Truth" and" Parresia. University of Chicago Press.
Frye. N.(1990). The Great Code: The Bible and Literature. Penguin Books.
Frye. N. (1951). The archetypes of literature. Vie Kenyon Review, 13(1), 92-110.
Frye, N. (2000). Anatomy of Criticism: Four Essays. Princeton University Press. Gamez-Djokic, M., Waytz, A., & Kouchaki, M. (2025). Poets over quants: Automation and AI threats increase the value people place on creativity. Personality and Social Psychology Bulletin. Advance online publication, https://doi.org/10.1177/01461672251337126
Gay, P. (2006). Freud: A Life for Our Time. W W. Norton & Company.
Grant, C. B. (2016). Information Theory. The International Encyclopedia of Communication Theory and Philosophy· 1-11.
Gregory. R. W. & Henfi-idsson. O. (2021). Bridging art and science: Phenomenon-driven theorizing. Journal of the Association for Information Systems· 22(6), 1509-1523.
Grimal, P. (1990). The Concise Dictionary of Classical Mythology. Basil Blackwell.
Hatch, M. J. (2018). Organization theory: Modern, symbolic, and postmodern perspectives. Oxford university press. https://doi.org/10.1017/CBO9781107415324.004
Hermans. H. J M (2001). The dialogical self: Toward a theory of personal and cultural positioning. Culture & Psychology, 7(3), 243-281. https://doi.orfi/10.1177/1354067X0173001
Hesiod. (2006). Theogotiy. Works and Days. Testimonia. Harvard University Press.
Homer. (2002). The Odyssey. The University of Michigan Press.
Horgan. J. (2023. June 26). A 25-Year-Otd Bet about Consciousness Has Finally Been Settled - Scientific American. Scientific American. https://www.scientilicamerican.com/article/a-25-ye-ar-old-bet-about-consciousness-has-linallv-been-settled/
Ignatow. G.. & Mihalcea. R. (2018). An Introduction to Text Mining Research Design Data Collection and Analysis. Sage Publications.
Jones. M. L. (2018). How we became instrumentalists (again) data positivism since World War II. Historical Studies in the Natural Sciences, 48(5), 673-684.
Jovchelovitch, S. (2007). Knowledge in context: Representations, community and culture. Rout-ledge. https://doi.orp/10.4324/9781315173368
Jung. C. G. (1971). Vie spirit in man, art, and literature. Princeton University Press.
Jung. C. G. (1972a). Vie Structure and Dynamics of the Psyche. Princeton University Press.
Jung. C. G. (1972b). Two essays on analytical psychology. Princeton University Press.
Jung. C. G. (1975). Psychology and Religion: West and East. Princeton University Press.
Jung. C. G. (1978). Civilization in Transition. Princeton University Press.
Jung. C. G. (1979). Aion: Researches Into the Phenomenology of the Self. Princeton University Press.
Jung. C. G. (1980a). Psychology and alchemy. Princeton University Press. Jung. C. G. (1980b). Vie archetypes and the collective unconscious. Princeton University Press. Jung. C. G. (1985). The Practice of Psychotherapy. Princeton University Press. Lakoff. G. & Johnson. M. (2008). Metaphors we live by. University of Chicago press.
Langer, J. (2015). Dicionário de mitologia nórdica: símbolos, mitos e ritos [Nordic Mythology ictionary: Symbols, Myths and Rites]. Hedra.
Latour. B. (1988). Science in Action - How to Follow Scientists & Engineers Through Society. Harvard University Press.
Latour. B. (2005). Reassembling the Social: an Introduction to Actor-Network-Theory (1st ed.). Oxford University Press.
Meletinski, E. M. (2015). Os arquetipos literdrios [The literary archetypes]. Atelie.
Mills. S. J.. Pankake. A., & Schall. J (2010). Children's books as a source of influence on gender role development: Analysis of female characters using lung's four archetypes. Journal of Women in Educational Leadersliip, 264(1), 99-117.
Moloney, G.. Hall. R.. & Walker. I (2005). Social representations and themata: 'Ihe construction and functioning of social knowledge about donation and transplantation. British Journal of Social Psychology. 44(i), 415441.
Nazar. R. & Renau, I. (2012). Google books n-gram corpus used as a grammar checker. Proceedings of the Second Workshop on Computational Linguistics and Writing (CIA&W 2012): Linguistic and Cognitive Aspects of Document Creation and Document Engineering, 27-34.
Nelson. L. K. (2020). Computational Grounded Theory: A Methodological Framework. Sociological Methods and Research, 49{ 1). 3-42.
O'neil C. (2016). Weapons of math destruction: How big data increases inequality and threatens democracy. Crown.
Ovid. (1958). The Metamorphoses. The Viking Press.
Peters. B. G. (2017). What is so wicked about wicked problems? A conceptual analysis and a research program. Policy and Society, 36(3), 385-396.
Plato. (2003). The Republic. Cambridge University Press.
Plomp. R. (1986). A signal-to-noise ratio model for the speech-reception threshold of the hearing impaired. Journal of Speech, Language, and Hearing Research, 29(2), 146-154. Psaltis, C. (2012). Culture and social representations: A continuing dialogue. Culture & Psychology, /am. 253-262. httpW/doi org/10 1177/1354067X] 2446233
Propp. V. (2009). Morphology of the Folktale. University of Texas Press.
Puspawarni. R. & Amelia, D (2023). Archetype analysis of main characters in the secret garden (2020) movie by frances hodgson burnetL Linguistics and Literature Journal, 4(1), 1-13.
Rahwan, I., Cebrian. M.. Obradovich. N.. Abeliuk. A.. Alvarez, J. M., Bar-Yam. Y. ... & LeVeck. B. L. (2019). Machine behaviour. Nature, 568(7753). 477-486. https://doi.orp/10.1038/ S41586-019-U38-V
Repko. A. E. Newell. W. H., & Szostak, R. (2011). Case studies in interdisciplinary research. Sage Publications.
Revuz. D. (2008). Markov chains. Klsevier.
Rudder. C. (2014). Dataclysm: Love, sex, race, and identity-What our online lives tell us about our offline selves. Crown.
Ryan. M.-L. (1987). The heuristics of automatic story generation. Poetics, 16(6), 505-534.
Sardinha, T. B. (2000). Linguistica de Corpus: historico e problematica (Corpus Linguistics: History and Problematics]. DELTA: Documentacdo de Estudos em Linguistica Teorica e Aplicada, 16(2), 323-367.
Sarker. I. H. (2021). Machine learning: Algorithms, real-world applications and research directions. SN Computer Science, 2(3). 160.
Scheufele, D. A., Jamieson, K. H.. & Brossard, D. (2021). The science of science communication in the age of artificial intelligence. PNAS, 118(15), e2008056118. https://doi.org/10.1073/ pnas.2008056118
Shakespeare, W. (2003). Hamlet, Prince of Denmark. Cambridge University Press.
Smaldino, P. & O'Connor, C. (2020). Interdisciplinarity can aid the spread of better methods between scientific communities.
Tamkin, A., Bnmdage. M.. Clark. X, & Ganguli. D. (2021). Understanding the capabilities-limitations, and societal impact of large language models. ArXiv:2102.02503.
Tolkien. X R. R. (2005). Vie Lord of The Rings. HarperCollins.
Tsoukas. H. (2009). A dialogical approach to the creation of new knowledge in organizations. Organization Science, 20(6), 941-957.
Venturini. T. (2010). Diving in Magma: How to Explore Controversies with Actor Network Theory. Public Understand. Sci., 19(3), 258-273. https://doi.ore/https://doi. orp/10.1177/0963662509102694
Von Franz. M.-L. (2017). The Golden Ass ofApuleius: Vie Liberation of the Feminine in Man. Shambala.
Weingart. P. (2018). 2. Interdisciplinarit)': The Paradoxical Discourse. In Practising interdisci-plinarity (pp. 25-42). University of Toronto Press.
Wetherell. M. (2021). Critical discursive psychology and the analysis of social inequality. Socio/ and Personality Psychology Compass, 15, e!2543. https://doi.org/10.111 l/spc3.12543
Wittmayer, J. M... B.ickhaus, X.. Avelino. E. Pel. B-. Strasser. T.. Knnze. I.. & Zinjder%vijk. L. (2019). Narratives of change: How social innovation initiatives construct societal transformation. Futures, 112,102433.
