Dr. Dirk Hovy chats with #SPSM about the important differences between predictive versus explanatory models (from a data science perspective), and we’ll discuss how this applies to suicidology, 2/5/17, at a special time, 2pCT.

Why is this important? Well, it turns out that explanatory models are focused on explaining the causes of a phenomena (identifying risk factors, for example). Predictive models just predict that phenomena. A model that *explains* something may not necessarily be good at *predicting* it. And, when it comes to suicide prevention this is a big deal. These models are often confused, and actually use different kinds of statistical modeling.

Explanatory models on suicide, such as Joiner’s Interpersonal Theory, or other models that aim to identify risk factors have never really been successful at predicting a suicide. However, we really haven’t taken good advantage of data science and possible predictive modeling. If we still struggle with the math for identifying who is likely to die by suicide, perhaps it’s time to try a different approach. Let’s chat about this! For more reading about applications of this approach in psychology and suicide prevention, check out these articles:

yarkoni_westfall_choosing_prediction

mtlhealth

You can watch Dr. Hovy LIVE here:

Dr. Hovey works in Natural Language Processing (NLP), a subfield of artificial intelligence. His research focuses on computational sociolinguistics, i.e., the intersection of sociolinguistics and statistical natural language processing (NLP). The goal of his research is to integrate sociolinguistic knowledge into NLP models. He uses large-scale statistics to detect and model the interaction between people’s demographic profile and their language use (see here or here). He is also interested in semantics (modeling what words mean in context), and non-standard language. He works as an associate professor at the computer science department (DIKU) at the University of Copenhagen.