Academically Productive Talk (APT) is a content-independent framework of conversational behaviors that emphasizes the importance of social interaction to facilitate
accountability to a community of learners, to knowledge, and to accepted standards of
reasoning through collaborative discourse (Michaels et al., 2008; Adamson et al. 2013;
Resnick et al. 1993). APT codifies a body of conversational behaviors and interactions3
(i.e., “moves”) that 15 years of research show promote reasoned participation, foster
effective discourse, provide access to improved knowledge structures, yield increases in academic achievement, and facilitate transfer to other domains (Michaels et al., 2008;
3 Core moves include: “Say More,” “Press for Reasoning,” “Revoice,” “Restate,” “Add More,”
Adamson et al. 2013; Resnick et al. 1993, 2013; Kumar and Rosé, 2011; Rosé et al., 2015). Recent research in CSCL has employed the APT framework to design
collaborative “scripts4” and “intelligent conversational agents5” that confirm and extend prior findings. Studies using such tools report that APT effectively supports
collaborative learning (Adamson et al., 2014; Hamalainen, 2006; Weinburger et al., 2005) and yields substantial gains in learning based on pre/post test scores (Chaudhuir et al., 2009; Ai et al., 2010; Kumar et al., 2010).
Of particular interest is Adamson et al.’s (2014) summary of findings from a number of experimental studies employing dynamic APT scripts with conversational agents in high school and undergraduate STEM-focused courses. These studies tested the effects of two facilitative moves, Revoice and Agree/Disagree, on learning gains and student interactivity. In the experimental conditions the conversational agent, based on the automated evaluation of prior student contributions, inserted these moves into the flow of students’ synchronous chat discourse as they solved a CSCL task. The researchers present evidence that suggests that the use of these moves is followed by “pockets of intensive discussion” and yield gains in achievement, though the effects varied by context (i.e., difficulty of the material and age level of students).
For instance, in studies of 9th grade science learners (see Dyke et al., 2013), the Revoice move was found to be an effective support for collaborative learning when students engaged with challenging and/or unfamiliar material. In particular, the researchers report a significant and positive correlation between the number of
4 Schemas that facilitate collaborative learning and activity.
5 CSCL interfaces designed to provide interventional (i.e., facilitative) support for students engaged
in collaborative tasks/activities such as Basilica (see Kumar and Rose, 2011) and more recently, Bazaar (see Adamson and Rose, 2012).
“revoicable assertions” contributed by a student and those contributed by their teammates in their discussions, suggesting that the initiation of the Revoice move by the
conversational agent elicited more substantive and critical interactions between students. These effects were diminished, however, when the CSCL task material was more familiar (i.e., easier) in the same student sample.
Further studies conducted with college freshman found mixed effects (see
Adamson et al., 2014). For example, in a CSCL task in an engineering design course using material familiar to students, there was a negative effect on learning outcomes related to the use of the Revoice move by the agent. However, a significant and positive correlation in the control condition between the frequency of a student’s unprompted use of this move in the course of conversation and that of their teammates suggests that the endogenous use of the Revoice move by students increased interactivity in students’ discourse. The authors conclude that “more advanced learners are already good at articulating their own ideas [and that the Revoicing support] is unneeded for them. Rather, they need to be pushed beyond that to connect to the reasoning of their partner students” (ibid, p. 120).
In a fourth study (again with college freshman) the experimental condition centered on the use of the Agree/Disagree move in a chemistry-focused CSCL task. Findings indicated that the use of this move by the agent intensified student interactions and yielded increases in learning. As above, this study also indicated a significant and positive correlation between the number of “revoicable assertions” contributed by a student and those contributed by their teammates in their discussions, indicating that the Agree/Disagree intervention “precipitated pockets of intensive discussion.”
Although further research is needed to establish the effects of other APT moves in the framework on student discourse, this emerging work is highly promising. Of
particular interest is that these initial findings suggest that once an APT move is
introduced into student discourse there may be a group level effect on their usage. Said another way, as students introduce collaborative moves in the flow of conversation, as exemplified by APT, there is emerging evidence that a tone may be set within the group for their more frequent use in subsequent interactions. Following from this, it is also reasonable to consider whether more collaborative interactivity during group discourse yields unique types of meaning making and/or characteristically different knowledge structures in a given context, such as is found in games and simulations for learning. As summarized above, research has established that there are numerous factors that influence social interaction during collaborative activity. Furthermore, research shows that APT effectively supports collaborative social interaction and promotes
increased learning gains for individuals. Little is known, however, about the relationship between APT and the learning of, or knowledge structures developed by, groups through discourse in situated social activity. From a socio-cognitive perspective this is an
important consideration and implies that the group, not the individual, may be the appropriate unit of analysis in the study of particular types of social interaction on the process, or outcomes, of collaboration (Stahl, 2006; Dillenbourg et al., 1996;
Dillenbourg, 1999). This position is further supported in literature that suggests, because a group’s discourse is a representation of knowledge in and of itself (i.e., group
cognition), it is the “property” of the group and its collective achievement and meaning making, and not merely the sum of each individual’s knowledge and contributions to the
group’s discourse (Clark, 2001; Scardamalia and Bereiter, 2010). By extension, Stahl (2008) further suggests that, if knowledge building is situated in groups, “we can observe the construction and evolution of the knowledge in the artifacts that are produced, in the sentences spoken, sketches drawn, and texts inscribed” (p. 70).
Finally, another question that has yet to be considered is whether the experience of APT in discourse affects an individual’s attitudes and perception in relation to the domain of study around which they are engaged in discourse. One way to fill the aforementioned gaps in the literature is to apply the APT framework for analysis in an
environment that reflects a socio-cognitive theory of collaborative activity, such as those found in games and simulations.