Do U.S. and Indian Twitter differ in what users share about others?
This blog post summarizes the paper “A Tale of Two Cultures: Comparing Interpersonal Information Disclosure Norms on Twitter” which unveil and compares what is culturally appropriate to share on social media about people you know. This work will be presented at the 26th ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW 2023). This paper will also be published in The Proceedings of the ACM on Human Computer Interaction. You may access the paper here.
Cultural background heavily impacts human behavior and sharing of information. In particular, cultures are likely to differ when considering what information is appropriate to share about themselves or others (e.g., about their parents). With the increasingly global reach of social media, mismatched communication styles and disclosure norms become quite probable — -they can lead to misunderstanding, hindering the establishment of new relationships across cultures, or worse, leading to unintentional conflict.
Sociologists track much cultural difference in disclosing information to individualism (focusing on rights and concerns of individuals) vs. collectivism (stressing importance of community). In fact, recent sociological research uncovered how information disclosure norms differ for U.S or European participants. Unfortunately, little work goes beyond the global north and usually focuses on information the user reveals about themselves. However, social media users also often post about the people they know — -about relationships that range from family to strangers they meet on street. To that end, our work focuses on how cultures from global south and global north differ during Interpersonal Information Disclosure, i.e., disclosing information about someone else on Twitter. We focus on India and U.S., two large countries (representative of different cultures) from global south and global north respectively.
We faced a challenge even while planning this study — it was not clear what are the relationship words used and how they should be grouped into relationship categories in each culture. So, we created a survey and deployed it to users from India and U.S. to identify the key relationship words and their broad relationship categories. We started seeing differences even in how the words are grouped — a typical example is the relationship word uncle, it is considered Family in India, but Extended Family in the US.
Having identified these relationship words and their categories, next we designed and deployed novel automated techniques to collect more than 2.3 million tweets posted by users from India and U.S. which talk about these interpersonal relationships. We then started analyzing the content of tweets talking about interpersonal relationships to identify what users disclose about other people in each culture.
We first eyeballed the topics extracted from these tweets using automated topic detection coupled with manual analysis. We show a side-by-side comparison of topics in Table 1. Some topics are the same (shown in gray in Table 1). However, there were marked difference too, e.g., ’Patriotism’ and Expressing love’ seem to be particular to the disclosure norms of India. However, the tweets “Expressing love” are often towards kids, best friends, but almost never towards their partner, hinting at differing norms of what information is appropriate to share about personal relationships.
Our analysis of the emotions of these tweets (detected using deep learning) further found interesting analogous patterns of emotional expression across different relationship groupings within each culture. Furthermore, when we focused on tweets with potentially sensitive financial or location information we found that disclosure patterns vary dramatically for tweets containing disclosures about someone else’s sensitive data, as opposed to disclosures about sensitive data in general. Strikingly, 30.6% of location-related Indian tweets involved memories of some form, as contrasted with only 6.6% of U.S. location tweets.
Our paper contains more unexpected, interesting (and fun) norms of what is culturally appropriate to reveal about people social media users know.
Our findings encourage social media designers coming from individualist societies (e.g., U.S.) to consciously consider the collectivist cultural norms that may be relevant for many of their users rather than relying on default assumptions that may be deeply rooted in individualistic norms. Designers might even consider culturally-aware design of recommendation algorithms driving the display of specific posts or meta-information in users’ input streams. For example, posts expressing anger about coworkers may not be surprising to U.S. Twitter users, but may perhaps be culturally jarring to Indian users.
