Social diversity for reducing the impact of information cascades on social learning

TL;DR

This paper investigates how social diversity can mitigate the negative effects of information cascades in social learning, enhancing collective decision-making in online social networks.

Contribution

It introduces the concept that social diversity's stochasticity can improve collective inference by reducing herd behavior and information cascades.

Findings

Social diversity decreases the likelihood of information cascades.

Increased social diversity improves collective inference accuracy.

Situations where social diversity is not beneficial are identified.

Abstract

Collective behavior in online social media and networks is known to be capable of generating non-intuitive dynamics associated with crowd wisdom and herd behaviour. Even though these topics have been well-studied in social science, the explosive growth of Internet computing and e-commerce makes urgent to understand their effects within the digital society. In this work we explore how the stochasticity introduced by social diversity can help agents involved in a inference process to improve their collective performance. Our results show how social diversity can reduce the undesirable effects of information cascades, in which rational agents choose to ignore personal knowledge in order to follow a predominant social behaviour. Situations where social diversity is never desirable are also distinguished, and consequences of these findings for engineering and social scenarios are discussed.