AAPT Conference - How Effective is Productive Failure Learning in Physics Education

Introduction

Studying computationally driven disciplines, such as physics and math, requires well-rounded learning techniques that ensure proper conceptual understanding and transfer – beyond acquiring good grades only (Hardiman et al., 1989). There is a wide gap between instructors’ perceptions of students’ understanding and students’ actual learning (McDermott, 1991). Several conventional instructional methods utilized in classrooms might produce decent grades, giving out an impression of successful learning. However, learning and performance are not always commensurable, especially for problem-solving disciplines (Kapur, 2014). This means that despite producing results, most traditional techniques are not engaging and inspiring enough to develop conceptual understanding and inculcate interest in students (Hoellwarth et al., 2005). For example, repetitive problem-solving was once considered an efficient strategy to improve learning for such disciplines. However, recent studies have revealed that although these techniques improve scores, they usually fail to build a strong connection between conceptual learning and problem-solving skills (Freeman et al., 2014; Stamovlasis, et al., 2005). Instead, such techniques ‘promote the memorization of context-specific solutions with minimum generalization’ (Bao & Koenig, 2019, p.4). This means even if students could solve problems similar to what they have practiced, they would still struggle or fail to transfer those learnings to other problems or interpret their results (Sweller & Cooper, 1985). In fact, instructional practices focusing on performance are one of the crucial reasons many students lack an in-depth understanding of the content. This situation is more prevalent in higher education where most classes are conducted in a linear manner: weeks of lectures followed by exams (CITATION). This issue calls for smarter learning strategies.

In recent years, the development in education research has resulted in the emergence of numerous novel and more efficacious learning techniques, several of which are extensions to discovery-based learning (Klahr & Nigam, 2004). These include comparing contrasting cases (Schwartz & Bransford, 1998), inventing to prepare for learning (IPL) (Schwartz & Martin, 2004), and productive failure (PF) learning (Kapur, 2014). Out of all these methods, productive failure has produced remarkable results for computationally and conceptually intense disciplines such as math and physics in K-12 education (Kapur, 2012; Pathak et. al, 2011; Kapur & Bielaczyc, 2012; Kapur, 2008). Productive failure learning dwells upon the notion of creating planned, short-term, hindrances for learners when they are learning new concepts, causing them to fail at the start. Research shows that short-term failure in performance at the beginning can result in long-term learning, especially if failure is desirable and tactfully located (Kapur, 2016).

Problem Statement

So far productive failure learning has mostly been implemented in K-12 settings and its efficacy is yet to be explored in higher education. Moreover, there are only a few studies that have used the contrasting-case design to implement productive failure learning (Kapur, 2008). Hence, this study aims to bridge the gap. This mixed-method study aims to measure the effectiveness of productive failure learning in the context of undergraduate physics education to evaluate its efficacy for solving ill-structured and well-structured questions.

The proposed hypothesis for the study is that the group that receives treatment in the form of ill-structured problems followed by well-structured problems (a contrasting structural design) would outperform groups receiving treatments in a non-contrasting structural design.

Guiding Research Questions

The research question guiding this study is: a) can solving ill-structured problems without any facilitating structures be used as a productive failure exercise; and b) how efficacious is the contrasting-cases design—implemented via ill-structured problems followed by well-structured problems—to extract the usefulness of productive failure learning when compared to a noncontrasting-cases design i.e., 1) well-structured problems followed by more well-structured problems and 2) ill-structured problems followed by more ill-structured problems?

Research Plan

This is a mixed-method experiment with three groups, each receiving a different treatment. As it is an extension of Kapur’s (2008) existing work on productive failure learning in physics education, it builds upon his existing design of two conditions. There are three treatment groups in the study, as shown in Figure 1: in Group 1, participants will first solve ill-structured problems followed by well-structured problems; in Group 2, participants will solve well-structured problems followed by more well-structured problems, and in Group 3, participants will solve ill-structured problems followed by more ill-structured problems. After completing the treatment phase, all the participants from the three groups will take a post-test that will comprise both well-structured and ill-structured problems.

The independent variable of the study is the difference in the structuredness of the three treatment conditions (ill-structured followed by well-structured vs. well-structured followed by well-structure vs. ill-structure followed by ill-structure, as shown in Figure 3) whereas the performance of participants in the post-test is the dependent variable of the study. The post-test will comprise both ill-structured and well-structured questions, but the participants will be evaluated for their score on well-structured problems and the quality of their ill-structured problems. This is to ensure the internal validity of the study i.e., the experiment should not be designed to cause other groups to fail. Furthermore, since it is a mixed-method study, along with the final scores of the participants, the quality of their solutions and discussions will also be analyzed qualitatively to find how if they were using novice-like or expert-like skills during the process.

The participants in the study will be future elementary teachers enrolled in an undergraduate physics course. There will be around 100 participants, 95% of them females, who will be randomly assigned to one of the three treatment conditions. In the first phase of the study, participants will be working in groups of four and will deliberate with each other. However, in the second phase of the study, they will be working individually on their respective tasks. Since the study is part of the course, it is assumed that the participants will be putting in their best efforts, and their submissions will truly represent their understanding and learning of the material.

While Kapur chose Newtonian mechanics as the topic of study for the participants, this study aimed to opt for circuits as the topic of interest. As productive failure learning has never been used as an instructional strategy to teach circuits, the study would also uncover its effectiveness in topics other than kinematics.