Puetz2025

From emcawiki
Revision as of 03:29, 26 January 2026 by JakubMlynar (talk | contribs) (Created page with "{{BibEntry |BibType=ARTICLE |Author(s)=Ole Pütz; |Title=Co-textual dopes: How LLMs produce contextually appropriate text in chat interactions with humans without access to co...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Puetz2025
BibType ARTICLE
Key Puetz2025
Author(s) Ole Pütz
Title Co-textual dopes: How LLMs produce contextually appropriate text in chat interactions with humans without access to context
Editor(s)
Tag(s) EMCA, AI Reference List, Conversation analysis, Chatbots, LLM
Publisher
Year 2025
Language English
City
Month
Journal Linguistic Anthropology
Volume 36
Number 1
Pages e70036
URL Link
DOI 10.1111/jola.70036
ISBN
Organization
Institution
School
Type
Edition
Series
Howpublished
Book title
Chapter

Download BibTex

Abstract

This paper asks how LLM-based systems can produce text that is taken as contextually appropriate by humans without having seen text in its broader context. To understand how this is possible, context and co-text have to be distinguished. Co-text is input to LLMs during training and at inference as well as the primary resource of sense-making for humans in interaction, collaboratively produced by both human and machine during chat. Systems can also passively participate in contextualization, insofar as cues are found in the co-text and the user guides the system as to what is the appropriate context for them.

Notes

Dieser Artikel fragt, wie LLM-basierte Systeme Texte erstellen können, die von Menschen als kontextuell angemessen empfunden werden, obwohl sie Texte nie in einem umfassenderen Kontext sehen. Um zu verstehen, wie dies möglich ist, ist eine Unterscheidung zwischen Kontext und Kotext notwendig. Der Kotext ist Input für LLMs während des Trainings und der Inferenz, sowie die primäre Ressource für die Zuschreibung von Sinn in der Interaktion; der Kotext wird im Chat gemeinsam von Mensch und Maschine erzeugt. Systeme können sich zudem passiv an Kontextualisierungsprozessen beteiligen, sofern sich Cues im Kotext finden lassen und Nutzende das System hinsichtlich des für sie relevanten Kontexts anleiten.