Productive Failure

Abstract

This study demonstrates an existence proof for productive failure: engaging students in solving complex, ill-structured problems without the provision of support structures can be a productive exercise in failure. In a computer-supported collaborative learning setting, eleventh-grade science students were randomly assigned to one of two conditions to solve problems in Newtonian kinematics. In one condition, students solved ill-structured problems in groups followed by well-structured problems individually. In the other condition, students solved well-structured problems in small groups followed by well-structured problems individually. Finally, all students solved ill-structured problems individually. Groups who solved ill-structured problems expectedly struggled with defining and analyzing the problems, resulting in poor quality of solutions. However, despite failing in their collaborative efforts, these students outperformed their counterparts in the well-structured condition on individual near- and far-transfer measures subsequently, suggesting a latent productivity in what initially seemed to be failure.

Acknowledgments

The research reported in this article was funded in part by the Spencer Research Training Grant and the Education Policy Research Fellowship from Teachers College, Columbia University to the author. I would like to thank the students, teachers, and principals of the participating schools for their support for this project. I am also deeply indebted to Professor Charles Kinzer for his support and guidance throughout the project. Finally, I am grateful to Professors David Hung, Donald J. Cunningham, Florence Sullivan, Liam Rourke, John Voiklis, and the reviewers of this manuscript for their insightful comments and suggestions.

Notes

¹Participants in this study were half-way through their eleventh grade; their tenth-grade CBSE national exams provided the most recent standardized test scores for comparison.

²Even though grade 1 onward, the medium of instruction in the participating schools is English, English is not the native language. Because students had to communicate with each other using text-only chat, English proficiency was included in the analysis to capture any variation that may be explained by differences in English proficiency.

³The sum of proportions, by definition, equals one. Hence, it introduces linear redundancy in multivariate analysis. Therefore, the proportion of NT activity was excluded.

⁴As a rule of thumb, partial η² = .01 is considered a small, .06 medium, and .14 a large effect size (Cohen, 1977).

⁵The software program Multiple Episode Protocol Analysis (MEPA) developed by Gijsbert Erkens was used for carrying out the LSA. See http://edugate.fss.uu.nl/mepa/index.htm.

⁶The derivation becomes clearer when one conceives the functional categories as component states of an evolving Boolean network; Boolean in the sense that, at any point in time, a component state (PA, PC, etc.) may be present or absent in the group discussion, and interactions between the component states may be represented in terms of probabilistic logical functions. This, in many ways, is what LSA attempts to do; it looks at the probability of how certain types of interactions (or component states) follow others at a rate that is significantly above chance level: an if–then probabilistic logical function. As a result, the collaborative process can be examined as an evolving, multi-state, Boolean network, and the greater the number of significant transitions between the component states, the greater the complexity of how its evolution unfolds (Kauffman, 1995).

⁷All excerpts presented here have undergone some minimal editing to make them more readable. In particular, spelling mistakes have been corrected, and unclear short and abbreviated forms have been expanded.

∗Significant effect.

∗Significant effect.