Identifying the Affective Dimension of Data Mining Practice: An Exploratory Study

• Published in: Transforming Digital Worlds pp. 243 - 252
• Main Authors: Bates, Jo, Elmore, Jess
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Affect; Data mining; Data practices; Data science; Emotion
• ISBN: 3319781049; 9783319781044
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-78105-1_28

The paper aims to illuminate how feeling, emotion and affect influence the practice of data mining. While data mining is sometimes presented as an objective and neutral technique by which to rationally understand and predict phenomena, we observe that there is an important affective dimension in how people understand, engage in and respond to data mining practices. We report the findings of a small exploratory pilot study conducted in 2016 in which we used ethnographic methods to observe the culture of a collaborative project between data scientists and a small digital marketing company. The project aimed to explore potential uses of data mining techniques in the process of telesales lead generation. Thematic analysis of collected data indicates that even in the case of a small scale project, the practice of mining data is deeply influenced by an underlying affective dimension. While these affective dynamics rarely surfaced explicitly in discussions between team members, it is clear from our interview data that feelings and emotions had a significant impact on how participants experienced and engaged with the practice of data mining. Our findings point to the necessity for a much deeper understanding of, and reflexivity in relation to, the affective dimension of data mining practice and how it emerges in the cultures and practices of data science projects. We argue that a deeper awareness of, and openness about, this affective dimension could benefit practitioners’ understanding of their own practice and motivations in decision making, and thus has the potential to improve data science practice.