Primary Submission Category: Causal Inference Education
Does association really imply non-causation? The power of language in causal attribution.
Authors: Jennifer Hill, George Perrett, Stacey Hancock, Le Win, Yoav Bergner,
Presenting Author: Jennifer Hill*
While most statisticians and, arguably, all causal researchers have been taught to be cautious in making unwarranted causal attribution, many of us are not as careful as we might be in describing results of descriptive (i.e. non-causal) evidence. For instance, it is still common practice to use the word association to describe non-causal relationships, even if this wording is combined with causal words such as “change,” “increase/decrease,” or “gain/loss.” Are these wording choices innocuous or do they inadvertently lead less sophisticated readers to assume that these links are causal? The current study investigates the connection between the wording of study findings and causal attribution by the reader using a series of randomized experiments involving several samples of students from two large U.S. universities. It also provides evidence about the association between statistics instruction and the ability to understand appropriate causal attribution. The results suggest that specific wording choices used to describe study results noticeably impact the level of causal attribution by the reader. Moreover, the degree of causal attribution is strongly linked to the research topic. These results suggest ways to tailor wording of research findings to help decrease the probability of causal misattribution.