Predicting information diffusion in social networks is a theoretical and practical concern, especially in the context of identifying and thwarting barely licit attempts to influence individuals’ political and purchasing decisions. As the authors demonstrate, the predominant diffusion models employ fixed factors, usually one of two. These models can only adapt to new situations by changing the inputs to the models. Chou and Chen propose another approach. They improve the predictions generated by fixed factor models with a multiple and variable factor model. The variability is that the number of factors is node and environment dependent. The model also includes the probable uncertainty of information flow between two nodes in a network. Their model is thus multidimensional. They call it a multiple factors-aware diffusion (MFAD) model.

The MFAD model includes such influencers on diffusion as the preexisting popularity of information and the length of time for an individual to embrace a piece of information. MFAD can select or ignore factors.

The authors buttress an already compelling theoretical case with empirical evidence. Using both synthetic and actual data, the MFAD model outperforms competing models, especially in its ability to predict information diffusion, a measure of how rapidly information propagates through a network.

The authors point to future research directions. These include online learning and maximization of information flow. They suggest that their method could be applied to natural phenomena, such as epidemics, that spread through a network.