Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies

Alicia DeVrio, Myra Cheng, Lisa Egede, Alexandra Olteanu, Su Lin Blodgett

TL;DR

This paper addresses the problem of conceptual ambiguity around anthropomorphism in language technologies. It develops an empirical, linguistically grounded taxonomy of textual expressions that can contribute to anthropomorphism, supported by 50 in-the-wild sources and 3954 annotations, plus a literature synthesis. The core contributions are 19 expression types organized under five guiding lenses (internal states, social positioning, materiality, autonomy, and communication skills) and a practical framework for researchers and designers to identify, measure, and mitigate harmful anthropomorphism in real-world outputs. The work advances precise discourse around humanness in AI text, enabling targeted interventions, design guidelines, and nuanced interpretation of system capabilities and limitations with potential impact on policy, UX, and AI safety research.

Abstract

Recent attention to anthropomorphism -- the attribution of human-like qualities to non-human objects or entities -- of language technologies like LLMs has sparked renewed discussions about potential negative impacts of anthropomorphism. To productively discuss the impacts of this anthropomorphism and in what contexts it is appropriate, we need a shared vocabulary for the vast variety of ways that language can be anthropomorphic. In this work, we draw on existing literature and analyze empirical cases of user interactions with language technologies to develop a taxonomy of textual expressions that can contribute to anthropomorphism. We highlight challenges and tensions involved in understanding linguistic anthropomorphism, such as how all language is fundamentally human and how efforts to characterize and shift perceptions of humanness in machines can also dehumanize certain humans. We discuss ways that our taxonomy supports more precise and effective discussions of and decisions about anthropomorphism of language technologies.

A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies

TL;DR

This paper addresses the problem of conceptual ambiguity around anthropomorphism in language technologies. It develops an empirical, linguistically grounded taxonomy of textual expressions that can contribute to anthropomorphism, supported by 50 in-the-wild sources and 3954 annotations, plus a literature synthesis. The core contributions are 19 expression types organized under five guiding lenses (internal states, social positioning, materiality, autonomy, and communication skills) and a practical framework for researchers and designers to identify, measure, and mitigate harmful anthropomorphism in real-world outputs. The work advances precise discourse around humanness in AI text, enabling targeted interventions, design guidelines, and nuanced interpretation of system capabilities and limitations with potential impact on policy, UX, and AI safety research.

Abstract

Recent attention to anthropomorphism -- the attribution of human-like qualities to non-human objects or entities -- of language technologies like LLMs has sparked renewed discussions about potential negative impacts of anthropomorphism. To productively discuss the impacts of this anthropomorphism and in what contexts it is appropriate, we need a shared vocabulary for the vast variety of ways that language can be anthropomorphic. In this work, we draw on existing literature and analyze empirical cases of user interactions with language technologies to develop a taxonomy of textual expressions that can contribute to anthropomorphism. We highlight challenges and tensions involved in understanding linguistic anthropomorphism, such as how all language is fundamentally human and how efforts to characterize and shift perceptions of humanness in machines can also dehumanize certain humans. We discuss ways that our taxonomy supports more precise and effective discussions of and decisions about anthropomorphism of language technologies.

Paper Structure

This paper contains 40 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Anthropomorphic Design in Technology
  4. Negative Impacts from Anthropomorphism of Technologies
  5. Types of Anthropomorphism of Technologies
  6. Methods
  7. Taxonomy
  8. Guiding Lenses
  9. Suggestive of internal states
  10. Suggestive of social positioning
  11. Suggestive of materiality
  12. Suggestive of autonomy
  13. Suggestive of communication skills
  14. Expressions Contributing To Anthropomorphism
  15. Expressions of intelligence
  16. ...and 25 more sections

Figures (1)

  • Figure 1: Overview of the five guiding lenses used in our taxonomy, along with examples of relevant quotes from our sample of cases and associated types of expressions present in those quotes. We emphasize that the same type of expression can be associated with more than one guiding lens, and text outputs can be associated with more than one type of expression.