Perhaps it’s understandable that work committed to helping with human misery and self-destructiveness should want to have certainties and confident causal attributions. It’s probably good to have active discussions of what causality is and how to be confident of having attributed it correctly. However, my/our position in the book is that this wish has been allowed to dominate our field and to create the idea of a “hierarchy of evidence” in a way that is unhelpful.
There are many areas of life where causal attribution is vital: understanding heavier than air flight and how how parachutes work need strong causal attributions to be safe. Philosophers and methodologists note that there are different sorts of causal attribution, as often the wikipedia entry on causality is a nice place to start.
In our field randomised experiments have been given a dominant position but there is work using path analysis, or its derivative, structural equation modelling (SEM) and some work using “directed acyclical graph” (DAG) methods and the idea of “Granger causality” in time series can be useful. I’m particularly fond of the paper Avdi, E., & Evans, C. (2020). Exploring Conversational and Physiological Aspects of Psychotherapy Talk. Frontiers in Psychology, 11, 591124. https://doi.org/10.3389/fpsyg.2020.591124 (open access) I co-wrote with Prof. Avdi that combines qualitative and quantitative methods and time series data. However, we were very careful there to make no causal attributions as even though we had time series data and interesting phenomena we were certainly way short of methods even to attribute Granger causality to things.
Though SEM, DAG and time series methods are important and useful, they almost always either require research situations very far removed from routine practice (not intrinsically wrong of course but limiting generalisability) or else turn on assumptions, particularly that anything that might have affected things has been measured, that are implausible.
Try also #
Mostly pertinent to chapter 10.