Papers
Topics
Authors
Recent
Search
2000 character limit reached

Social Norms in Cinema: A Cross-Cultural Analysis of Shame, Pride and Prejudice

Published 17 Feb 2024 in cs.CY | (2402.11333v5)

Abstract: Shame and pride are social emotions expressed across cultures to motivate and regulate people's thoughts, feelings, and behaviors. In this paper, we introduce the first cross-cultural dataset of over 10k shame/pride-related expressions, with underlying social expectations from ~5.4K Bollywood and Hollywood movies. We examine how and why shame and pride are expressed across cultures using a blend of psychology-informed language analysis combined with LLMs. We find significant cross-cultural differences in shame and pride expression aligning with known cultural tendencies of the USA and India -- e.g., in Hollywood, shame-expressions predominantly discuss self whereas shame is expressed toward others in Bollywood. Women are more sanctioned across cultures and for violating similar social expectations.

Citations (3)

Summary

  • The paper introduces a method for extracting culture-specific social norms in cinema by analyzing 43 million tokens from over 5,400 films.
  • It employs LIWC and GPT-4 based clustering to identify around 10,000 norms, highlighting differences in gender roles, family honor, and individual achievements.
  • The findings provide actionable insights for developing culturally aligned NLP systems and sensitive AI applications in diverse film contexts.

A Cross-Cultural Analysis of Social Norms in Bollywood and Hollywood Movies

This paper provides a detailed examination of how social norms are variably expressed in cultural contexts, specifically through Bollywood and Hollywood films. The study proposes a culture-agnostic approach for norm discovery using moral emotions—specifically shame and pride—as proxies to unearth and compare cultural norms embedded within cinematic expressions. The research introduces a comprehensive dataset derived from over 5,400 films and identifies approximately 10,000 social norms, marking a significant contribution to the understanding of cultural influence on normative expectations.

Approach and Methodology

The authors employ a multifaceted methodology to achieve their research aims:

  1. Data Collection and Preprocessing: The study meticulously gathers subtitles from over 5,000 Bollywood and Hollywood movies to form a dataset rich in examples of social interactions. A total of approximately 43 million tokens are analyzed, segmented into specific dialogues linked to shame and pride.
  2. Linguistic Analysis via LIWC: Linguistic Inquiry and Word Count (LIWC) is used to measure the linguistic expressions correlating with shame and pride in the dialogues. This enables an understanding of psychosocial categories and their correlation strength in expressions tied to shame and pride across different cultural contexts.
  3. LLMs: GPT-4 is utilized to extract reasons for expressions of shame and pride, mitigating cultural bias inherent in LLM outputs that predominantly reflect Western ethical contexts. This involves careful prompt engineering to capture normative expectations rather than directly outputting culturally biased "norms."
  4. Clustering and Identification of Themes: Agglomerative clustering on extracted norms identifies major themes that characterize the normative differences and similarities between the two cultures represented by Bollywood and Hollywood movies.

Key Findings

Cross-Cultural Linguistic Expressions

The linguistic analysis reveals distinct cultural patterns:

  • Shame: In Hollywood, shame is more self-centered with a past-focus and strongly associated with negative emotions and incompetence. Conversely, Bollywood depicts shame as other-centered, tied to social roles and family honor, with a present-focus and anger-linked associations.
  • Pride: Hollywood associates pride with individual achievements and duties, whereas Bollywood ties pride to collective achievements, family honor, and national identity.

Social Norms and Themes

The study identifies a range of social norms, highlighting cultural dichotomies:

  • Shame Norms: Bollywood often associates shame with gender roles and family expectations, while Hollywood links it to incompetence and poverty.
  • Pride Norms: Bollywood shows pride in the context of collective success and relational roles, while in Hollywood, pride is related to personal achievements and ethical behavior.

Gender Analysis

The research highlights the gendered nature of shame and pride:

  • Both cultures show a tendency towards shaming females more than males for deviations from social norms, particularly relating to sexuality in Bollywood.
  • While males are often depicted with pride concerning achievements and valor, subtle cultural nuances in Bollywood attribute family-centered pride to females.

Implications and Speculations

The implications of these findings extend to practical and theoretical domains:

  • Culturally Aligned NLP Systems: Echoing the need for culturally nuanced NLP systems, the research underlines the importance of embedding cultural sensitivities in model training processes, crucial for applications in multilingual LLMs.
  • Moral Emotion Data Utilization: By conjecturing that moral emotions could serve as effective indicators of culture-specific social norms, the findings suggest an increased role for nuanced emotional datasets in enhancing cultural understanding in AI.

In conclusion, this paper illuminates the complexity and depth of cross-cultural variations in social norms through the lens of cinema. The unique integration of moral emotions in capturing these distinctions offers valuable pathways for future AI research and applications, stressing the necessity for culturally diverse and sensitive machine understanding. As global interactions increase, research such as this will become indispensable in fostering multicultural alignment and understanding in technological development.

Paper to Video (Beta)

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We found no open problems mentioned in this paper.

Collections

Sign up for free to add this paper to one or more collections.

Tweets

Sign up for free to view the 11 tweets with 131 likes about this paper.