What interdependence can tell us about collaborative learning: a statistical and psychological perspective

Lenka Schnaubert,Daniel Bodemer

doi:10.1186/s41039-018-0084-x

Lenka Schnaubert, Daniel Bodemer

Open Access

https://doi.org/10.1186/s41039-018-0084-x

Copy DOI

Abstract

When learning collaboratively, learners interact and communicate transactively. Interventions to foster collaborative learning frequently target such interactive processes and thus may drastically change how learners engage with and thus influence each other. One statistical phenomenon related to collaborative learning is the interdependence of data gained from learners collaborating. Often viewed as a mere statistical phenomenon, on a conceptual level, statistical interdependence is a similarity between learners mainly resulting from the mutual influence learners have on each other while collaborating and is thus closely related to collaborative practices. In this paper, we report data of an exemplary study (N = 82) to illustrate how information on interdependence and within- and between-dyad variance may add to data interpretation. The study examined how providing metacognitive group awareness information during collaboration affects individual learning outcomes. We found indications that the information fosters knowledge gain, but not confidence. Surprisingly, the data revealed different levels of interdependence between conditions, which led us to assume interdependence to be part of the treatment effect resulting from differential collaboration processes. We discuss reasons and implications of varying levels of statistical interdependence and their impact on inferential and descriptive statistics.

Highlights

Collaborative learning (CL) yields a lot of potential to foster knowledge construction
In this paper, we argue that (1) statistical interdependence after collaboration is something to be expected and even hoped for in CL; (2) assessing the intra-class correlation (ICC) on a sample level is flawed on principle, because variance caused by the treatment will taint the results and lead to overestimations of interdependence within dyads; (3) interdependence can be highly influenced by interventions targeting collaborative learning processes and may differ dramatically between experimental conditions; and (4) information on interdependence is valuable and indicative of collaborative processes, and should be explicitly and critically reviewed
While we are aware that multi-level approaches may account for such differences, we argue that statistical interdependence is not primarily a statistical nuisance to be eliminated from our data, but a valid diagnostic outcome to be explicitly discussed in research on collaborative learning, as it is the core of collaboration

Summary

Introduction

Collaborative learning (CL) yields a lot of potential to foster knowledge construction. They exchange and commonly build knowledge and/or skills. One important issue is that collaboration is an interactive activity of learners that is thought to foster group performance and individual learning (Hesse 2007). The data collected is frequently on different levels (individual and group) and/or heavily intertwined (like in turn-taking during discussion) (cf Strijbos and Fischer 2007). This poses a great challenge for quantitative research, because traditional analyses

Objectives

Methods

Results

Discussion

Conclusion