Deixis machines and enunciation without a speaker in large language models

How do large language models (LLMs) enact deixis and enunciation while lacking the embodied and temporal continuity that, in Benveniste’s sense, constitutes a speaker? Drawing on Agamben’s analysis of enunciation as suspended deixis and Viveiros de Castro’s account of perspectivism as cosmological deixis, this article examines what the displacement of the deictic origo reveals about language, agency, and responsibility in human–machine interaction. When LLMs generate expressions such as “I” and “you,” they organize interpretation and obligation without grounding them in any embodied or fictional subject that endures across time. Unlike literature, theatre, or ritual, which stabilize deixis within narrative or symbolic frames, LLMs leave deictic expressions unanchored, producing transient forms of address that dissolve with each exchange. In this sense, LLMs function as deixis machines: systems that reproduce the structure of enunciation while lacking an embodied and accountable subject of enunciation or a stabilizing frame of interpretation and attribution. To investigate the effects of unanchored deixis, we conducted a pilot study in which deictic framings were systematically varied across six ethical dilemmas and three LLMs, producing 162 responses. The results indicate that pronoun shifts reorganize moral reasoning: first-person prompts produce consequentialist trade-offs, second-person formulations emphasize duties, reflexive constructions encourage self-examination, and cosmological framings invoke universal or virtue-based justifications. In LLM discourse, deictic framing can shape normative reasoning independently of alignment procedures or regulatory constraints. Deictic force persists even in the absence of an accountable subject or a stabilized order of attribution. To address this condition, we introduce the figure of the machine shaman to name the mediating function through which machine-generated address is interpreted, situated, and redistributed among the actors implicated in its circulation. Rather than attempting to recover a fictive origin or relying on regulatory control alone, the machine shaman designates the practices through which deictic force is interpreted, responsibility for responding to it is allocated, and positions of speaking, responding, and responsibility are organized.

Authors

Institutions

Publication Details

Journal
Discover Artificial Intelligence
Published
2026-04-01
DOI
https://doi.org/10.1007/s44163-026-01157-8
Primary Topic
Language, Metaphor, and Cognition
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Deixis machines and enunciation without a speaker in large language models

Denisa Kera
Discover Artificial Intelligence
Language, Metaphor, and Cognition
article

Deixis machines and enunciation without a speaker in large language models

Denisa Kera
article en

Abstract

How do large language models (LLMs) enact deixis and enunciation while lacking the embodied and temporal continuity that, in Benveniste’s sense, constitutes a speaker? Drawing on Agamben’s analysis of enunciation as suspended deixis and Viveiros de Castro’s account of perspectivism as cosmological deixis, this article examines what the displacement of the deictic origo reveals about language, agency, and responsibility in human–machine interaction. When LLMs generate expressions such as “I” and “you,” they organize interpretation and obligation without grounding them in any embodied or fictional subject that endures across time. Unlike literature, theatre, or ritual, which stabilize deixis within narrative or symbolic frames, LLMs leave deictic expressions unanchored, producing transient forms of address that dissolve with each exchange. In this sense, LLMs function as deixis machines: systems that reproduce the structure of enunciation while lacking an embodied and accountable subject of enunciation or a stabilizing frame of interpretation and attribution. To investigate the effects of unanchored deixis, we conducted a pilot study in which deictic framings were systematically varied across six ethical dilemmas and three LLMs, producing 162 responses. The results indicate that pronoun shifts reorganize moral reasoning: first-person prompts produce consequentialist trade-offs, second-person formulations emphasize duties, reflexive constructions encourage self-examination, and cosmological framings invoke universal or virtue-based justifications. In LLM discourse, deictic framing can shape normative reasoning independently of alignment procedures or regulatory constraints. Deictic force persists even in the absence of an accountable subject or a stabilized order of attribution. To address this condition, we introduce the figure of the machine shaman to name the mediating function through which machine-generated address is interpreted, situated, and redistributed among the actors implicated in its circulation. Rather than attempting to recover a fictive origin or relying on regulatory control alone, the machine shaman designates the practices through which deictic force is interpreted, responsibility for responding to it is allocated, and positions of speaking, responding, and responsibility are organized.

Discover Artificial Intelligence
Bar-Ilan University (IL)
Openalex Percentile: Top 25%
Language, Metaphor, and Cognition
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.