Development and implementation of a user interface prototype for an interactive machine learning system that supports Grounded theory practitioners with building theories
Qualitative coding in the era of big data is a challenging task for grounded theory practitioners. Interactive machine learning systems can provide researchers the perks of employing machine learning methods on the data, and thus assist them with some processes, but nevertheless, grant them control over the model. In this work, we draw inspiration and context from previous works in the eld in order to implement such a system that makes use of semantic interaction on a spatial layout to infer the intent of the user to train the model. We explore how users react on the system, the data and themselves in connection to the data while using it.
We seek to provide further insights in the eld of designing technologies that take the reactions of users into consideration. To explore that, we perform a think-aloud user study using a pair-analytics approach. The participants receive an introduction to the system and then are asked to interact with the system. The collected data from the user study is then analyzed in terms of whether and how the participants reacted during the interaction.
The results of the user study are examined and further discussed. As the results displayed great variability, the reactions of the participants provided plentiful insights. Some of which are regarding to their prior knowledge, the difference between human and machine in measuring similarity, the data, the system's purpose and shared control in an interactive machine learning context.