Language dynamics of immigration discourse
Title: Language dynamics of immigration discourse
SNIC Project: SNIC 2022/22-297
Project Type: SNIC Small Compute
Principal Investigator: Miriam Hurtado Bodell <miriam.hurtado.bodell@liu.se>
Affiliation: Linköpings universitet
Duration: 2022-04-01 – 2023-04-01
Classification: 50401
Keywords:

Abstract

This project is meant to be used as a resource for studying the language dynamics between the news, social media users, and politicians. Based on Steven Pinker (1994): "People invent new polite words to refer to emotionally laden or distasteful things, but the euphemism becomes tainted by association and the new one that must be found acquires its own negative connotations." We will test this hypothesis by (1) studying the similarity of language use when discussing immigration between 3 corpora; (i) Swedish national newspapers; (ii) Swedish parliament speeches, and (iii) social media posts from the online discussion forum Flashback and Twitter between 2014 and 2018. (2) find descriptive statistics of how patterns of convergence is a response to (a) the actions of the ego-actor, (b) the actions of the alter-actor, or (c) the actions of both. Lastly, (3) we will explore if social influence can be proven to be the causal driver of the descriptive dynamics patterns described in (2). We conduct (1) by identifying trigrams used by different which include known immigration keywords. We can extract the "context words" surrounding the immigration keyword in the trigram and compare their distribution among different actors. We compare the context word distribution by Spearman rank correlation (2) By studying the changing rank of context words among different actors, we can identify how many context words are increasingly/decreasingly used by the ego/alter/both actor(s). Studying these different categories over time and between different groups allow us to describe the dynamics generating the similarities found in (1). We test (3) to find evidence that an ego actor (i) are influenced by what an alter actor (j), by contrasting time periods where j "treats" i with a change in language with non-treatment periods. This requires running a language model per actor and time point, a resource-intensive task. Understanding the diffusion of new language helps us understand the role of opinion homophily in language use and the politicizing power of languages.