Knowledge Co-Construction During Collaborative Learning

Heisawn Jeong

Learning Research and Development Center

University of Pittsburgh


This thesis work aims to study the process and outcome of collaborative learning. College dyads collaborated to learn a biology concept in this study. Detailed analyses of their answers during pre- and post-test and the interaction during collaboration are planned to address (1) the kind and extent of shared knowledge constructed after collaboration, (2) the relationship between the construction of shared knowledge and individual learning, and (3) the underlying processes of knowledge co-construction.

Keywords–collaboration, learning, knowledge co-construction


A basic premise of social interaction is the construction of shared knowledge (Hardin & Higgins, 1997). Collaborative learning is often characterized as a process of constructing shared knowledge in which people converge on a shared meaning and representation of the materials (Roschelle, 1992).

Traditionally, the researchers who study collaborative learning have focused on what happens to individuals as a result of collaboration (e.g., individuals learn better when collaborate). However, given the finding that dyads or groups process information together, it is increasingly important to examine the processes and outcomes of collaboration both at the individual level and at the dyad or group level.

The aim of this thesis is, thus, to study the outcome of collaboration and its underlying mechanism in more detail. In this study, dyads collaborated to learn the human circulatory system. Through detailed analysis of their answers and their interaction, this thesis plans to address the following three questions: (1) Does collaboration result in the construction of a set of shared knowledge? If so, then what kind and how much shared knowledge results from collaboration? (2) Is there a relationship between the amount of shared knowledge and the amount of learning after collaboration? (3) What is the process of co-construction like?


Twenty-two dyads, made up of University of Pittsburgh undergraduates, participated in this study for course credits. Participants had not taken any college-level biology or nursing classes. All the dyads were of the same gender (9 male and 13 female dyads) and race (3 African American and 18 Caucasian dyads).

During the first session, participants were individually tested on the following two pre-tests: (1) Terms task in which participants were asked to tell the experimenter everything they knew about 19 terms relevant to the human circulatory system and (2) Blood Path (BP) task in which participants were asked to draw the path of the blood in the circulatory system.

In the second session, participants were paired with another student of the same gender and race and were asked to collaborate to learn a text passage on the human circulatory system. The text, originally from a high school biology textbook, consists of 73 sentences and was adopted from Chi, de Leeuw, Chiu, & LaVancher (1994).

The third session was scheduled roughly a week after the collaborative session. Participants were tested individually on the Terms task and BP task. Participants also answered a set of knowledge questions which were designed to tap into different levels of understanding that students acquired about the materials.


The preliminary findings refer to a portion (using three pairs) of the proposed analysis reported in Jeong and Chi (1997) in this proceeding.

What kind and how much shared knowledge results from collaboration?

One way to answer this question is to capture and compare the mental models that students construct (see Chi et al., 1994 for a list of mental models that students construct in this domain). The preliminary findings show that although students all had different initial models in the pre-test, all the pairs shared the same mental model in the post-test. In other words, students converged on their model of the circulatory system after collaboration, although they sometimes converged on an incorrect model.

Another index of shared knowledge can be obtained by examining individual pieces of knowledge. The proportion of the terms shared at the post-test will be computed by using only the knowledge pieces shared on the post-test because the knowledge pieces shared from the pre-test cannot be considered an outcome of co-construction. The preliminary findings show that there exists some variability in the amount of shared knowledge and that the amount of sharing is not simply a function of how much individuals know.

Do successful pairs tend to share more knowledge?

As the preliminary findings show, not all pairs would share equal amounts of knowledge after collaboration. The next question, then, is how are the amounts of shared knowledge related to other outcomes of collaboration such as learning? To answer this question, the amount of shared knowledge pieces will be correlated with various measures of learning gains. Preliminary findings show that when pairs shared more knowledge, they also tended to learn more on average.

Self-construction versus co-construction of knowledge

It has been widely accepted that engaging in active learning such as generating self- explanations is beneficial to learning (Chi et al., 1994). Thus, first, all the explanations generated during collaboration will be coded to determine whether generating explanations is also an effective predictor of learning in collaborative learning situation. If this is the case, then it will demonstrate that the same underlying learning process is happening both when one learns alone and when one learns with others.

Generating explanations during collaboration, however, is more complicated than generating explanations to oneself: what one generates is often dependent on one’s partner’s action and explanations are often generated together with a partner rather than alone. Depending on how the explanations are generated, two different types of construction activities will be identified: self-construction versus co-construction of explanations. Self-constructed explanations will be defined as explanations that are generated by either member of the pair alone, whereas co-constructed explanations will be defined as an explanation to which both members of the pair contributed, which could be either in the form of supplementing the other partner’s explanation or answering the questions that the other partner posed. Both self-constructed knowledge and co-constructed knowledge are expected to be important in learning. However, self- and co-construction would lead to the construct! ion of different types of knowledge.


This thesis addresses the issues of knowledge co-construction in collaborative learning. The preliminary findings show that people achieve a shared understanding after collaboration both at the level of the mental model and at the level of individual knowledge pieces. In addition, the amount of shared knowledge seems to correlate with the amount of learning. Further analysis is planned to investigate the process of knowledge co-construction, specifically, the role of self-constructed and co-constructed explanation in learning.

A deeper understanding of the outcomes and process of collaboration would help us to take better advantage of this learning environment as well as to provide important clues for designing more effective computer supported collaborative learning environments. And this thesis hopes to make a contribution to this goal.


Chi, M. T. H., de Leeuw, N., Chiu, M., & LaVancher, C. (1994). Eliciting self explanations improves understanding. Cognitive Science, 18, 439-477.

Hardin, C., & Higgins, E. T. (1996). Shared reality: How social verification makes the subjective objective? In R. M. Sorrentino & E. T. Higgins (Eds.), Handbook of motivation and cognition: The interpersonal context (Vol. 3). New York: Guildford.

Jeong, H. & Chi, M. T. H. (1997). Construction of shared knowledge during collaborative learning. In Proceedings of Computer Support for Collaborative Learning 1997.

Roschelle, J. (1992). Learning by collaborating: Convergent conceptual change. The Journal of the Learning Sciences, 2(3), 235-276.