Although collaborative learning has been proved significant for students both for social and cognitive reasons (Slavin, 1996), collaborating students usually fail to engage in productive learning interactions when left without teachers' consistent support and scaffolding (e.g., Hewitt, 2005). Currently, issues regarding the adaptive operation of CSCL (computer-supported collaborative learning) systems attract the increasingly intense efforts of various research groups (e.g., Walker et al., 2009). These efforts advance the tradition of Adaptive Hypermedia Environments toward CSCL and expand the perspective of the field while setting innovative research agendas. In general, adaptive collaboration support techniques aim to model the major aspects of the collaborative activity and activate learner/group support interventions when needed and in the form it is needed (Soller et al., 2005). Although, there have been reported some encouraging first results (e.g., Kumar et al., 2007), there are also implementations that do not prove that such type of interventions lead to enhanced learning outcomes (e.g., Baghaei et al., 2007). Moreover, most of these systems are research prototypes that demonstrate possible system architectures or have been used to showcase their beneficial learning impact but are not widely available outside the research laboratory.
Based on the above drawbacks, we investigate if the integration of adaptive domain-specific support in a scripted collaborative activity would lead to better learning outcomes compared to a fixed support mechanism. In the following, we present (a) the theoretical background of our research, (b) the study design and results, and (c) a discussion analyzing the learning impact of the collaboration support method.
Support collaboration using collaboration scripts
Collaborative learning has been proved important for students for social, cognitive and meta-cognitive reasons (Slavin, 1996). However, when students are engaged in collaborative learning they need significant support and guidance since they are rarely engaged in productive interactions such as asking each other questions or reflecting upon their knowledge (Hewitt, 2005; Liu & Tsai, 2008).
A first step toward providing the kind of student support necessary in collaborative processes has been to script the activity (Fischer et al., 2007). Scripts structure the collaborative process by defining sequences of activities, by creating roles within groups and by constraining the mode of interaction among peers or between groups (Dillenbourg & Tchounikine, 2007). Implementing CSCL scripts has been reported to result in improved learning outcomes (Fischer et al., 2007; Hernandez-Leo et al., 2006). However, CSCL scripting has been criticized for its loss of flexibility (Dillenbourg & Tchounikine, 2007), and also the danger of "over-scripting" collaborative activity (Dillenbourg, 2002).
Supporting collaborative learning through adaptive/intelligent interventions
Current CSCL efforts have focused on supporting groupwork through the use of adaptive and/or intelligent systems. Adaptive and intelligent interventions tailor the collaborative learning process to the needs of the individual students or groups. The target of the adaptive/intelligent interventions varies and can be classified into 2 main categories: 1) peer interaction support (i.e., help peers to "learn to collaborate"), 2) domain knowledge support (i.e., help peers to deepen their domain understanding) (Magnisalis et al., 2011).
Peer interaction support refers to the actions taken by the system in order to help learners improve their interaction and develop domain-general knowledge and skill (Soller et al., 2005). For the purpose of this study, we focus on the second category of support. Domain-specific support refers to the actions taken by the system in order to help learners understand the domain better. …