Tuesday Oct 18 1300
- Bart Russell
- Deputy director
https://zed.uchicago.edu/data/pub_drafts_//YFA/cognet.pdf
Abstract
Society is a complex system shaped by the beliefs of its members. Diverse opinions can facilitate social learning and adaptation, but polarization can perpetuate separation and conflict. Despite the mechanics of how social and cultural factors contribute to societal polarization remains unclear. Additionally, the historical use of diverse polarization measures, and the lack of an objective standard to homogenize their analysis, has hindered our understanding. Here, we introduce a computational framework (CogNet) to discover dependencies between individual opinions on diverse topics, to ultimately craft a theory of belief shift, that is predictive at the level of individuals. Automatically analyzing responses from > 65K US participants over approximately half a century (1972-2021) from the General Social Survey, and > 3.9M participants from the European Union over a similar time period from the Eurobarometer surveys, we distill emergent dependency structures in collective opinion, and demonstrate that our models can reason with, and reliably predict responses to thousands of controversial social questions, for specific individuals with a limited amount of information on their worldviews available a priori. CogNet’s opinion prediction capability, demonstrated by its ability to improve occluded worldview estimates for greater than 90% and 99% of GSS and Eurobarometer participants respectively, arises from a data-inferred metric (Gibbs-distance) to measure deviations in the space of opinions. The Gibbs-distance quantifies the likelihood of spontaneous shift between worldviews, yielding objective computable measures of societal polarization, which includes the variation of “distance” between two fixed extreme worldviews over time, offers quantitative evidence that faltering economies anticipate greater polarization. Additionally, we discover that affective polarization is a likely precursor to ideological polarization, yielding a valuable insight into how social structures respond to economic stress: opinion clusters increase in numerosity and move apart, leading to widening gap between dominant ideologies. Such computable frameworks, validated to capture the intricate dependencies between opinions and the ensuing evolutionary dynanics, may help foster more effective socio-economic policy, and illuminate new frontiers in social theory.
Points that we need
| simulation of social and economic systems |
|---|
| digital twins |
| knowledge settings |
|---|
| opinion systems |
| high value conjectures |
Dr. William Corvey Information Innovation Office (I2O) Program Manager
Dr. Mary R. Schurgot Strategic Technology Office (STO) Program Manager