An Opinion Diversity Enhanced Social Connection Recommendation Re-Ranking Method Based on Opinion Distance in Cyber Argumentation with Social Networking

摘要

The quality of crowd wisdom extracted from online communities decreases as the community becomes less ideologically diverse, which is an issue in many online spaces. One cause of this decline is that users tend not to engage with diverse, idea-challenging content that contrasts their prior opinions. However, they do tend to engage with content endorsed by their social connections, even if it goes against their personal opinion. Thus, by increasing the diversity of opinion in a user's social network, they will likely engage with more diverse content. We are developing a cyber argumentation system with social networking and present a social connection recommendation re-ranking method that promotes opinion diversity. We use artificial intelligence and data mining techniques to mine and analyze user opinions from argumentation data on important issues, then use furthest opinion distance to re-rank the recommendations. Our method is designed to easily integrate with existing social connection recommenders, which preserves platform specific criteria. We compare the opinion diversity of recommendations from five types of social connection recommendation methods, with and without our re-ranking method, on a large empirical dataset. Our results show that our method improves the recommended diversity by around 15% for five existing social connection recommendation methods, while only reordering around 50% of the initial social connection recommendations.