Paper Image

Observations of AI-generated replies to identity-related language

Published on:

23 October 2023

Primary Category:

Computation and Language

Paper Authors:

Li Lucy,

Su Lin Blodgett,

Milad Shokouhi,

Hanna Wallach,

Alexandra Olteanu

Bullets

Key Details

Tests AI system behaviors when identity-related language features are varied in inputs

Finds tensions between adapting vs being invariant to these features

People's expectations for system behaviors diverge

Some want adherence to social norms and cultural sensitivity

Others prefer generic replies to avoid assumptions

Explore the topics in this paper

AI generated summary

Observations of AI-generated replies to identity-related language

This paper examines how AI systems respond when identity-related language features like names, roles, dialect, and style are varied in system inputs. Through 5 case studies, it finds tensions between adapting outputs to these features versus behaving invariantly. People's expectations also diverge: some want systems to follow social norms and be sensitive to cultural differences, while others prefer generic replies to avoid assumptions.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up