Source-linked AI summary
Incorporating Machine-Others: An Enactive Exploration of Human-LLM Interactions
Elena C. Cuffari, Marta Pérez-Verdugo, Xabier E. Barandiaran
TL;DR
The paper asks how human cognition and linguistic selves are affected when people engage in conversation and collaboration with LLMs. The authors contrast AI’s textual presence, which seemingly acts for itself, with invoking Aristotle through what the author has read or learned. The authors propose that low-friction LLM interactions can require users to perform more semiotic labor, while habitual engagement makes a difference to their linguistic practices.
Problem
The paper asks how human cognition and linguistic selves are affected when people engage in conversation and collaboration with LLMs.
Source: §1 · p. 3Method
The authors contrast AI’s textual presence, which seemingly acts for itself, with invoking Aristotle through what the author has read or learned.
Source: §4.1.1 · p. 17Results
The authors propose that low-friction LLM interactions can require users to perform more semiotic labor, while habitual engagement makes a difference to their linguistic practices.
Source: §3.2 · p. 11, §4.3 · p. 23Takeaways & Limitations
LLM-mediated communication may reduce opportunities to practice regulation and flexibility in human interactions, potentially affecting users’ self-regulation.
Source: §5 · p. 24The authors leave unresolved tensions in the manuscript, including passages they describe as relatively contradictory.
Source: §5 · p. 25
Abstract
from arXiv · showhide
New concepts based in interactional thinking are required to understand how humans are changed by relating to LLMs. Standard theories of technological impact upon users' autonomy are inadequate here, because this technology is capable of inducing linguistic interactions with humans in an unprecedented manner. Traditional theories of social cognition also fall short, because LLMs have neither autonomy nor minds nor sociality as such. And yet, humans interact, talk, and create with them. We base ourselves on enactive theory, particularly on the theory of Linguistic Bodies and recent developments of enactive accounts of AI, to explore how we make sense with machines. We argue that we incorporate machine-others in two ways: in the tool-like sense of incorporation as enabling particular skills, but also in the more linguistic and identity-tied sense of incorporating (and incarnating) LLMs' ways of languaging. However, LLMs are themselves excorporated and excarnated phantoms of our own languaging; we give a (machinic) body to our language. As such, we are not just incorporating machinic others; we are incorporating excorporated others. These conceptual advances draw a new map of our linguistic relation to LLMs, putting them in continuation with other forms of languaging involving other linguistic artifacts (e.g., religious books). We attend to some particular forms that this relation takes today, and how they may be impacting the autonomy and becoming of LLM-engaged humans. We introduce the idea of "susceptive sense-making" to characterize increased semiotic labour that is carried by the lack of friction that we sometimes encounter within these services.
1. Introduction
The introduction asks how human cognition and linguistic selves are changed by conversational and collaborative engagement with LLMs. It argues that people incorporate LLMs both as skill-enabling tools and through their ways of languaging, while LLMs themselves embody excorporated human language.
- The paper asks what humans incorporate and excorporate through meaning-making activities with LLMs, and what those processes mean for selves formed through linguistic interaction.Source: §1 · p. 3
- LLM incorporation has a double status: it enables particular skills and involves taking up LLMs’ ways of languaging in identity-tied ways.Source: §1 · p. 3
- LLMs are described as excorporated and excarnated traces of human languaging that people give a machinic body to.Source: §1 · pp. 3–4
- The paper considers the societal, existential, and linguistic consequences of incorporating these “excorporated others.”Source: §1 · p. 4
2. Human-LLM interaction
The section characterizes LLMs as fluent, context-adaptive but non-autonomous systems used in conversational, directive, and collaborative ways. Enactive theory helps explain why these interactions matter for human meaning-making despite LLMs not being autonomous participants.
2.1 LLMs
- LLMs support smooth, personalized, open-ended conversation, and their availability can invite users to vent about everyday problems.Source: §2.1 · p. 4
- Commercial chatbots can maintain a consistent tone or be configured to agree, unlike friends who may tire, worry, or reframe a user’s account.Source: §2.1 · p. 4
- LLM use includes doing-with, where generated text enters a shared product, and doing-for, where the system follows a user’s directive.Source: §2.1 · pp. 5–6
- LLMs generate context-sensitive, purpose-structured language but lack autonomous agency, subjective experience, and intrinsic norms.Source: §2.1 · p. 6
2.2 Participating in sense-making
- Participatory sense-making explains social cognition through interactive dynamics that can influence, enable, or constitute cognition.Source: §2.2 · p. 7
- Human meaning-making remains open to others through linguistic practices, which extend coordination while also creating conditions for misunderstanding.Source: §2.2 · pp. 7–8
3. Making sense with machines?
The section argues that LLM engagements resemble participatory sense-making without meeting its requirement for autonomous participants. It compares these engagements with textual interaction and proposes susceptive sense-making to capture the uneven human labor and influence involved.
3. Making sense with machines?
3.1 Textual engagements
- Readers can enact a narratorial agency with a text, experiencing its pacing as guidance that interacts with their own meaning-making norms.Source: §3.1 · p. 9
3.2 Less friction, more semiotic labor
- The authors hypothesize that less friction from an LLM can require more human semiotic labor, including taking the lead in chatbot exchanges.They point to additional prompting, evaluation, and repair work reported in cited research.Source: §3.2 · p. 11
- Their account centers on ordinary chatbot exchanges; longer LLM co-produced creative work is a more complex, separate case.Source: §3.2 · p. 11
3.3 Susceptive sense-making
- Repeated exchanges can make an LLM chatbot a virtual other experienced as present even when the user is not directly engaging it.Source: §3.3 · p. 13
- Susceptive sense-making names an interactional dynamic in which the human, unlike in participatory sense-making, takes up and carries the whole meaning-making burden.The human alone has a stake, body, history, and becoming in this relation.Source: §3.3 · pp. 13–14
4. Incorporation, Incarnation, Excorporation, Excarnation
The section extends linguistic-bodies theory to human–LLM interaction, distinguishing tool-like and linguistic incorporation from the excorporation and excarnation of human languaging into machines. It traces how these processes can shape users’ linguistic agency and co-regulation.
4.1 A matrix of susceptive sense-making
- Linguistic-bodies theory treats identity as shaped through others’ utterances, distinguishing incarnating another’s agency from incorporating utterances into one’s own linguistic repertoire.Incarnation embodies another’s agency in one’s utterances; incorporation makes utterances part of one’s own agency and identity.Source: §4.1 · pp. 14–15
4.1.1 Excarnation
- Excarnation uses an LLM to give another agency textual presence, such as asking Claude trained on Aristotle’s works to respond as Aristotle.The authors distinguish machine bodies from organic bodies while describing processes formerly carried out by organic bodies as externalized and automatized into machines.Source: §4.1.1 · p. 15
4.1.2 Excorporation
- LLMs embody collective human textual production and can be personalized to externalize an individual’s agency; these processes can recursively intertwine with incarnation and incorporation.The collective corpus is described as generic, computationally automatized, and incomplete and biased; personalized systems may be trained on a person’s prior communications.Source: §4.1.2 · pp. 17–18
4.2 Incorporation and Incarnation Dialectics with Machine Others
- Human–LLM interaction can involve incorporating or incarnating what has itself been excorporated or excarnated, linking users to machine-mediated forms of human languaging.The section illustrates this with students invoking an author’s thinking through a chatbot trained on that author’s work.Source: §4.2 · p. 18, §4.3 · p. 23
4.2.1 Incarnating LLMs
- AI-generated passages can let users incarnate an abstract voice in their own utterances, replacing some human interactional subtleties while also making tone choices more open to reflection.For example, AI may remove contractions when rewriting an email as professional.Source: §4.2.1 · pp. 18–19
4.2.2 Incorporating LLMs
- LLMs are incorporated both as tools that enable skills and as linguistic influences that can reshape users’ agency, with possible effects on autonomy and cognitive habits.Users may plan around capabilities such as summarizing 20 articles quickly, while risks include cognitive gaps and overload.Source: §4.2.2 · pp. 20–21
4.3 What this means for co-regulation
- An abstract, authoritative-seeming LLM voice may be rigidly incarnated and contribute to externalizing regulation onto AI or assimilating its style beyond AI interactions.The authors compare this possibility to beginning with rigidly repeating a doctrine and eventually reorganizing identity around it.Source: §4.3 · pp. 22–23
- Habitual ChatGPT use shapes emotional, pragmatic, and syntactic/lexical practices; AI-bolstered utterances can pre-determine exchanges and affect co-regulation, though the authors caution that cases are uncertain.The receiver may bear the weight of the LLM’s projected authority.Source: §4.3 · pp. 22–23
5. Making sense with machine others: Considering open-ended consequences
Using enactive Linguistic Bodies theory, the authors describe LLM interaction as susceptive sense-making and examine how LLMs may shape human languaging, autonomy, and social life. They present these consequences as open questions, urging vigilance rather than definitive conclusions.
- Using ChatGPT to mediate communication may reduce opportunities to practise self-regulation and adapting to others, while excorporated work could also increase regulatory demands.The authors present these as possible effects, alongside the possibility that new forms of self-regulation may emerge.Source: §5 · pp. 23–24
- LLMs’ norm-compliant responses may contribute to more homogeneous expression and reduced creativity, but this possible drift needs further study (Sourati et al., 2026).The authors contrast accepted complementary responses with the open-ended, creative moves possible in human interaction.Source: §5 · p. 24
- LLM tutoring studies report positive learning outcomes, with stronger effects among underserved students, while adult users may gain epistemic correction and critical thinking (Henkel et al., 2024; Kestin et al., 2025; Wang et al., 2024; Costello et al., 2024; Li et al., 2025).The authors note that tutoring benefits often involve scaffolding and personalization, and that potential benefits do not rule out other possible consequences.Source: §5 · pp. 24–25
- The authors lack experience from the standpoint of deeply vulnerable, needful embodiment and leave unresolved tensions in their account.They identify this as a limit on their perspective while summarizing their effort to think through LLM–Linguistic Bodies dynamics.Source: §5 · p. 25
- Susceptive sense-making names a distributed interaction in which humans supply extra semiotic labour because LLMs lack true linguistic agency.Users must find suitable prompts and assess whether generated text fits the conversational context; the work of sustaining interaction is distributed differently than in human participatory sense-making.Source: §5 · p. 26
- Humans may incorporate LLM-generated text and use LLMs to regulate communication, while also excorporating their sense-making styles into systems that return textual products.The authors describe these processes as recursive, but say their longer-term effects remain speculative because the technologies are recent.Source: §5 · pp. 26–27
- At social scale, LLM incorporation may reshape expectations and identities around work and communication, while relying on human semiotic labour at both user and producer ends.The authors raise questions about corporate ownership of these infrastructures and the consequences of their integration into cognitive and social life.Source: §5 · p. 27
- The authors call for vigilance and democratic participation in shaping and governing LLMs, while explicitly rejecting drastic conclusions about how humans will change.They characterize these technologies as potentially transformative and deeply incorporated into social life.Source: §5 · p. 27
Declarations
The authors declare no financial or non-financial interests and approve this manuscript version.
- The authors certify no affiliations or involvement with entities having a financial interest in the manuscript’s subject matter or materials.Source: Declarations · p. 28
- The authors certify no affiliations or involvement with entities having a non-financial interest in the manuscript’s subject matter or materials.Source: Declarations · p. 28
- All authors approve this version of the manuscript.Source: Declarations · p. 28