Each social relationship is complex and unique. Yet, most social media research has characterized social interactions with simple models of sentiment or strength, and interpreted emergent collective phenomena in the light of such models. Even though it is hard to get an algorithmic understanding of the complex nature of social interactions, a solution might lie at the intersection of AI and psychology.
We found that 10 distinct types of interaction extensively discussed in psychology are sufficient to account for the way most people conceptualize their relationships. These types are: Trust, Respect, Knowledge, Power, Support, Identity, Similarity, Fun, Romance, and Conflict. They built a Deep Learning tool to extract the expression of the 10 dimensions from any conversational text. By applying their tool to a variety of source material (including Reddit posts, Tweets, and movie scripts), they found that the combination of the predicted dimensions suggests both the types of relationships people entertain and the types of real-world communities they shape. For example, some of the dimensions are predictive of societal outcomes in US States, such as education, wealth, and suicide rate.
In the future, social media researchers could use this tool to enrich their interpretation of collective dynamics of social networks under the light of these 10 explainable dimensions of interaction. For example, they could shed more light on why fake news spread and on the dynamics of conflict in social media. Social media providers could go beyond limiting antisocial behavior on their platforms and use this tool to identify and promote opportunities for positive and meaningful social interactions.
N.B.: If you do not receive the instruction message within a few hours, please check your junk/spam e-mail folder just in case the email was moved there.