Abstract:

Student motivation is foundational to effective learning across both classroom and self-directed contexts. Maintaining, let alone improving, student motivation is a challenge that remains from the earliest modern educational research. In the present study, we establish predictors of student motivation by examining data collected by Shonan Seminar (SHO-zemi) as part of their efforts to characterize their instructional programs. Compared to public educational institutions, SHO-zemi experiences strong pressures to optimize student motivation and provides extensive training to teachers for implementing motivation-boosting strategies. As in any educational context, however, there are marked difficulties in measuring transient student energy levels and emotions. Historically, such assessments have been performed by expert observers and are time-intensive. More recently, however, experience sampling methods have provided a technologically-mediated, less resource-intensive approach for individuals to report their internal state. While such methods are not without shortcomings, they have enabled experiments and observations that were previously infeasible.

Over the past two years, SHO-zemi has implemented tablet-based surveys of student emotion, arousal, and attitudes in both 4-day residential and semester-long weekly tutoring contexts. This approach is highly streamlined, works without an internet connection, and requires only about five minutes of classroom time to administer. There is also potential for massive scale in both of these contexts, given approximately 1000 students in the summer residential program and 200 relatively homogenous classrooms offering weekly tutoring. In the present report, however, we focus on an initial trial including 39 students in a 4-day residential program. This context offered a structured opportunity to collect surveys three or four times a day at standardized times as students studied for an upcoming high school entrance exam.

The primary aim of this report is to characterize the emotional context experienced in this residential tutoring intervention where student motivation was not only maintained, but generally increased. Broadly speaking, students reported marked emotional shifts following administration of mock exams. As an initial step, a Principal Component Analysis (PCA) was used for each reported emotion to characterize the time structure of emotional responses. Scree plots reveal that the first component generally explains student differences much more than other components. Moreover, the largest components correlated better between emotions as compared to raw emotion scores, presumably because the decomposition highlighted underlying events that were driving these emotions. Subsequent to this decomposition, stepwise linear regression revealed a robust effect across model sizes in which happiness, sadness and feeling “able to achieve good results” significantly predicted student motivation. Additional analyses are also provided, including demonstration of the limitations of this approach in situations where students are engaged in a more heterogeneous daily experience. These methods and results lay the foundation for future experiments in which a massive parallel class structure can be leveraged to test causal relationships via interventions that may emotionally support improved motivation, and thus improved learning outcomes.

Keywords:

Motivation, emotion, surveys, data collection, experience sampling, PCA, tablets, tutoring.