Understanding and predicting the dissemination of scientific papers on social media: a two-step simultaneous equation modeling-artificial neural network approach

Social media platforms have had an enormous impact on the dissemination of scientific work and have fared well in covering scientific papers. However, little is known about the general dissemination process from academia to social media and how various factors affect the dissemination of scientific papers at different stages. In this paper, we proposed a two-staged dissemination process to profile the diffusion of scientific papers from academia to social media. A two-step simultaneous equation modeling-artificial neural network approach was adopted to predict the retweet scale of scientific papers on Twitter by combining source-related and content-related factors. The analysis in the field of oncology suggests that the artificial neural network algorithm (ANN) with the input units generated from the simultaneous equation model (3SLS) can predict the retweet scale of scientific papers on Twitter with an accuracy of 78.05%. According to the normalized importance obtained from the ANN, we found that most factors related to the information source play critical roles in promoting the dissemination of scientific papers. The number of first-generation tweets has the most remarkable impact on subsequent dissemination. As for the content-related predictors, tweets attached with more URLs can provide richer information for audiences, thereby increasing the retweet scale of scientific papers. Besides, the influence of research topics on dissemination varies with different audiences. The findings of this study contribute to the literature on the dissemination of scientific papers beyond academia and provide practical implications for scholarly communication.