Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Othering and low status framing of immigrant cuisines in US restaurant reviews and large language models

Yiwei Luo, Kristina Gligorić, Dan Jurafsky

Abstract

Identifying implicit attitudes toward food can mitigate social prejudice due to food's salience as a marker of ethnic identity. Stereotypes about food are representational harms that may contribute to racialized discourse and negatively impact economic outcomes for restaurants. Understanding the presence of representational harms in online corpora in particular is important, given the increasing use of large language models (LLMs) for text generation and their tendency to reproduce attitudes in their training data. Through careful linguistic analyses, we evaluate social theories about attitudes toward immigrant cuisine in a large-scale study of framing differences in 2.1M English language Yelp reviews. Controlling for factors such as restaurant price and neighborhood racial diversity, we find that immigrant cuisines are more likely to be othered using socially constructed frames of authenticity (e.g., "authentic," "traditional"), and that non-European cuisines (e.g., Indian, Mexican) in particular are described as more exotic compared to European ones (e.g., French). We also find that non-European cuisines are more likely to be described as cheap and dirty, even after controlling for price, and even among the most expensive restaurants. Finally, we show that reviews generated by LLMs reproduce similar framing tendencies, pointing to the downstream retention of these representational harms. Our results corroborate social theories of gastronomic stereotyping, revealing racialized evaluative processes and linguistic strategies through which they manifest.

Othering and low status framing of immigrant cuisines in US restaurant reviews and large language models

Abstract

Identifying implicit attitudes toward food can mitigate social prejudice due to food's salience as a marker of ethnic identity. Stereotypes about food are representational harms that may contribute to racialized discourse and negatively impact economic outcomes for restaurants. Understanding the presence of representational harms in online corpora in particular is important, given the increasing use of large language models (LLMs) for text generation and their tendency to reproduce attitudes in their training data. Through careful linguistic analyses, we evaluate social theories about attitudes toward immigrant cuisine in a large-scale study of framing differences in 2.1M English language Yelp reviews. Controlling for factors such as restaurant price and neighborhood racial diversity, we find that immigrant cuisines are more likely to be othered using socially constructed frames of authenticity (e.g., "authentic," "traditional"), and that non-European cuisines (e.g., Indian, Mexican) in particular are described as more exotic compared to European ones (e.g., French). We also find that non-European cuisines are more likely to be described as cheap and dirty, even after controlling for price, and even among the most expensive restaurants. Finally, we show that reviews generated by LLMs reproduce similar framing tendencies, pointing to the downstream retention of these representational harms. Our results corroborate social theories of gastronomic stereotyping, revealing racialized evaluative processes and linguistic strategies through which they manifest.
Paper Structure (11 sections, 3 equations, 15 figures, 10 tables)

This paper contains 11 sections, 3 equations, 15 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Data and Methods
  3. Results
  4. Study 1A: Othering of immigrant cuisines in Yelp
  5. Study 2: Low status framing of non-European cuisines
  6. Study 3: Reporting bias in LLMs
  7. Discussion
  8. Broader perspective
  9. Further related work
  10. Acknowledgments
  11. Appendix

Figures (15)

  • Figure 1: Example reviews showing frames detected in our analysis ( blue: luxury; green: prototypicality; pink: hygiene; gray: authenticity). Both customers gave 4 stars and both restaurants are designated as $ (on the 4-point scale $ to $$$$) with the same mean rating.
  • Figure 2: Distribution of restaurants over cuisine, state.
  • Figure 3: Distribution of restaurants along price point (corresponding to Yelp-designated $), mean star rating, neighborhood income (2020 USD), and neighborhood racial diversity from 2020 (computed as the Simpson Diversity Index).
  • Figure 4: Othering of immigrant cuisine: Linear regression coefficients predicting othering in review text from cuisine region, showing immigrant cuisines receive more othering. Note the different y-scale for Authenticity framing. Error bars are 95% CIs. Significance values for US (the reference level) indicates whether being US has a significant effect on the outcome variable, other values indicate whether a cuisine is significantly different from US with respect to its effect on the outcome. ns=p$>$0.05, **=p$<$0.01, ***=p$<$0.001.
  • Figure 5: Othering by outsiders: Linear regression coefficients predicting othering of As/ Lat within neighborhoods with a high/low % (i.e. above/below the median) of Asian and Hispanic residents. Error bars are 95% CIs. Significance determined by a Wald test comparing coefficients within the same model. ns=p$>$0.05, **=p$<$0.01, ***=p$<$0.001.
  • ...and 10 more figures