According to the ‘selective engagement hypothesis’, as individuals experience early symptoms of cognitive decline, they may divert newly limited cognitive resources towards essential everyday functioning, and away from other nonessential tasks. As such, prior research has suggested that social interaction may constitute one such nonessential task that cannot be maintained in the face of emergent cognitive decline. With the advent of virtual social networks platforms like Facebook, it is now possible to derive rich and accurate measures of an individual’s level of interaction in these platforms, collected over several years. This presents an unprecedented opportunity to investigate the relationship between objective daily activity patterns and similarly objective markers of cognitive decline.
Aside from gains in data quality and quantity, this methodology presents a unique opportunity to collect a time series of historical data, from which we can generate models that can make predictions about future cognitive status. The goal of which is to develop tools that can be used to identify at-risk patients and provide a pathway to early intervention. The aim of the proposed pilot is to study changes in language use and social interaction patterns using Facebook history in a cohort of early diagnosed dementia patients, with the goal of developing a predictive model aided by machine learning.
If successful, this pilot will lay the foundation for a larger study that will (a) refine this algorithm in a larger sample and critically, (b) test its performance in a hold-out sample that were not used for training.