The effects of task characteristics on online discussion

A key guidance factor of computer supported collaborative learning (CSCL) is the specification of a discussion task. Aspects of the discussion task may affect the quality of group discussion for higher-order learning. This experiment investigated the effects of two aspects of discussion task on asynchronous text discussion of an online higher-education course. Groups completed discussion assignments that varied in degree of task context and outcome specification. Content analysis was used to assess conceptual conflict and level of information processing of online messages. Results indicate that conceptual conflict is associated with higher-order discussion, but differences in task context and product do not have large effects on the quantity or quality of online discussion. Introduction and Research Overview Wiley and Bailey (2006) describe process loss as the less effective performance of groups in completing some tasks. A significant body of research, however, has shown that collaborative learning groups, including online groups, can foster shared understanding, retention of learned material, and deeper processing compared to noncooperative learning activities (Johnson & Johnson, 1994; Johnson, Maruyama, Johnson, Nelson, & Skon, 1981; Slavin, 1987, 1992; Yeager, Johnson & Johnson, 1985). Other research supports the assertion that collaborative learning can promote higher-order learning such as critical thinking (e.g., Anderson, Howe, Soden, Halliday & Lowe, 2001; Gokhale, 1995; Meyer, 2003). Such research suggests that collaborative learning groups have characteristics that result in process gain in comparison to other group efforts. A major focus of collaborative learning research is to identify what characteristics result in process gain and how learning can be designed to maximize such gain in addition to meeting learning goals. Computer supported collaborative learning (CSCL) often relies on peer-to-peer discussion as the key activity supporting achievement of higher-order learning objectives. Hammond’s (2005) survey of online discussion studies lists several that cite evidence of higher-order knowledge construction and learning advantages of group discussion. Efforts to improve process gain of learning by discussion include efforts to understand how aspects of discussion task may affect the quality of peer discussion. This study investigated how two aspects of discussion task affect asynchronous online discussion (AOD) associated with higher-order learning. Characteristics of High-Quality Discussion for Higher-Order Learning Models of learning by discussion, such as the Process of Controversy model of Johnson and Johnson (1979) or the Collaborative Knowledge Building model of Stahl (2000), indicate that for higher-order learning to occur, information expressed in discussion must vary (diverge) sufficiently to achieve conceptual conflict among students. Conceptual conflict occurs when students encounter ideas and information that do not fit with what they believe to be true (Johnson & Johnson, 1979). These models assert that collaborative learning occurs when students encounter cognitive conflicts and then engage in group processing of information to identify or produce a shared interpretation that completes the discussion task. Discussions in which ideas and assertions diverge and conflict tend to promote learning, especially higher-order learning. In completing a group learning task, group members process shared information to identify or generate information that members agree resolves the task (i.e., information converges to a task solution). A group can process information by negotiation, questioning, and argumentation (Andriessen, 2006, Andriessen, Baker & Suthers, 2003; Spatariu, Hartley & Bedixen, 2004), but CSCL discussions often do not converge. Andriessen (2006) found that online discussion messages tend to be both unconnected (do not reference each other) and non-argumentative. Hewlitt (2005) found that students tend to focus only on the most recently posted messages, while older messages tend not to be reexamined or referenced. Lobry de Bruyn (2004) found low levels of analysis, synthesis, and summarizing (“convergent processes”) displayed in discussion messages. Online discussions often fail to integrate diverse ideas, opinions, and suggestions into new group knowledge that indicates higher-order learning. Peer discussion may fail to support higher-order learning because information does not sufficiently diverge to create conceptual conflict. When peer discussion does diverge sufficiently, students often do not connect the different ideas expressed and do not return to explain, summarize, or reach conclusions about issues. Measuring Discussion Quality Discussion divergence is beneficial if it stimulates conceptual conflict within group members (Spatariu et al., 2004), so one measure of quality of discussion for learning is the amount of conceptual conflict evident in the discussion. Such conflict, however, is insufficient for collaborative learning. Group members must also process information to identify or generate information that the group agrees resolves the discussion task. Information convergence is difficult to measure, but researchers can measure the type of information processing present in a discussion. Such processing, displayed in online text-only discussion messages, suggests whether information is diverging and converging. The type of information processing displayed in discussion messages also indicates whether higher-order learning is occurring. This study measured discussion conflict and level of information processing observed in online messages and indicators of discussion quality for learning. Measuring Collaborative Information Processing Several studies use content analysis to analyze the quality of online discussions or to access more detail about the collaborative learning process (De Wever, Schellens, Valcke & Van Keer, 2006; Rourke, Anderson, Garrison & Archer, 2001). Veerman, Andriessen, and Kanselaar (1999) classify “constructive activities” in messages into three categories: 1) added, explained, or evaluated; 2) summarized; 3) transformed. This scheme can be seen as an information processing approach in that constructive activities can be viewed as levels of information processing. Comparison of content analysis instruments used in CSCL studies reveals that several instruments tend to agree on only two basic classifications that are similar to the added and transformed categores of the Veeman et al. (1999) classification scheme (Jorczak, 2008). The measurement approach adopted for this study, therefore, used two categories of information processing displayed in messages: 1) adding/clarifying diverse information to the discussion from knowledge sources; and 2) generating (creating) information new to the group (and not obtained from a source) that resolves conceptual conflict and achieves group goals. The adding/clarifying category involves processing to obtain information including judging it relevant to the discussion task and clarifying or stating the information in a way that is meaningful to all group members. The generating category requires group or individual cognitive processing involving inferring or elaborating, resulting in relevant new information that comes from neither a source outside the discussion nor any member’s prior knowledge. Following this approach, this study created a content analysis instrument by which discussion messages were placed into one of three levels of information processing: repetitive (no additional information added to the discussion), additive, and generative. The differences in additive and generative levels are consistent with the distinction between lowerand higher-order learning adopted for this study. Measuring Conceptual Conflict Andriessen (2006) coded messages into six categories of “dialog moves” including statements, checks, challenges, counters, acceptances, and conclusions. Three of these categories (check, challenges, and counters) display disagreement (Andriessen, 2006; Veerman et al., 1999). The amount of disagreement expressed in a discussion is related to the amount of conceptual conflict present in the discussion. Agreement and disagreement are often explicitly expressed in discussion messages. Implicit disagreement can be identified by messages that check, challenge, and counter statements of other students (Veerman et al., 1999). Only two categories of messages are defined for measuring discussion conflict in this study: 1) neutral/agreeing and 2) disagreeing. Variables that Affect Discussion Quality Several variables have been suggested and investigated as affecting the quality of online discussion. For example, Wiley and Bailey (2006) suggest that task coordination, group interdependence, and amount of argumentation are factors that determine if collaborative learning displays process loss or gain. Lobry de Bruyn, (2004) found that instructional interventions can improve discussion convergence. Hewitt (2005) agrees that instructor interventions can shape electronic discourse, and he lists course design, software interface design, and individual student differences as factors that affect learning by discussion. Similarly, Veerman and VeldhuisDiermanse, (2006) suggest four categories of such factors: instructors, communication medium, students, and learning task. Instructional guidance has been identified as a necessary component of any instructional design (Kirschner, Sweller & Clark, 2006), and it is rare to observe effective interaction in spontaneous unguided student discussions (King, 2007; Weinberger, Stegmann, Fischer & Mandl, 2007). Discussion tasks are a key means for instructional designers to guide discussion toward more divergence and convergence of information. To date, lacking direction from research, task specification for productive online discussion has often been inadequate. Kirschner, Beers, Boshuizen, & Gijselaers (2008) opine: With respect to tasks, it is too often the case that the learning tasks are not suited to collaboration....They are often too closed (i.e., there is little room in the problem space to discuss), too easy (i.e., it can more efficiently be carried out by one person than by a team), or too controlled (i.e., there is little room for learner initiative).... (p. 404) CSCL researchers have suggested several task characteristics that may affect discussion quality, including task scripting, function, and goal. Scripting can improve discussion (King, 2007). Scripting involves detailed instructions that guide student discussion and may include a template of expected student responses, such as the labeling or diagramming of the discussion (e.g., Fischer & Mandl, 2005; Suthers, 2003). Tasks can be scripted to scaffold students to adopt specific modes of interaction (e.g., argumentative). Highly-structured tasks lessen the management burden on students and let them spend more time on the task (Veerman & Veldhuis-Diermanse, 2006; De Wever, 2003). Veerman and Veldhuis-Diermanse (2006) suggest a positive correlation of task structure with knowledge construction and also found that tasks with designated student roles or perspectives (a type of structuring) resulted in more discussion. Differences in task function are also thought to affect discussion. For example, the type of discussion task can directly affect the amount and quality of question asking and argumentation in discussion (e.g., Rose & Flowers, 2003; Wiley & Bailey, 2006). Controversial tasks can stimulate argumentation. Nussbaum (2005) found that tasks that specify different discussion goals substantially affect characteristics of student discussion. The goals “to persuade” and “to generate reasons” had the strongest effect on argumentation. The persuasion goal resulted in more conflict and debate. The goal “to explore” increased discussion divergence and resulted in more connected messages (Nussbaum, 2005). Task Context Naidu and Oliver (1999) are among the researchers who stress the importance of operating within a context during instruction. From a cognitive perspective, highly contextualized tasks (those providing specific and realistic details) promote the recall and sharing of student ideas and experiences, because the additional details of context stimulate students’ episodic memories of events within the proposed or similar contexts. Providing details of context should serve to activate schema in long-term memory and therefore enable students to provide more information about the discussion topic. Highly-situated tasks may increase the amount and quality of discussion by increasing the introduction of new and diverse information, thereby increasing opportunities for conceptual conflict and knowledge negotiation. Increased context of discussion tasks should make discussions more realistic which would, according to some proponents of situated learning, improve learning (e.g., Greeno, More & Smith, 1993). Theoretical models, such as social constructivism, suggest that detailed, or at least more realistic, contexts can be expected to promote learning. Social constructivist theory posits that “authentic” activities (those similar to activities encountered outside the classroom) have learning benefits such as the more realistic use of social resources and increased meaningful connections (Ormrod, 2008, p. 343). Online discussion tasks provide an opportunity to test whether increased detail of task context affects aspects of discussion associated with learning. Task Product One approach to the lack of discussion convergence is to specify tasks that require the creation of a final product or statement of group consensus. Wiley & Bailey (2006) point out that successful collaboration occurs when students must cooperate to achieve a goal (i.e., accomplish an interdependent task). It is likely that a specific end product or goal stimulates students to share information and to discuss and learn from the knowledge, experiences, beliefs and values of other students (Veerman & Veldhuis-Diermanse, 2006). Andriessen (2006) suggests that effective argumentative discussion requires that students share and maintain a focus on the themes and problems of the discussion task. Specification of a group product (e.g., a written statement of consensus, or creation of a product such as a slide presentation) may act to strengthen group focus on a topic. A specific task outcome may promote a merging of effort with clarity of goal that promotes the cognitive learning processes of knowledge negotiation and synthesis.