Successful instruction should help students acquire robust knowledge and prepare them for future learning opportunities. However, we are yet to find a winning strategy for systematically achieving robust learning (Bransford & Schwartz, 2001). Accumulated evidence suggests that discovery learning does not help most students acquire the basic foundations, and direct instruction, on the other hand, often leads to a relatively rigid body of knowledge (c.f., Tobias & Duffy, 2009). Instructional technologies are in a similar pursuit of robust learning (Koedinger & Aleven, 2007). However, students working with discovery environments often do not receive adequate support and thus fail to achieve desired learning gains (De Jong & van Joolingen, 1998). Students working with intelligent tutoring systems receive appropriate support, but on tasks that may not prepare them enough to make sense of new situations.

Recently, Schwartz and colleagues devised a hybrid method called Invention as Preparation for Learning (IPL; Schwartz & Martin, 2004). In IPL students attempt to develop novel mathematical methods prior to (and not instead of) receiving direct instruction. While Schwartz and Martin (2004) showed that IPL is successful in preparing students for future learning, questions regarding the mechanisms and scalability of IPL remain largely unanswered.

This thesis focuses on understanding the sources of IPL's effectiveness, and using that to design technology that can scale up IPL. To address these issues, I conducted a series of classroom experiments to assess the effect of IPL on students' domain knowledge, motivation, and general invention skills, and to identify under what conditions and by what cognitive mechanisms IPL accelerates future learning; I contrasted different versions of IPL in order to identify its core components; and I created and evaluated the Invention Lab, a unique intelligent tutoring system for IPL.

This thesis makes contributions to cognitive science by better understanding the mechanisms and effects of inventions in learning. It contributes to the learning sciences by conducting comprehensive evaluations of a novel pedagogy. And it contributes to the field of human-computer interaction by designing, implementing, and evaluating a novel type of intelligent system, capable of adapting to users' knowledge in scientific inquiry tasks.

160 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu