What does partial co-presence remove from the ritual obligations of talk?
This explores what AI conversation loses when the partner isn't fully 'there' with you — specifically, which of the small social obligations that normally bind a conversation (repairing misunderstandings, taking turns, being held accountable for what you said) quietly disappear.
This explores what AI conversation loses when the partner isn't fully 'there' with you — which of the everyday social obligations of talk quietly fall away when co-presence is only partial. The sharpest answer in the corpus comes through Goffman's lens: human talk runs on ritual machinery we barely notice — corrective rituals when something goes wrong, entrainment as speakers fall into each other's rhythm, the accountability built into adjacency pairs (you asked, so I owe an answer), and the constant cue-reading of being in a shared space. LLM dialogue skips all of it, and that gap is exactly why the fluency can feel real while the communication has actually failed What happens to social order when AI removes ritual constraints?.
What gets removed, concretely, is the obligation to *jointly repair and update* what you share. Grounding isn't just trading words — it's the collaborative labor of calibrating that your words point at the same thing as mine Why do speakers need to actively calibrate shared reference?. Partial co-presence drops that labor: studies find LLMs perform roughly 77.5% fewer grounding acts than people, and the very thing we optimize them for — fluent, confident answers — actively erodes the rest Does preference optimization damage conversational grounding in large language models?. Worse, the model can't symmetrically update the shared scoreboard at all; it reads every later turn inside a fixed initial frame, so when you pivot or contradict yourself, *you* are left as the sole keeper of common ground Can LLMs truly update shared conversational common ground?.
The deeper claim is that the missing element isn't a feature you could bolt back on — it's constitutive. You can describe a 'quasi-belief' by stripping belief down to its function, but you can't describe a 'quasi-communication,' because removing the mutual, intersubjective orientation doesn't weaken communication, it eliminates it, leaving only text you must interpret one-sidedly Why does the quasi-prefix fail for communication?. That one-sidedness is the signature of partial co-presence: the ritual obligations were always reciprocal, and reciprocity is the first casualty.
Where the corpus gets genuinely interesting is that it won't let co-presence stay purely negative. Several notes argue the missing piece is grounded in having a *body in a shared world* — consciousness-talk and full linguistic agency presuppose co-presence and joint triangulation on real objects, which no amount of usage supplies to a disembodied model Can disembodied language models ever qualify as conscious? Do LLMs gain true linguistic agency through integration?. And there's no biological host to carry a relationship across the gaps between sessions, so each conversation is reconstituted from stored text rather than resumed by someone who was waiting Does an LLM have anything that persists between conversations?. Yet against all this sits a dissenting, time-indexed wager: that social grounding is *acquired* through participation in language games, so as LLMs become established partners in our linguistic practice they accrue an elementary, child-like version of it Can LLMs acquire social grounding through linguistic integration? Does language create subjects or express them?.
So the thing you didn't know you wanted to know: 'co-presence' isn't one switch but a bundle of obligations — repair, turn-accountability, joint updating, reciprocal interpretation — and partial co-presence removes them unevenly. It strips the reciprocal, body-and-world-anchored ones cleanly, while leaving open the radical possibility that the social ones might slowly grow back through sheer use.
Sources 10 notes
Goffman's framework reveals that LLM-based dialogue skips corrective rituals, entrainment, adjacency pair accountability, and co-presence cues that humans use to build trust and repair understanding. This ritual gap explains apparent fluency masking actual communicative failure.
The same words can mean different things to different speakers because referential grounding is person-specific. True communicative grounding demands collaborative negotiation of how language connects to the world, not mere surface-level word sharing.
Research shows LLMs generate 77.5% fewer grounding acts than humans, and RLHF preference optimization actively worsens this gap. The optimization target—fluent, confident responses—directly undermines the communicative work of establishing shared understanding.
LLMs interpret all subsequent conversational turns within a fixed initial prompt frame, preventing them from symmetrically proposing updates to shared assumptions. Even when users pivot topics or contradict earlier framings, the model cannot absorb revisions into jointly held background—making the user the sole maintainer of conversational scoreboard.
Unlike belief, which can be characterized functionally as quasi-belief, communication is constitutively relational. Removing the intersubjective element doesn't weaken communication but eliminates it entirely, leaving only text generation—which humans must interpret unilaterally.
Current disembodied LLMs cannot be candidates for consciousness because consciousness language originates from and applies only to entities sharing a world with us through co-presence and triangulation on shared objects.
Social grounding and linguistic agency are distinct properties. LLMs acquire more social grounding through integration into language communities, but remain categorically incapable of linguistic agency in the enactive sense, which requires embodiment and precariousness no amount of use can provide.
While humans have a continuous biological-phenomenological substrate that preserves interaction effects during dormancy, LLMs have no analogous carrier. The virtual instance is reconstituted from stored text each time, making resumed and new conversations structurally identical.
Social grounding is acquired through participation in language games rather than possessed innately. As LLMs become established communicative partners in human linguistic practice, they develop elementary social grounding comparable to young children, making the question of LLM understanding time-indexed.
Subjecthood is produced within communicative events, not possessed prior to them. This convergent position across philosophy, linguistics, and cognitive science inverts the standard picture of language as a tool used by pre-existing subjects.