2016/11/21 by John R. Hollenbeck, Patrick M. Wright · 2 voices · 4 citations
Decision Sciences · Computer Science · #Meta-analysis and systematic reviews #Explainable Artificial Intelligence (XAI) #scientometrics and bibliometrics research
paper · pdf · doi:10.1177/0149206316679487
openalex publication_date 2016/11/21 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/15
In this editorial we discuss the problems associated with HARKing (Hypothesizing After Results Are Known) and draw a distinction between Sharking (Secretly HARKing in the Introduction section) and Tharking (Transparently HARKing in the Discussion section). Although there is never any justification for the process of Sharking, we argue that Tharking can promote the effectiveness and efficiency of both scientific inquiry and cumulative knowledge creation. We argue that the discussion sections of all empirical papers should include a subsection that reports post hoc exploratory data analysis. We explain how authors, reviewers, and editors can best leverage post hoc analyses in the spirit of scientific discovery in a way that does not bias parameter estimates and recognizes the lack of definitiveness associated with any single study or any single replication. We also discuss why the failure to Thark in high-stakes contexts where data is scarce and costly may also be unethical.