Finding Patterns Within Patterning: Data Mining of Relational Reasoning Utterances in Engineering Design

• In Event: 51.024 - Educational Data Mining: Affordances and Constraints in the Analysis of Behavior and Performance in STEM

Abstract

Purpose & Theoretical Framework
In the workplace, complex problems are frequently the responsibility of teams who collectively labor to achieve a positive outcome. One area in which teams are core to the discipline is engineering design. While functioning as a unit, it often expected that each member brings foundational abilities and perhaps unique talents to the task. One foundational ability is relational reasoning, the ability to discern meaningful patterns within a stream of information. Alexander and colleagues (2012) identified four forms of relational reasoning that warrant attention. The first, analogy, refers to relational similarity; anomaly, to outliers within a set; antinomy, to incompatibilities or mutual exclusivity; and antithesis, to opposition. In particular, the use of analogies and antinomies has been linked to creativity and innovation (Christensen & Ball, 2016). This study uses social network analysis and sequence mining to examine the patterns of relational reasoning among team members conceptualizing a new product design.

Method
The sample included one six-person team of senior mechanical engineering students enrolled in an engineering design course. During three meetings, students were tasked with conceptualizing ten viable ideas for a new product (Meeting 1), narrowing ideas down to three (Meeting 2), and then to one (Meeting 3), which they would ultimately prototype. Specifically, their project was to reduce the difficulty of re-entering a kayak after capsizing.
The team meetings were audio- and video-taped. Researchers performed a content analysis examining task-related and relational reasoning verbalizations. Table 4 contains examples of each type of coded verbalization. Sociomatrices and bipartite graphs were constructed to illustrate co-occurrences of utterances and contributions made by each team member. Finally, a sequence mining tool was used to identify patterns of relational reasoning as they occurred to uncover more about how such reasoning unfolds in real time.

Results
Social network analysis demonstrated that task-related utterances had the highest degree centrality across all meetings, while anomalies had the lowest (see Figure 1). It also highlighted the relative importance of analogies in Meeting 1, task management in Meeting 2, and antinomies and antitheses in Meetings 2 and 3. Bipartite graphs showed that throughout the three meetings, the proportions of each type of verbalization in each person’s talk stayed relatively consistent (Figure 2). However, different individuals spoke in differing amounts across the meetings. Further, analogies were verbalized in higher proportions by everyone in Meeting 1, whereas antinomies and antitheses were more prominent in Meetings 2 and 3. Finally, a sequence mining tool was constructed to find strings of relational reasoning utterances two to six units in length in each meeting. Results showed that task related and task management units, analogies, and antinomies tended to each occur in strings of 3, and analogies often preceded antinomies.
Significance
Using data mining techniques, patterns of relational reasoning in engineering design were identified as they occurred in a real time design task. Specifically, the distribution of relational reasoning was reflective of the goals of each meeting. Challenges and affordances of applying data mining to patterns of relational reasoning will be discussed.

Authors

• Sophie Jablansky, University of Maryland - College Park

• Patricia A. Alexander, University of Maryland - College Park

• Linda Schmidt