Opinion dynamics of online social network users: a micro-level analysis

ABSTRACT

In this paper, we present an empirical study of the opinion dynamics of a large-scale sample of online social network users. We estimate users’ opinions as continuous scalars based on their subscriptions to information sources and analyze how friendship connections affect the dynamics of these estimations. Distinguishing between positive (toward friends’ opinions) and negative (away from friends’ opinions) opinion shifts, we find that the existence and magnitude of both types of shifts are positively related (largely through linear or inverted U-shaped form) to the distance in opinions between a user and their friends. The distance additionally moderates the balance between positive and negative movements: if the distance is within a certain moderate range, there is a relatively high chance of a positive shift.

KEYWORDS:

• Opinion dynamics
• online social networks
• repulsive influence

Acknowledgments

The authors are grateful to anonymous reviewers for their invaluable comments.

Notes

¹ In the general case, the probability of acceptance and the magnitude of opinion shift could follow different assumptions and thus take different mathematical forms. For example, the probability of acceptance could depend on $\left|x_i-x_j\right|$ linearly with a positive slope (linear assumption) while the magnitude of opinion shift may follow the moderated assumption..

² If the probability of acceptance and the magnitude of opinion shift both depend on $\left|x_i-x_j\right|$, then the effect on the expected opinion shift could be nonlinear. For example, if the probability of acceptance is defined as a linear function of $\left|x_j-x_i\right|$ and we assume the positive linear influence assumption in (1), then the expected opinion shift will depend quadratically on $\left|x_j-x_i\right|$.

³ Moussaïd et al. (2013) maintained also that the chances of opinion shifts follow the positive moderated assumption.

⁴ This observation does not contradict previously obtained results on EPOC and conditional magnitude – a puzzle may be because, after multiplication of two linear/quadratic functions, we should get a quadratic/cubic/tetrad function. The reason is that all curves that represent EPOC and conditional magnitude do not have a strictly linear/quadratic form.

⁵ https://www.bbc.com/news/world-europe-45342721

⁶ Note that an individual whose friends’ opinions are not different from their own is not necessarily in an echo-chamber. To ensure this, one must additionally claim that their friends also tend to be isolated, and friends of their friends are as well, and so on.

⁷ https://meduza.io/en

⁸ https://www.rt.com/

⁹ https://archives.cjr.org/feature/what_is_russia_today.php

¹⁰ https://www.rt.com/op-ed/377715-rt-gazillionaire-media-funding/, https://www.rt.com/russia/443562-meduza-editor-sexual-harassment/

Additional information

Funding

This work was supported by the Russian Foundation for Basic Research, project no. 18-29-22042.