Triadic influence as a proxy for compatibility in social relationships.

Area of Science:

• Social network analysis
• Computational social science
• Network science

Background:

• Social interactions form the basis of civilizations.
• Understanding relationship dynamics (friendships, enmities) is crucial but scientifically challenging.
• Large-scale social network data with individual attributes are difficult to obtain.

Purpose of the Study:

• To quantitatively study the formation and evolution of social networks.
• To investigate the predictive power of individual attributes and network structure on relationship signs.
• To introduce and evaluate a novel metric for social influence within networks.

Main Methods:

• Collected data from 13 schools, encompassing over 3,000 students and 60,000 declared relationships.
• Introduced 'triadic influence' metric to quantify the impact of neighbors on relationships.
• Employed neural networks and high-dimensional embeddings to predict relationship signs (friend/enemy).

Main Results:

• Triadic influence alone achieved the highest accuracy in predicting relationship signs.
• Incorporating personal attributes did not significantly improve prediction accuracy.
• Triadic influence appears to serve as a proxy for social compatibility.

Conclusions:

• Triadic influence is a powerful predictor of social network evolution.
• Network structure, particularly triadic influence, may be more critical than individual traits in relationship formation.
• This work opens new avenues for the quantitative study of social systems.