Luca Spliethoff & Stephan Abele


A common problem type workers are confronted with across different professional and vocational domains are diagnosis problems that require identifying the causes of undesired states, such as machine malfunctions or diseases (Jonassen, 2010, p. 78). In the technical domain, a field with high relevance of diagnostic problem solving is car mechatronics, where diagnosing the causes of car malfunctions is an important work task (Baethge & Arends, 2009). To regulate the diagnostic problem-solving process and find the cause of the undesired state, diagnosticians can apply different diagnostic problem-solving strategies. To date, empirical research on how to differentiate between different strategy types based on observable problem-solving behaviour is scarce and there is no agreement upon how to operationalize the strategies (Konradt, 1995; Schaper et al., 2004). Log data generated by a computer-based assessment system during a diagnosis task build an expedient basis for this research purpose, as they provide an extensive picture of the task completion process. In this talk, we demonstrate how a theory about the diagnostic problem-solving process (Abele, 2018) can serve to deduce which types of evidence are relevant for identifying different strategy types. In the presented framework (Abele & von Davier, 2019), idiosyncratic patterns of the problem-solving behaviour are theoretically defined for each diagnostic strategy. Subsequently, the patterns are used as search template to scan the log data and to identify which test takers showed the respective patterns (i.e., strategies). Specific challenges that are associated with the analysis and interpretation of computer-generated log data, as well as potential approaches to overcome these challenges, will be discussed.