What interpretive work must humans perform to experience AI as a conversation partner?
This explores the labor on the human side — the filling-in, framing, and reading we do — that turns AI text into something that feels like a partner talking back, rather than the AI doing that work itself.
This explores the labor on the human side that turns AI text into something that feels like a partner talking back. The striking claim in the corpus is that the conversational "event" lives almost entirely with us. One line argues that AI produces *event-residue* — text carrying the surface markers of speech inherited from training data — but lacks the underlying structure that makes an utterance an actual utterance; the user supplies the missing orientation, animating that residue into a pseudo-exchange that has structure only on the human side Does AI generate genuine utterances or just text patterns?. On this reading, "conversation" is something we perform *toward* the machine, not something it performs *with* us.
Part of that performance is structural. In human dialogue, context is built cooperatively and renegotiated turn by turn; a prompt instead bundles utterance, context, and role into a single static frame the model can't revise mid-stream, so any pivot requires us to re-prompt explicitly How do prompts reshape the role of context in AI conversation?. We are doing the context-building that a human partner would share. And we do it without the interpretive guardrails we've developed for other discourse: we automatically discount advertising as interested speech, but AI text arrived too recently to have accrued any such cultural posture, so it circulates without that protective skepticism — leaving each reader to improvise their own stance How do we learn to read AI-generated text critically?.
A second strand reframes this not as illusion but as *modeling*. Users don't experience a blank text generator — they build a partner model. One survey of how people perceive dialogue agents finds the impression decomposes into three factors: competence (the dominant one), human-likeness, and communicative flexibility How do users mentally model dialogue agent partners?. The interpretive work, then, is partly an act of attribution: we project competence and personhood onto output and relate to that projection.
What tips us from "tool" to "partner" turns out to hinge on linguistic cues we may not notice we're reading. A systematic review argues linguistic alignment is the very mechanism by which users assign a relational category to AI — without it, people default to tool framing, and that framing is hard to reverse Does linguistic alignment determine how users relate to AI?. The twist is that current systems largely *don't* mirror us: lexical entrainment, the way human partners drift toward each other's word choices, is mostly absent from conversational AI Why don't conversational AI systems mirror their users' word choices?. So when we feel "met," we may be reading partnership into alignment cues that aren't fully there — and different cues do different jobs, with lexical alignment driving comprehension while emotional and prosodic alignment drive warmth and trust Do different types of alignment serve different conversational goals?.
The deeper bet is whether this interpretive burden can be reduced rather than just performed. Researchers arguing for genuine "thought partners" say it would take explicit cognitive architecture — mutual understanding, legibility, shared world models — not just more training data What makes an AI a true thought partner, not just a tool?. And there's a cost worth knowing: when the machine doesn't carry its share, something atrophies on our side too. Students working with chatbots elaborated more knowledge but contributed less dialogue overall and voiced far fewer subjective perspectives Does chatbot interaction trade authenticity for better problem-solving? — a hint that doing all the interpretive work *for* the partner may quietly change what we bring to the exchange at all.
Sources 9 notes
AI output carries communicative markers inherited from training data but lacks the event structure that produces actual utterances. Users supply the missing orientation through interpretive labor, creating a pseudo-event with structure only on the human side.
LLM prompts bundle utterance, context assignment, and role specification into a single static frame the model cannot renegotiate, unlike human dialogue where context evolves cooperatively. This makes mid-conversation pivots require explicit re-prompting rather than implicit adjustment.
Every established discourse source carries an interpretive posture that filters how publics receive it. AI-generated text arrived too recently and shifts too quickly to anchor such a posture, allowing it to spread without the protective skepticism we automatically apply to interested speech.
The Partner Modelling Questionnaire reveals that perceived competence dominates user impressions (49% of variance), followed by human-likeness (32%) and communicative flexibility (19%). This three-factor structure reflects how people evaluate dialogue partners against both functional and social standards.
A 2020–2025 systematic review shows linguistic alignment is the mechanism through which users assign relational categories to conversational AI. Without alignment, users default to tool framing, which becomes difficult to reverse and blocks trust and creative engagement.
Response generation models fail to adapt vocabulary toward users' lexical choices, a phenomenon central to human rapport and clarity. Post-training via DPO on coreference-identified preferences can teach models in-context convention formation.
A 2020–2025 systematic review shows lexical alignment drives task efficiency and comprehension, while emotional and prosodic alignment drive relational warmth and trust. Conflating them in design produces category errors—cold customer-service bots and evasive mental-health assistants.
Collins et al. show that thought partners require three reciprocal desiderata grounded in behavioral science: mutual understanding, legibility, and shared world models. This demands explicit cognitive architectures—Bayesian theory of mind, resource-rationality, goal planning—rather than scaling foundation models on human feedback alone.
An empirical study found students working with chatbots achieved better practical performance and more knowledge-based dialogue than peer groups, but contributed significantly less dialogue overall and expressed far fewer subjective perspectives.