Adaptive Support and Task Generation in Exploratory Learning Environments

• In Event: Roundtable Session 98
  In Roundtable Session: Using Learning Analytics and Educational Data Mining to Understand Scripted and Exploratory Learning Environments: Toward a Common Theoretical and Methodological Framework to Investigate the Trajectory to Expertise

Abstract

As the benefits of inquiry learning become apparent, so does the need for better support for students’ inquiry behaviors (Tobias & Duffy, 2009). However, adhering to constructivist principles of learning, inquiry-learning activities offer students only limited guidance at the domain level. While the relative lack of domain-level support may be productive for students who manage to construct valid models, many students fail to achieve the desired learning goals (Kirschner, Sweller, & Clark, 2006; Klahr & Nigam, 2004). The tension between exploration and support is most apparent in Exploratory Learning Environments (ELE). ELE facilitate activities in which students are asked to develop an explanation or reveal an underlying model that reflects the structure of the given data (de Jong & van Joolingen, 1998). For example, Figure 1 shows a typical task in the Invention Lab (Roll, Aleven, & Koedinger, 2010). In this task students are asked to invent methods for calculating the spread of given data, prior to receiving instruction on variability.

Figure 1: A Typical Task in the Invention Lab asks students to invent a general method for calculating the variability of the given data.

ELE give students high agency over the learning process. The design space that students explore is virtually unlimited, and the solution space that needs to be evaluated by the system is under-defined. Thus, ELE typically offer only little support at the domain level (Mulder, Lazonder, & de Jong, 2009).
In this talk I demonstrate the potential of using students’ moment-by-moment traces to offer domain-level support in ELE. The Invention Lab analyzes the invention process and its outcomes, and generates, in real time, new data for students to explore. This form of automated Socratic tutoring creates tasks that help students realize the flaws of their invented methods and advance their understanding. The Invention Lab adapts its problems to individual students by combining modeling approaches: First, a cognitive model is used to evaluate students’ inquiry behaviors and give process-level feedback. Students’ invented methods are then evaluated using constraint-based modeling (Mitrovic, Koedinger, & Martin, 2003). The outcome of this analysis is a list of features that are missing from the students’ methods, and likely reflect gaps in their understanding of the domain. The third model is a model of an expert teacher. It uses the information regarding students’ invented methods to prioritize their knowledge gaps and create new tasks that target these gaps.
Log data from a classroom evaluation of the lab (N = 33) was used to validate the modeling approach of the lab. Detailed log files were also used to learn about students’ typical invention behaviors and outcomes. For example, while students were found to evaluate their methods and address the deep features of the domain, they often failed to create general methods that apply across examples and capture multiple features. This work demonstrates how moment-by-moment data from students’ learning trajectories can (i) improve our understanding of students’ inquiry behaviors, (ii) explain students’ learning form invention activities, and (iii) enable domain level support in ELE.

Author

• Ido Roll, The University of British Columbia