Interindividual Variation Refuses to Go Away: A Bayesian Computer Model of Language Change in Communicative Networks

Abstract

Treating the speech communities as homogeneous entities is not an accurate representation of reality, as it misses some of the complexities of linguistic interactions. Inter-individual variation and multiple types of biases are ubiquitous in speech communities, regardless of their size. This variation is often neglected due to the assumption that “majority rules,” and that the emerging language of the community will override any such biases by forcing the individuals to overcome their own biases, or risk having their use of language being treated as “idiosyncratic” or outright “pathological.” In this paper, we use computer simulations of Bayesian linguistic agents embedded in communicative networks to investigate how biased individuals, representing a minority of the population, interact with the unbiased majority, how a shared language emerges, and the dynamics of these biases across time. We tested different network sizes (from very small to very large) and types (random, scale-free, and small-world), along with different strengths and types of bias (modeled through the Bayesian prior distribution of the agents and the mechanism used for generating utterances: either sampling from the posterior distribution [“sampler”] or picking the value with the maximum probability [“MAP”]). The results show that, while the biased agents, even when being in the minority, do adapt their language by going against their a priori preferences, they are far from being swamped by the majority, and instead the emergent shared language of the whole community is influenced by their bias.

How Network Structure Shapes Languages: Disentangling the Factors Driving Variation in Communicative Agents

“…While in the prior study, we defined centrality as betweenness centrality, we complemented these results here by exploring whether various other node-level metrics (local clustering coefficient, node degree, eigenvector centrality, and closeness centrality) might exert different influences. These findings, available only in the Supplementary Materials (see Part 3.4), not only expand upon but also validate the results presented by Josserand et al (2021).…”

“…Previous studies (see fig. 13 in Josserand et al, 2021) have shown that in a type of Bayesian model of language change similar to the ones implemented here but where the linguistic feature was binary, language stabilized before 1000 rounds in all network types (see our Supplementary Materials, Part 3.1 for visualization). Here, we introduced a conservative margin and analyzed the inter and intraindividual variability after 3000 rounds (when all agents had produced 3000 utterances) in order to make sure that language would have stabilized.…”

“…Seeing enough data should be able to overwhelm even very strong a priori beliefs and the order in which the utterances in the data arrive does not matter. This generalizes the prior model studied in Josserand et al (2021) by allowing more than two possible utterances (i.e., from a Bayesian model with a binary feature using Binomial/Beta conjugate distributions, to a multinomial feature using Multinomial/Dirichlet conjugate distribution).…”

“…The color gradient reflects the density of random draws, with red denoting high density and blue indicating low density. sampler (or "SAM") strategy (Griffiths & Kalish, 2007;Josserand et al, 2021) in which the agent picks any of the possible k utterances proportional to its subjective probability (as given by the agent's current Dirichlet distribution).…”

A bifurcation threshold for contact-induced language change

“…It should be emphasized that the model presented and analysed in this paper is concerned with only one potential external ("environmental") predictor of contact-induced change, namely the proportion of L2 speakers in a speech community. Other possible external factors conditioning language change have been considered in the literature, ranging from population size (Lupyan & Dale 2010;Nettle 2012;Koplenig 2019) to social network geometries (Ke & Gong & Wang 2008;Fagyal et al 2010;Kauhanen 2017;Josserand et al 2021) or combinations of such parameters (Trudgill 2004). Although the present model sets such effects aside, this is not to deny their importance.…”