Chapter 2. Literature Review
2.6 Conceptualizations and Measures of Knowledge Construction in
2.6.1 Henri (1992)—Learning Process Framework
Henri (1992) developed five dimensions to describe the process of learning in CSCL environments, which were Participation, Social, Interactive, Cognitive, and
Metacognitive. The Henri’s scales for the Participation and Interactive dimensions are not discussed here because they are not relevant to knowledge construction. The details of the Social, Cognitive, and Metacognitive dimensions are given in Table 2-2 (The Social dimension did not describe knowledge construction, but played an important supplementary role in the analysis of knowledge construction). Henri suggested using a sentence as a unit of analysis, although his scale is applicable at other levels, such using a message as a unit of analysis.
Table 2-2 Henri's (1992) Dimensions of Learning
Dimension Subdimension Scale Levels Categories
Social Nominal related to the
subject, unrelated to the subject
Cognitivea Ordinal elementary clarification, in-depth clarification, inference, judgment, strategy
Metacognitivea Knowledge Categorical
assessment of the task, learner self-perception, means to succeed Metacognitivea Skill Categorical
evaluation, planning,
regulation, self-awareness aAssessed only for units related to the subject of discussion.
The Social dimension reflected if the content of the unit of analysis was related to the intended subject of the discussion. The Cognitive and Metacognitive dimensions were assessed only for units related to the subject of the discussion (and did not make sense for units that were purely social in nature).
The Cognitive dimension reflected the level of critical thinking exhibited in a sentence by its contributor and was measured on an ordinal scale, thus allowing comparison of the levels of critical thinking between units.
The Metacognitive dimension reflected awareness, self-control, and self-regulation of learning. The subdimensions of Metacognitive, Knowledge and Skill, were measured on categorical scales.
The known uses of the measure developed by Henri (1992) are discussed in the following subsections.
2.6.1.1 Bullen(1998)—ParticipationandCriticalThinking
Bullen (1998) adapted the Henri’s (1992) model to investigate the use of computer- mediated communication (CMC) to support the process of learning. Bullen aimed to determine whether the students were actively building on each other’s contributions and thinking critically about the discussion topics.
Bullen (1998) studied synchronous online discussions in a computer information systems class.
Table 2-3 Bullen’s (1998) Scale for Critical Thinking
Dimension Scale Levels
Critical Thinkinga Ordinal
low, moderate, high aMeasured using an adaptation of the Henri’s (1992) scale for cognitive.
Bullen (1998) adapted a subset of the Henri’s (1992) scale (the Cognitive dimension only) to measure the level of critical thinking (see Table 2-3). The data were analysed as qualitative data, with the level of the Critical Thinking dimension used as a
descriptive measure informing the qualitative analysis. Inter-rater reliability was not assessed.
Bullen (1998) found evidence of critical thinking.
2.6.1.2 Hara,Bonk,andAngeli(2000)—KnowledgeConstructionand Interaction
Hara, Bonk, and Angeli (2000) used the Henri’s model to analyse interactions between students and teachers in online discussions. Hara et al. focused on the following aspects of the interactions: (a) student participation rates, (b) electronic interaction rates, (c)
social cues within student messages, (d) evidence of cognitive and metacognitive processing in student messages, and (e) the depth of processing within student messages.
Content analysis was used for (c) and (d): Hara et al. (2000) used the Henri’s (1992) scales for Social, Cognitive, and Metacognitive. Hara et al. found it difficult to use the Knowledge sub dimension of Metacognitive together with Cognitive because they found them to be similar. Therefore, they used only the Skill sub dimension of Metacognitive in their final analysis.
Hara et al. (2000) analysed 271 messages collected over 15 weeks of discussion by 20 students enrolled in a graduate Psychology course at a university in the USA. A paragraph was used as a unit of analysis (the number of paragraphs in the messages analysed was not reported).
Independent researchers were used as coders. It was not reported if the coders were trained. The coders negotiated the final score, but the details of the negotiation protocol were not reported. Inter-rater reliability was assessed for the initial scores (the scores before negotiation) using percent agreement; the details of inter-rater reliability results are summarized in Table 2-4.
Table 2-4 Inter-Rater Reliability in Hara et al. (2000)
Measured by Social Cognitive Metacognitive / Skill
Percent agreement 0.78 0.75 0.71
Percent agreement has been criticized as a measure that overestimates inter-rater agreement (Riffe, Lacy, Fico, & Fico, 2005), which is particularly true for nominal scales. Hara et al. (2000) used a nominal scale for Social, and the scales for Cognitive and Skill subdimensions of Metacognitive involved, respectively, five levels and three categories. For a scale involving three choices, chance agreements are less likely than for a nominal scale (so that random sequences result in 33.33% agreement compared to 50% for a nominal scale). For a scale involving five choices, chance agreements are even less likely. Thus, one can judge the inter-rater reliability achieved by Hara et al.
for Cognitive and Metacognitive as relatively good, even though it was below the 0.8 threshold suggested by Riffe et al. (2005).
Based on a qualitative analysis of messages (informed by the outcomes of the content analysis), Hara et al. (2000) concluded that online discussions promote cognitive processes.
2.6.1.3 Pena‐Shaff&Nichols(2004)—KnowledgeConstructionand Participation
Table 2-5 Categories Retained by Pena-Shaff and Nicholls’(2004) That Matched Levels and Categories in the Henri’s (1992) Model
Pena-Shaff & Nicholl
(2004) Henri (1992)
Category Category or level subdimension Dimension /
clarification elementary
clarification and in depth clarification
Cognitive
interpretation inference Cognitive
judgment judgment Cognitive
reflection self-awareness Metacognitive / Skill
reply direct responses and
direct commentaries
Interactive
other unrelated with the
subject
Social
Pena-Shaff and Nicholls (2004) studied student interactions and meaning construction in online bulletin board discussions. Pena-Shaff and Nicholls developed 11 knowledge construction categories based on qualitative analysis of the messages. Qualitative analysis started by using a coding scheme that included categories and levels from the Henri’s (1992) model (along with categories from other related work), which were simplified for the purposes of the qualitative analysis. The coding scheme was allowed
to evolve inductively: New categories were allowed to emerge in the course of the analysis, and not all of the initial categories were retained.
Pena-Shaff and Nicholls (2004) analysed 152 messages in discussion threads randomly sampled from discussions by 35 students enrolled in an advanced communication course at a university in the USA. A sentence was used as a unit of analysis.
Many of the categories retained after the analysis was completed matched the
categories and levels of the dimensions in the Henri’s (1992) model. The categories in the Pena-Shaff and Nicholls’ (2004) analysis that matched the categories or levels of the dimensions in the Henri’s model relevant to knowledge construction are given in Table 2-5.
Pena-Shaff and Nicholls (2004) found the highest proportions of messages in the
clarification (44%) and interpretation categories (15%), which matched low and
middle levels of the Cognitive dimension in the Henri’s (1992) model.
Because Pena-Shaff and Nicholls (2004) used qualitative analysis, inter-rater reliabilities did not apply (and were not reported).
2.6.2 Gunawardena,Lowe,andAnderson(1997)—KnowledgeConstructionin