Productive Failure and Subgoal Scaffolding in Novel Domains.

The assistance dilemma asks how learning environments should "balance information or assistance giving and withholding" (Koedinger and Aleven 2007, p. 239). Minimal guidance (MG) methods posit that students learn best when exploring problems freely, while direct instruction (DI) methods provide canonical solutions early on to streamline students' efforts (problems later). Each method type provides unique benefits, but both are important (Schwartz and Martin 2004) and not easily delivered together. A relatively new MG-based method called "productive failure" (PF) is hypothesized to capture both sets of benefits by requiring students to struggle through problems early on and revealing canonical solutions afterward (Kapur 2008). Students using PF are hypothesized to more effectively transfer and retain information because balancing heuristics and formal knowledge produces diverse solution attempts (diSessa and Sherin 2000) and struggling during exploration pushes students to fill knowledge gaps (Kulhavy and Stock 1989). In the present studies, participants learned to perform tasks in two domains, cryptarithmetic (more traditional) and Rubik's Cube (psychomotor, less traditional) while using either PF or DI. Analyses revealed that (A) PF participants did not outperform DI participants on either immediate post-tests or retention tests, although they did report being more exploration-oriented and trying more unique strategies, (B) subgoal labels increased learning, but only for the relatively novel Rubik's Cube domain (and they sometimes increased workload in cryptarithmetic, surprisingly), and (C) effects of subgoal labels did not change with instruction type. Future research should determine how PF methods can be scaffolded to foster exploration mindsets and diverse solutions.