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

Sunny Rai, Khushang Zilesh Zaveri+4 more|arXiv (Cornell University)|Feb 17, 2024

South Asian Cinema and CultureArts and Humanities3 citations

TL;DR

This paper proposes a culture-agnostic, emotion-based approach to discover social norms using self-conscious emotions—shame and pride—in Bollywood and Hollywood films. By analyzing 5.4K movie dialogues and extracting 10,000 norms, it reveals that Indian cinema emphasizes shame for role deviation and pride in family honor, while American cinema focuses on shame for poverty and pride in ethical conduct, with women facing disproportionate social sanctions across both cultures.

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 large language models. 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.

Motivation & Objective

• To address the cultural bias in existing social norm datasets, which are predominantly Anglocentric and derived from Western, English-language sources.
• To investigate how self-conscious emotions—shame and pride—can serve as proxies for uncovering unspoken social norms across cultures.
• To examine cross-cultural variations in normative expectations between India (collectivist) and the USA (individualist), focusing on moral emotions.
• To analyze gender disparities in the attribution of shame and pride in cinematic portrayals across cultures.
• To release the first large-scale, cross-cultural dataset of self-conscious emotions and associated social norms, enabling more equitable AI alignment.

Proposed method

• A two-step approach: (a) a vocabulary-based search for words related to shame and pride in movie transcripts, followed by (b) prompting a large language model (LLM) to extract underlying social norms from context.
• The dataset was constructed from 5.4K movie dialogues (2,738 Bollywood, 2,697 Hollywood), with 10,000 norms extracted after filtering and validation.
• Native speaker validation was used to ensure norm accuracy and cultural relevance, with a focus on linguistic and emotional context.
• Linguistic Inquiry and Word Count (LIWC) analysis was applied to assess emotional valence and linguistic patterns in normative expressions.
• A control group of dialogues without shame/pride words was used to compare normative expression patterns.
• The study used a comparative framework to contrast normative expectations in collectivist (India) versus individualist (USA) cultural contexts.

Experimental results

Research questions

• RQ1 How do linguistic expressions of shame and pride differ between Bollywood and Hollywood films?
• RQ2 To what extent do cultural differences in collectivism versus individualism shape normative expectations as reflected in cinematic portrayals?
• RQ3 How are shame and pride distributed across genders in Indian and American cinema?
• RQ4 Can self-conscious emotions serve as reliable indicators for discovering culture-specific social norms?
• RQ5 How do the norms extracted from movies compare to real-world social expectations, and what are the risks of using cinematic data for AI alignment?

Key findings

• Bollywood films emphasize shame for violating traditional social roles and pride in family honor, while Hollywood emphasizes shame for poverty and incompetence, and pride in ethical behavior.
• Women are more frequently targeted by shame than men in both cultures, with a negligible difference in shame frequency (0.18 in Bollywood vs. 0.16 in Hollywood).
• The gender gap in pride-related dialogues is similar across both cultures, with men receiving more pride-related expressions than women.
• The study identified 10,000 social norms from 5.4K movie dialogues, forming the largest known cross-cultural dataset of self-conscious emotions and associated norms.
• LIWC analysis confirmed that shame and pride are strong indicators of normative expectations, supporting their use in reducing cultural bias in LLMs.
• Despite limitations in temporal and regional representativeness, the dataset reveals consistent patterns of social sanctions toward women and cultural dichotomies in normative values.