All Roads Lead to Polenta: Cultural Attractors at the Junction of Public and Personal Culture

Abstract

In the process of retelling information, individuals often inadvertently transform it to be more consistent with their cultural schemas. We explore the long-term cultural change inherent in this process, focusing on utterances about cultural tastes as our case study (e.g., music, food, and outdoor hobbies). We use a word embedding model to simulate a "telephone game" where each actor partially hears an utterance, uses their cultural schemas to guess the missing word, and tells the result to the next actor. While laboratory "telephone games" explore short transmission chains of approximately four steps, our approach lets us simulate these chains out to 1000 steps. We find that these chains are often pulled toward powerful "cultural attractors"-essentially points of least resistance where communications end up through transmission error alone. Moreover, some attractors operate across taste domains: transmission chains gravitate toward these attractors regardless of which cultural domain they begin in. The most powerful such attractor we located concerns high-status, broadly liked food. Taste in food may thus have an underappreciated centrality within personal taste: verbal accounts describing taste in food may be particularly stable across multiple retellings, while accounts about other taste domains may become transformed into accounts of taste in food.

Machine Learning as a Model for Cultural Learning: Teaching an Algorithm What it Means to be Fat

“…For example, in an agent-based model, each agent's "meaning-making" may be operationalized as a neural word embedding trained on some cultural diet. Or, for example, embeddings may be used to model agents' schemas in transmission chains, to identify meanings that may be more or less resistant to change across interaction (Boutyline, Cornell, and Arseniev-Koehler 2021).…”

“…Public culture is considered a key source of influence for how individuals make sense of their own and others' body weight and shape (e.g., Grabe et al 2008; López-Guimerà et al 2010; Ravary et al 2019). Indeed, a vast body of work demonstrates that Americans hold a range of pejorative and stigmatizing meanings around fatness (e.g., Azevedo et al 2014; Carels et al 2013; Charlesworth and Banaji 2019; Hinman et al 2015; Lieberman, Tybur, and Latner 2012; Nosek et al 2007; Ravary et al 2019; Schupp and Renner 2011; Schwartz et al 2006; Teachman et al 2003; Teachman and Brownell 2001). Thus, the social construction of obesity is an important and well-documented case to consider how public culture becomes personal culture.…”

Sociolinguistic Perspectives on Machine Learning with Text Data

“…Word embeddings, most often the word2vec (Mikolov et al, 2013) or GloVe (Pennington et al, 2014) varieties, have been used by social scientists to model cultural meaning, representations of intersectionality, and variation in meaning by author income (Kozlowski et al, 2019;Nelson 2021;Arthurs & Alvero, 2020). Sociologists have also been active in developing new word embedding methods, such as concept movers distance to compare how concept meanings in documents (Stoltz & Taylor, 2019), combining topic modeling and word embedding approaches , and showing how cultural associations between words can serve as attractors between words and concepts (Boutyline et al, 2021). Most of these studies leverage large word embedding datasets that were trained on massive amounts of text, such as all of the text on Wikipedia, the entire Google Books corpus, or every digitized newspaper article ever printed in the US.…”

A Tool Kit for Relation Induction in Text Analysis

“…22 Alternatively, we may want to down-weight frequent words. Therefore, we could use the inverse of word frequency (Boutyline et al 2021, Arora et al 2016b, Karipbayeva et al 2019.…”