Dataset for "Understanding Interaction with Machine Learning through a Thematic Analysis Coding Assistant: A User Study"

20 participants installed and interacted with a thematic analysis coding assistant (TACA), an interactive machine learning desktop application designed to train a classifier on user-defined coded datasets to generate additional coding suggestions. The interviews were conducted with the participants after they interacted with the tool for 20 minutes, or until no more benefits were perceived. The questions were aimed to understand the experience of the participants with TACA and their perceptions of the ML model.

• The coded_transcripts.docx file contains the anonymised interview transcripts coded with codes appearing as comments. The document is split into Study 1 (5 participants) and Study 2 (15 participants). The participants in Study 1 imported their own dataset into TACA, while the participants in Study 2 used a set of newspaper restaurant reviews that were given to them by the researchers. Participant IDs follow the structure "S[study number]_P[participant number]", e.g. "S2_P1".
  
• The themes.csv file shows all the codes below each corresponding theme, the result of conducting thematic analysis on the interview transcripts.
  
• The restaurant_reviews.docx file is the collection of 21 restaurant reviews from the newspaper The Guardian (Restaurants + Reviews | Food | The Guardian) that was given to 15 of the 20 participants who did not have their own dataset available for the study.
  
• The logs folder contains an anonymised interaction log file for each participant with the interface of TACA named with the corresponding participant ID. The interaction logs for participants S1_P4 and S2_P5 are missing due to an issue in data storage.