Social media enhanced collective intelligence

This project aims to explore the extent to which it is possible to maximize “wisdom of crowd” effects within a network of collaborating analysts by measuring and exploiting socio-cognitive diversity within the network, where socio-cognitive diversity is inferred based on the content and patterns of analysts’ communications (in particular, social media communications).

Background

Wisdom of crowds (WoC) refers to a form of collective intelligence in which the aggregate judgment of a group is reliably superior to that of any one of its individual members [1]. WoC effects have been proposed to explain, for example, the underperformance of individually managed stock funds compared to passive index funds, the superiority of prediction markets versus political pundits in predicting election outcomes, and even the relative success of democracies over other forms of government [2].

For a crowd to be wise, however, it is essential that its members possess diverse information. The explanation linking diversity to collective intelligence is that analysts who possess diverse information are more likely to produce uncorrelated errors that cancel one another out, thereby yielding a collective judgment close to the truth [3]. Although one way of maximizing diversity is to have analysts work independently, this is often impractical given the increasingly collaborative nature of today’s analytic work environments [4]. But even if independence could be enforced, doing so would risk forfeiting the well-documented benefits of collaboration [5]. What is needed therefore are methods for identifying subgroups (or subnetworks) of diverse individuals within larger collaborative networks.

This project aims to develop socio-cognitive framework for understanding and modeling diversity in human social networks. We emphasize the social aspect of diversity because differences in individuals’ knowledge and viewpoints are often shaped as much by differences in their peer networks as by their individual experiences and thought processes [6] – such that it is possible to infer a person’s viewpoints in part by the company he or she keeps. Similarly, we stress the cognitive dimension of diversity because what a person knows, and how a person thinks, are measurable cognitive constructs that directly influence diversity of judgments across individuals.

Research Objectives

At a theoretical level, we seek to understand what are the appropriate dimensions (or features) for characterizing social-cognitive diversity in human networks. At an empirical level, we want to know which approaches for computing diversity produce metrics that are most predictive of collective intelligence. More specifically, we aim to answer following questions,

(1) Do current popular social media platforms such as Twitter contain sufficient data for measuring socio-cognitive diversity?

(2) Can socio-cognitive diversity be computed automatically using natural language processing (NLP) techniques and social network analysis?

Planned Contributions

Diversity analysis and effects on collective intelligence

The project primarily aims to develop theory-based methods for measuring diversity in socio-cognitive networks and demonstrate that such measures can be used to enhance the collective intelligence of the network. In service of these goals, Novel user modeling and clustering techniques are developed.

Social data clustering

User modeling and clustering can identify diverse set of people in a network. This project aims to investigate various user modeling techniques along dimensions such as social interactions, semantic content similarity, domain specific similarity etc. Efficient large scale graph clustering techniques are adopted as well as extended for clustering users based on different dimensions.

Dataset

Fantasy football dataset is chosen as an initial dataset because of the availability of prediction and outcome data in fantasy football. Users are allowed to select soccer players in their “fantasy” team. The players get a score based on their performance in soccer games which earns reward for users. Hence users are motivated to select player such that their reward is maximum. The project aims to investigate how the diversity among these users can help coming up with a better player choice.

People

PI

Brandon Minnery, Wright State Research Institute [1]

Co-PIs

Amit Sheth, Kno.e.sis, Wright State University [2]

Valerie Shalin, Kno.e.sis, Wright State University [3]

Researchers

Shreyansh Bhatt, Kno.e.sis. Wright State University [4]

Beth Bullemer, Wright State University

Software Engineer

Srikanth Nadella, Wright State Research Institute

Publications

1. Bhatt S., Minnery B., Nadella S., Bullemer B., Shalin V., Sheth A., (2017, August). Enhancing Crowd Wisdom Using Measures of Diversity Computed from Social Media Data. In Web Intelligence (WI), 2017 (To be appear)

Funding

This project is sponsored by Army Research Office Grant No. BAA# W911NF-12-R-0011-03.

References

[1] Francis Galton. 1907. Vox populi (The wisdom of crowds). Nature 75, 7 (1907), 450–451.

[2] James Surowiecki. 2005. The wisdom of crowds. Anchor.

[3] Richard P Larrick, Albert E Mannes, Jack B Soll, and JI Krueger. 2011. The social psychology of the wisdom of crowds. Social psychology and decision making (2011), 227–42.

[4] Fischhoff, Baruch, and Cherie Chauvin. "Intelligence Analysis." (2011).

[5] Barbara Mellers, Eric Stone, Pavel Atanasov, Nick Rohrbaugh, S Emlen Metz, Lyle Ungar, Michael M Bishop, Michael Horowitz, Ed Merkle, and Philip Tetlock. 2015. The psychology of intelligence analysis: Drivers of prediction accuracy in world politics. Journal of experimental psychology: Applied 21, 1 (2015), 1.

[6] Pentland, Alex. Social Physics: How Social Networks Can Make Us Smarter. Penguin, 2015.