Twittérature : autopoïèse, allopoïèse et générations de Twitterbot littéraire

The paper situates twitterature as a digital literary genre born with Twitter (2006), produced by human tweeters and Twitterbots. It revisits the long-standing question of whether machines can be poets, novelists, or dramatists, drawing on Couffignal’s (1965) claim that machines can select words and ideas to produce literature, Balpe’s (1997) meta-author theory, and Wenaus’s (2021) framing of Twitter as autopoietic and allopoietic. The study’s purpose is to examine and classify literary Twitterbots, their generative mechanisms, and reading modalities, arguing for recognition of robotic twitterature within pedagogy and highlighting the significance of algorithmic creativity and AI in contemporary digital literature.

The paper reviews bot typologies and prior work on classification and practice: early webbots (Leonardo, 1997; Schuessler, 2010), multiple taxonomies for Twitterbots by Veale & Mike (2018), Lampi (2017), Hansen (2017), MacPherson (2016), Enström (2019), and Bohacek/Botwiki (2015). It discusses how these schemes vary by technical dynamism, functional mechanisms, content types, and independence from Twitter. It also references movements and communities that support bot making and literary practices (#botALLY, #Botwiki/#Botmakers, #Moot, #NNNGM, #NaBoMaMo, #BotSociety, CABF, and Festival Cultures Remix), showing the intercultural and multilingual spread of generative literary practices on Twitter.

Hypothetico-deductive methodology. Between August 2019 and November 2022, the author queried Botwiki (botwiki.org) and Twitter using hashtags and accounts such as #Bot, #botALLY, #CBDQ, #BotSociety, #Moot, #Botmakers, #Botwiki, #CulturesRemix, and @CercleBoteursFR to establish the availability of literary Twitterbots. The study defined key terms (nanoliterature, twitterature, Twitterbot, autopoiesis, allopoiesis, deep learning, GPT) and categorized bots by multimodal content, functionality, and recursive procedural generativity. The objective was to inventory literary Twitterbots in the Twitterverse and inform readers/educators about robotic twitterature for use in digital creative writing pedagogy.

• Proposed seven literary Twitterbot categories: génératifs, interactifs, aléatoires, itératifs, combinatoires, créatifs, artistiques.
• Proposed ten generic content categories: twittextuels; d’arts; icôno-twittextuels; iconiques; mémétiques; format image numérique (GIF); technographiques; ludiques; chromatiques; pépins informatiques (glitch).
• Presented examples and creators for each category (e.g., @RealMagicoBot; @EmojiMeadow; @Poem_exe; @n7bot; @BonsGenresBot; @chaquemot; @AestheticBot_22).
• Proposed six generational stages of Twitterbots based on coding processes, linguistic/semantic capabilities, and platform policies:
  1. Twitterbots 1.0: simple, mechanical, list/rhyme-based, limited semantics, short-lived (e.g., @everyword, @chaquemot, @LaFontaineBot, @BotDuCoeur).
  2. Twitterbots 2.0: richer linguistic/semiotic techniques, able to read/learn from tweets and generate metaphors/observations (e.g., @pentametron, @MetaphorMagnet).
  3. Twitterbots 3.0: human-like engagement, grammar correction, cultural commentary, activism (e.g., @Boetien, @IKnowTheseWords, @Protestitas).
  4. Twitterbots 4.0: AI-driven (GPT-2, machine learning, TAL), sophisticated, multitask, critique/translate/manipulate multimodal inputs; examples include @poettranslator, @BigramPoetry, @Wibbitz.
  5. Twitterbots 5.0: Internet of Things connectivity, large-scale models/training (e.g., Google Meena chatbot @meena, Microsoft’s @Tay), conversational focus.
  6. Twitterbots 6.0: advanced LLMs (GPT-3/ChatGPT), autopoietic coding/writing across languages, complex tasks (e.g., @GPT3ChatBot, @imagealttext, @diffusionbot).
• Reading modalities for generative twitterature elaborated: close reading (including outside original media via tools like @threadreaderapp and muskviewer) and proposed “symmetrical reading” that accounts for signs, icons, spaces, times, and device/process elements as part of the digital literary text.
• Establishes autopoiesis/allopoiesis frameworks for Twitterbots in literary creation and identifies educational relevance of robotic twitterature.

The findings demonstrate that Twitterbots can function as literary creators (authors/poets) within autopoietic and allopoietic systems, supporting the hypothesis that procedural and AI-driven mechanisms produce legitimate digital literary works. The expanded taxonomies and six generational stages address the research aim to evaluate and classify literary Twitterbots, clarifying their modalities of production and interaction. By articulating reading strategies—close reading and symmetrical reading—the study provides a critical framework for interpreting bot-generated texts that incorporates technical display, metadata, and device processes, thereby validating robotic twitterature as a pedagogically relevant form and enriching electronic literature studies.

The paper proposes a detailed categorization of literary Twitterbots, outlines six procedural generations, and introduces both close and symmetrical reading modalities that treat on-screen transients and source/code/device elements as part of the literary text. Robotic twitterature and its reading modes are thereby demystified. Observing Twitterbot transformations (2019–2022), the author predicts future developments driven by AI and large language models, with bots evolving into conversational agents capable of recognizing human emotions and assisting in decision-making.

The author notes that hyperlinks functioned during the research period but may not remain accessible due to Twitter’s changing policies, and that some bot accounts or content may have disappeared or changed. The study relies on inventories from 2019–2022 and platform-dependent data, which are subject to volatility in the Twitter ecosystem.