Computational Models for Characterization of Idea Evolution in Collaborative Learning and Research Communities
March 8, 2018
Light Engineering room 250
Advisor: Prof. Alex Doboli
Understanding the flow of ideas in collective learning and group activities has been an important yet difficult topic. It offers novel and important insights on devising management strategies that optimize the quality and productivity of collective learning outputs by promoting a certain composition, structure, and interaction patterns for a community. This dissertation aims at designing an integrated platform consists of data acquisition functionality as well as novel methodologies to characterize and model the evolution of ideas in communities.
This thesis first presents a novel model and related methodology to characterize the evolution of circuit design communities, expressed as citation graphs, over time. The model explains knowledge evolution using two operators, combination/improvement for knowledge expansion and blocking for concept elimination. A set of new metrics were proposed to describe the effect of the operators. Experiments considered three large data sets of circuit design publications.
Secondly, an experimental study on the roles that research groups have in the idea flow within a community is presented. The study analyzes two facets: the contribution of different types of groups on the impact research ideas make on other groups, and the importance groups have in linking sets of groups with each other. Results showed that research groups can be clustered depending on the number of received citations, group size, and other related features. Groups of the mid-tier categories are important in tying a community together as they improve idea impact and bridging between groups. Policies that aggressively eliminate such groups reduce the effectiveness of idea flow, even though research communities include a certain amount of robustness for moderate levels of group removals. Similar results achieved for a forum-posts study are also presented.
In order to further expand the analysis of idea flow and role modeling to human interactions during group activities, we devised an integrated Cyber-Social System solution for group interaction tracking. The system integrates speaker identification and speech emotion recognition algorithms that utilized deep learning approaches. A method to design adaptive control strategies that produce reliable data collection and modeling for interaction-tracking applications is discussed.