Message-Based Motivation Modeling

Social networks contain a multitude of messages that can be utilized to motivate learning. However, while some messages may increase a learner´s motivation, other messages could undermine it. How can we tell which is which? Conceptual motivation models provide many answers, but how to translate these models into a concrete programmatic implementation (required by e-Learning systems) is often unclear. We approach the problem from a different angle, taking a data-driven approach by (1) assembling a corpus of over 100,000 messages, and (2) applying machine learning methods to this data to create a first-of-its-kind message motivation classifier. The constructed corpus and classifier provide for a new empirical way of studying text-based motivation, developing new models, and empirically evaluating such models on a large-scale.