Teachable Agents • Hobbies
Our research on Teachable Agents has found them to be very engaging to students. One of the goals of our group is to understand this form of motivation. Most theories of motivation are not tightly linked to learning. They describe external incentives like grades or internal incentives like fantasy that, if harnessed correctly, can indirectly lead to better learning. However, in some instances, people appear to be directly motivated to learn. Hobbyists embody the kind of engaged learning that we would like to see in classrooms. Some people who would otherwise be completely uninterested in microbiology, for example, can become quite engrossed in the subject when understanding some basics of microbiology is one of the keys to making better beer. The goal of this project is to look at ways to move this kind of engagement and directed learning into the classroom.
We started by surveying hobbyists who subscribed to hobby-related e-mailing lists on the Internet who were asked to rate a set of things which might contribute to the enjoyment of their hobbies. A list of potential sources of satisfactions was taken from different areas of the literature on motivation and engagement as well as some things that we thought were missing. These areas, loosely defined, are (1) intrinsic motivation and flow, (2) extrinsic and achievement motivation, (3) social motivations, (4) motivations to learn, and finally (5) the motivation to create artifacts. The survey had a 1-7 ranking for each item as well as a field for respondents to include written comments.
The results from the adult hobbyists showed considerable consistency in their most- and least-valued satisfactions. The top four motivations across a variety of hobbies were (1) the creation of a product, performance or artifact; (2) learning the methods necessary to create the artifact; (3) sharing the artifact with others; and (4) appropriate levels of challenge.
There was some variance between different hobbies-motorcycle racers, for example, rated highly the satisfaction they got from competition and being better than others, items that were rated much lower by members in other hobbies. Nevertheless, they also rated designing their motorcycles and sharing as extremely significant. The degree to which the same items were valued by members of all hobbies suggests that there are a few things that contribute to enjoyment in different types of hobbies. This suggests that what makes hobbies enjoyable is not entirely determined by the specific hobby.
To see if this list of satisfactions was also applicable to students in, or out of class, we gave a similar survey to high school seniors twice-once for their favorite non-academic activity and once for their favorite class. Again, there were similar patterns.
Data from these surveys suggest that creating an artifact or performance, combined with the prospect of sharing it with others, are tightly linked with the motivation to learn the methods for creating the artifact. However, this finding is based on activities into which people self-select. An important question is whether it is possible to create these motivation-to-learn conditions in settings that are not optional.
Our next project was to design instruction that maximizes and minimizes these factors to test their effect on student engagement in the classroom. To do this we asked students to work with NetLogo in each of three conditions. In the observation condition, the 7th-grade students observed and took notes to answer questions about several runs of a simulation on population dynamics. In the simulation condition, the students changed parameter values to affect the simulation and also took notes to answer questions. In the programming condition, the students added new features to the simulations and again took notes and answered questions. In a post-survey, students overwhelming preferred the programming condition. Additionally, we took real-time measures of student engagement (a pop-up window asked them to rate their engagement) and of students' use of learning resources. A striking finding is that high levels of engagement predicted the use of learning resources, but only in the programming condition. This work is currently described in a dissertation thesis by Jay Pfaffman. Contact him for further information.
