Professional Documents
Culture Documents
doi: 10.1111/j.1365-2729.2006.00206.x
Abstract In this study, constraint-based argumentation scaffolding was proposed to facilitate online
argumentation performance and ill-structured problem solving during online discussions. In
addition, epistemological beliefs were presumed to play a role in solving ill-structured
diagnosis–solution problems. Constraint-based discussion boards were implemented to scaf-
fold pre-service teachers’ online discussions about behaviour management (diagnosis–
solution) problems. The scaffolded discussion group generated more evidence notes and also
generated more hypothesis messages and hypothesis testing messages as well as problem space
construction messages. There was a relationship between epistemological beliefs and ill-
structured problem solving. Simple knowledge, omniscient authority, and fixed ability signifi-
cantly predicted problem-solving performance. A significant negative relationship between
simple knowledge and individual problem-solving performance was found. This implies that
individuals who believe in simple knowledge may be less inclined to explore more solution
alternatives. However, contrary to prediction, omniscient authority and fixed ability beliefs
were positively associated with problem-solving processes.
© 2007 The Authors. Journal compilation © 2007 Blackwell Publishing Ltd Journal of Computer Assisted Learning (2007), 23, 95–110 95
96 S. Oh & D.H. Jonassen
(Elstein et al. 1978). Argumentation (also known as jus- problem representation, justifications, and monitoring
tification) involves constructing, refuting, and compar- and evaluation. Davis (1998) examined whether reflec-
ing arguments using a variety of types of reasoning tion prompts promoted knowledge integration for stu-
(Andriessen et al. 2003). Argumentation has the poten- dents working on science projects in the Knowledge
tial for engaging students, modifying underlying Integration Environment. She found that the reflection
beliefs, making students’ thinking visible, and refuting prompts affected students’ ability to critique solutions.
misconceptions (Baker 1999). Nussbaum and Sinatra Cho and Jonassen (2002) showed that providing
(2002) believe that argumentation has potential as a constraint-based argumentation scaffolds during col-
conceptual change intervention because students get laborative problem-solving sessions increased the
involved in the deep processing or high engagement. generation of coherent arguments.
Baker (1999) found that students’ flawed beliefs or An emergent technology for scaffolding argumenta-
claims were weakened as a result of argumentation. tion is known as computer-supported collaborative
Arguing helps students to compare their conceptualiza- argumentation (CSCA). CSCA environments preclassi-
tions with others and recognize potential points of fying conversational attributes to fit sets of canonical
conflict for further discussion (Suthers 2003). Argu- relations constrains the nature of verbal interactions
mentation is also essential to achieving agreement on a among conversants. For example, the Belvedere envi-
specific diagnosis for problems as well as deciding on ronment provides four predefined argumentation
the most appropriate solution. The ability to argue effec- constraints (‘hypothesis’, ‘data’, ‘principles’, and
tively for different positions is tantamount to solving ‘unspecified’) and three links (‘for’, ‘against’, and
diagnosis–solution problems. ‘and’) (Suthers 1998). These constraints form the links
Several studies have shown positive effects of argu- or relations between the ideas that conversants produce.
mentation during problem solving (Burnett 1993; Other discussion-based tools, e.g. CSILE (Scardamalia
Veerman & Treasure-Jones 1999). Burnett (1993) found et al. 1994), Knowledge Forum (Scardamalia 2002),
that students who were involved in substantial conflict and Future Learning Environment (FLE), support social
during the writing assignment considered more alterna- negotiation and the explanation of informal reasoning.
tive solutions and finally produced better papers. Knowledge representation tools, such as IBIS (Rittel
Students who did not engage in the substantial conflict 1984), ArgueTrack (Bouwer 1998), Belvedere (Suthers
compromised too soon and produced poor solutions. 1998), and Compendium (Conklin 2003), help students
Two disparate approaches to learning argumenta- see argumentation structure, thus facilitating the con-
tion skills exist. The first, direct instruction, does not struction of rhetorical argumentation. In this study, we
always improve argumentation skills as expected. Some investigated the effects of scaffolded argumentation on
research indicates that direct instruction enhances argu- argument construction and problem solving.
mentation skills (Sanders et al. 1994), whereas other An important factor that has been shown to affect
research demonstrates no positive effects for direct argumentation and problem solving is the learner’s
instruction on improving argumentation skills epistemological beliefs (Kitchener 1983; Schraw et al.
(Knudson 1991). 1995). Schraw et al. (1995) found that well-structured
Another innovative approach to supporting argumen- and ill-structured problems engaged different epistemo-
tation in online environments is to scaffold argumenta- logical beliefs. Individuals who view knowledge from a
tion performance among students by question relativistic perspective adopted multiple strategies to
prompting or constraining the kinds of comments that analyse contradiction and ambiguity on ill-structured
students can contribute to an online discussion (Jonas- problems. Epistemological beliefs affect written argu-
sen & Remidez 2005). A small amount of research has ments about an ill-defined problem (Bendixen &
shown the positive effects of scaffolded argumentation Schraw 2001). Schommer and Dunnell (1997) found
during problem solving. Ge and Land (2003) investi- that the more students believed that the ability to learn is
gated the effects of question prompts in scaffolding fixed at birth, that learning is quick or not-at-all, and
college students’ problem-solving process on an ill- that knowledge is unchanging, the more likely they
structured task. They found that students working with wrote overly simplistic solutions to problems. Finally,
question prompts outperformed the other groups in Schommer-Aikins et al. (2005) found that general
Desmond is a 10-year-old fourth grader attending a large elementary school. His teacher
has asked for help with his behavior. He is not a bad kid but is constantly disrupting the
classroom. For example, if the class is having a discussion, Desmond will add a gross tidbit
to the conversation. One day he described in graphic detail how the family cat looked after
a car hit it. Desmond seemed to enjoy the disgusted looks of his classmates. Desmond is
constantly out of his seat bothering other students, poking them with his pencil or ruler,
blurting out answers, commenting without raising his hand, and interrupting lessons with
irrelevant comments. Desmond constantly complains that other students are bothering him
but when the complaint is investigated, he is the instigator. Desmond's peer relations are
poor. Other students do not want to be near him or work with him. They complain that he
makes sexual comments. Desmond denies making these comments. When he is caught
misbehaving, he always blames his problems on another student. Academically, Desmond
is not passing the fourth grade. Most written assignments are not completed or poorly
completed. He does not proofread or check his answers. Only 25% of his homework is on
time. He is highly distractible, has a very short attention span, and is impulsive. Fig 1 A case description of the individual
problem-solving test.
was classified by type of argumentation (evidence, reliability of each student’s problem-solving processes
elaboration, verification, and rebuttal), based on the was calculated. Two raters generated these ratings, k =
argument coding schemes developed by Meyers et al. 0.55–0.69.
(2000) and Toulmin (1958). For example, the postings Third, an Epistemic Beliefs Inventory (EBI) (Schraw
that determine the value of an argument were coded as et al. 1995) was administered before the treatment to
verification, and the postings that refute or challenge examine students’ epistemological beliefs. The EBI
others’ arguments were coded as rebuttal. The postings consists of 32 items that reflect beliefs about the nature
that ask questions seeking additional information for of knowledge and learning. Their exploratory factor
clarifying ideas were coded as elaboration, and the post- analysis yielded five factors, including certain knowl-
ings supporting an argument with a scholar’s work, per- edge (e.g. truth means different things to different
sonal experience, individual beliefs, and research people), simple knowledge (e.g. things are simpler than
findings were coded as evidence. We assumed that each most professors would have you believe), quick learn-
message from the scaffolded discussion group was one ing (e.g. if you don’t learn something quickly, you
argumentation type because the participants were won’t ever learn it), omniscient authority (e.g. when
required to select one type of argumentation before they someone in authority tells me what to do, I usually do
submitted their message. Therefore, the unit of analysis it), and innate ability (e.g. smart people are born that
for each discussion group was each individual message. way) (Schraw et al. 2002). Each item was rated on a
A total of 277 messages were collected from the discus- five-point Likert-type scale. The overall mean of each
sion boards across the two discussion sessions and clas- factor was used as the independent variable. Our repli-
sified by two raters. Although per cent agreement is cation analysis revealed that the EBI provides the same
used widely in research that involves the coding of number of factors, the same item-to-factor loadings for
behaviour, it misleads measures that overestimate true each test item, the same amount of sample variation
interrater agreement (Lombard et al. 2002). Therefore, explained, and an acceptable test–retest correlation
Cohen’s k was used to examine agreement between two among the five factors. The instrument has 0.63 to 0.87
raters on coding individual messages with argumenta- Cronbach’s a for items within each factor.
tion coding scheme. The interrater reliability between Fourth, a case scenario (see Fig 1) was administered
the two raters’ coding was good, k = 0.81. to each participant after the treatment. Students’ indi-
Second, each written posting during the discussions vidual problem-solving skills were assessed using their
was then reclassified by type of problem-solving activ- essay responses to the problem case. Based on the case
ity (problem space construction, hypothesis generation, description, the following questions were asked:
hypothesis testing, solution generation, and solution
verification) based on the diagnosis–solution problem- 1 What do you think the problem is? What does
solving process. In order to verify the reliability of Desmond do that causes concern? Please specify
the assessment of problem-solving process, interrater what behaviour is causing concern and identify any
other information you feel is relevant (known factors a full description of the scaffolds). Sentence openers
and constraints). were provided in the body of the message input box,
2 What are the possible causes of the problem in Des- accompanied with the chosen message type. At the
mond’s case? Please list the most possible cause(s) top level, students may select between ‘What is the
and support your hypothesis with argument and cause . . .?’ or ‘What should I do about it?’ When con-
evidence. structing messages responding to one of those claims,
3 What does Desmond seem to gain from behaving this students must select an appropriate warrant, such as ‘I
way? Support your argument with evidence. agree because . . .’ or ‘I don’t agree because . . .’ They
4 What strategies or solutions do you think are effec- enter text that elaborates their message type. When
tive? What do you expect from your solution? Please responding to a warrant, students must select a kind of
justify your solution(s) with evidence and evaluate evidence to support the other students’ statements. They
your solution(s). can select from four options: ‘My experience is . . .’; ‘I
5 What are alternative solutions? How are they com- believe . . .’; ‘Research shows . . .’; ‘A scholar says . . .’
pared with your proposed solution? What argument These sentence openers were expected to constrain
can you make or what evidence can you provide to students’ discussion moves and learning interactions
support that your solution is the most viable? to those believed to be productive for learning
(Constantino-Gonzales & Suthers 2002). In a study
Individual problem-solving skills were assessed examining the effects of sentence openers and elabo-
using scoring rubrics for assessing diagnosis–solution rated cases on online discussions, sentence openers
problem-solving process: problem space construction affected the frequency of disagreement, thus stimulat-
(defining the problem, identifying relevant information, ing reflective discussion (Nussbaum et al. 2002). The
searching and selecting needed information, developing result indicates that sentence openers are useful for stu-
a justification); hypothesis generation (generating dents with low degrees of assertiveness and who are not
initial hypotheses); hypothesis testing (generating alter- overly anxious. Students build dialogues and debate by
native hypotheses, activating fault schema and consid- posting their arguments into a shared database. Students
ering implications of hypothesis from fault schema, think in structured ways that mimic argumentation
justifying reasoning in hypothesis); solution generation skills by defining the relationships between knowledge
[developing solution alternatives, selecting solution(s)]; types and constraining knowledge types used to
solution verification [evaluating solution(s), construct- response to other message types.
ing argument(s), providing evidence, assessing alterna- Students in the threaded discussion group discussed
tive solutions]. The interrater reliability ranged from the same cases online using a threaded discussion board
0.54 to 0.71. without constraint scaffolds. Students in the control
group engaged in no discussion about the cases. They
were required to analyse the case individually (Figs 2
Materials
and 3).
The online discussion environments of the two treat-
ment groups were exactly the same except for the
Procedure
argumentation scaffolding provided to the scaffolded
discussion group. Scaffolded students posted their After the initial meeting to gain consent, participants
analyses of two diagnosis–solution problems and solu- were asked to complete the EBI to measure their episte-
tions with the help of the constraint-based discussion mological beliefs. After completing the EBI instrument,
board, FLE3. FLE3 is a web-based threaded conferenc- participants in the scaffolded group were trained on how
ing environment where all student postings must be to use the FLE3 environment and those in the threaded
typed before submission. Typing is a process of select- group were trained on how to use the threaded discus-
ing the most appropriate type of statement or response sion board.
to make. Students could select from one of six message During the 2 weeks that the topic of classroom man-
typed: Hypothesize cause; Solution generation; Verifi- agement was taught by the professor, the participants
cation; Rebuttal; Evidence; Elaboration (see Table 1 for received regular classroom instruction on classroom
Problem identification Constructing and articulating The problem is How do I define the problem?
(What information do the problem space in which What do you think is the problem?
I need to solve the problem solvers work out a What are the assumptions of the problem?
problem?) solution
Hypothesize cause Identifying the causes of the The problem is caused by What causes the problem?
(What is the cause?) problem by generating initial
hypotheses
Solution generation Generating the solution(s) My solution is What are possible solutions?
(What should I do to the problem
about it?)
Verification A statement that analyses I agree with what you said How do you verify your hypothesized cause?
(I agree because . . .) and determines value of an because How can you justify your solution?
argument How can you verify the accuracy or value
of your solution?
Are there any alternative solutions?
Rebuttal A statement that examines I disagree with what you What are conflicting issues?
(I don’t agree because) and finds flaws or weakness said because How can you support your rebuttal?
in others’ argument Why do you not agree with the argument?
Evidence Statements that provide A scholar says Is your evidence from an authoritative source?
(A scholar says . . . scholars’ work, personal My experience is How relevant are your experiences?
My experience says . . . experience, individual beliefs, I believe that What are your personal beliefs?
I believe . . . and research findings to Research shows that What are the research findings?
Research shows . . .) support or rebut a stated
argument
Elaboration A statement that asks Could you please clarify What are missing information in
(I can’t understand your questions seeking additional what you said? explanation?
point. Can you describe information for clarifying ideas What is unclear explanation?
more specifically?)
S. Oh & D.H. Jonassen
Pre-classified
knowledge type
Sentence opener
Guided questions
and behaviour management. Each week of the 2-week board, FLE3, or the threaded board. All participants
period of the study, the participants in the two treatment from these two treatments posted initial messages
groups were requested to solve one classroom behav- describing their problem-solving process on the class-
iour management problem using an online discussion room and behaviour management case, so the initial
postings were available for all of the small-group Some participants were eliminated from the sample
members to review. Each student was also required to because of lack of participation. No univariate outliers
read the initial postings and to ask questions or to were found. By using Mahalanobis distance with P <
provide different opinions to any member in the small 0.001, no multivariate outliers were found in the data
group. The participants’ online verbal interactions and sets. With no outliers and the case with missing values
their problem-solving and argumentation process were eliminated, 44 cases remained in the three groups.
recorded in a database in a computer for later analysis. An assumption of linearity was examined using scat-
Students in both treatment groups were asked to terplots and found to be satisfactory because the residu-
define the problem, identify the cause of the problem, als scatterplots did not show any nonlinear relationships
and propose solutions (Fig 4). The difference in the among variables. In the same way, in order to detect het-
treatments among the three groups was that the treat- eroscedasticity, the scatterplots of residuals were used.
ment A group received argumentation scaffolding in the There were no patterns to the residuals plotted against
form of structured interaction; the treatment B group the fitted values, which indicates that the variance of the
received no argumentation scaffolding in the online dis- residuals is homogenous across levels of the predicted
cussion board; and the control group did not participate values, known as homoscedasticity.
in online discussions. They were required to submit an
individual case study report for problem analysis and Effects of the constraint-based argumentation
solution of the classroom and behaviour management scaffolds on online argument building
cases. During the online discussions, students in both
treatment groups analysed and discussed the same The research question addressed the effect of the
behaviour management case studies, but only students in constraint-based argumentation scaffolds on online
the treatment groupAwere guided with online argumen- argument building and diagnosis–solution problem-
tation scaffolds. Following the 2 weeks of discussion, solving process. In order to answer these questions, all
participants completed a problem-solving essay test. students’ postings were collected from the two case-
study discussions. The postings were classified as one
of the four types of argumentation (verification, rebut-
Results
tal, evidence, and elaboration). After coding all of the
Before data analysis, frequencies of online argumenta- postings, the frequency of each type of argumentation
tion types, frequencies of problem-solving types, scores was counted from each of the two discussions,
of EBI, and scores of individual problem-solving respectively.
performance were summarized. The preliminary Table 2 shows descriptive statistics for the frequency
data analysis was to detect missing values and the accu- of different types of argumentation in the online discus-
racy of data entry, and to test the assumptions of multi- sions and indicates that verification was the most fre-
variate analysis. quent type of posting generated by the students.
First discussion
Scaffolded 14 31 10 22 2 65
Threaded 15 60 17 2 3 82
Total 29 91 27 24 5 147
Second discussion
Scaffolded 14 42 7 9 2 60
Threaded 15 51 9 5 5 70
Total 29 93 16 14 7 130
Total discussion
Scaffolded 14 73 17 31 4 125
Threaded 15 111 26 7 8 152
Total 29 184 43 38 12 277
The scores of the four types of argumentation were there was no difference between the two groups’ perfor-
summed and included in the statistical analysis. mance across the two discussions. Specifically, the
Descriptive statistics for the scores of argumentation treatments did not affect students’ performance in argu-
performance between the two treatment groups across mentation building throughout time. The main effect of
two discussions are presented in Table 2. A manova time of measurement was not significant, F1,27 = 0.413,
was employed to analyse the effects of the constraint- P = 0.526, h2 = 0.015.
based argumentation scaffolds on students’ online argu- In order to assess differences among the argumenta-
mentation types. The combined dependent variables tion types used by the students, an anova showed that
were significantly affected by argumentation scaffold- there was a significant main effect for argument,
ing, F1,27 = 5.99, P < 0.01. To investigate the impact of F2,157 = 60.215, P = 0.000, indicating that there was
the use of argumentation scaffolds on the dependent a significant difference among the four types of
variables, follow-up anova tests were performed. argumentation. The verification and the rebuttal scores
Univariate tests revealed that there were significant sta- were higher than the other argumentation types in the
tistical differences in the evidence variable (F = 16.55, argumentation performance. This indicates that stu-
P < 0.001) and the verification variable (F = 4.58, dents provided more verification and rebuttals than pro-
P < 0.05). The scaffolded discussion group posted viding evidence and asking more elaboration. There
significantly more evidence notes than the threaded dis- was also a significant effect for the argument-by-group
cussion group. However, the threaded discussion group interaction, F2,157 = 4.161, P = 0.013, indicating that
posted significantly more verification notes than the there were statistically significant differences between
scaffolded discussion group. No other argumentation the two groups in the argumentation performance. The
types differed significantly between groups. scaffolded group produced more evidence messages,
A multiple repeated measures (two discussions) while the threaded group produced many more verifica-
anova with between-subjects factors (two groups) for tion messages. There was a three-way interaction
the argumentation performance scores was conducted between time, argument, and group, F2,69 = 5.328,
because there were two discussions within subjects and P = 0.002. The threaded group produced proportion-
two groups with different treatments. The result for ately more verification messages in the first case than
within-subjects by between-subject interaction effects they did in the second.
(an interaction for time by group effects and time by
argumentation by group effects) was performed to test
Effects of the constraint-based argumentation
the hypothesis. The results showed that there was no
scaffolds on problem-solving process
significant effect for the time-by-group interaction,
F1,27 = 0.887, P = 0.355, or for the time-by-argument The second research question addressed the effect of
interaction, F2,69 = 2.105, P = 0.116. This means that the constraint-based discussion board on the types of
PSC, problem space construction; HG, hypothesis generation; HT, hypothesis testing; SG,
solution generation; SV, solution verification.
problem-solving processes engaged in by the students. significant, F1,27 = 2.914, P < 0.05. The univariate
Frequency counts of each problem-solving message anova tests indicate that there were significant differ-
type are included in Table 3. A multiple repeated mea- ences in problem space construction, F = 4.801, P =
sures (two discussions) anova with between-subjects 0.037, hypothesis generation, F = 10.134, P = 0.004,
factors (two groups) for the problem-solving processes and hypothesis test, F = 10.153, P = 0.004. The univari-
was conducted because there were two discussions ate F-tests of group differences showed that the scaf-
within the same subjects and two groups with different folded discussion group generated more problem space
treatments. The result for within-subjects-by-between- comments, more hypotheses, and more testing of
subjects interaction effects (an interaction for hypotheses.
time-by-group effects and time-by-problem-solving
process-by-group effects) was performed.
Relationship between epistemological beliefs and
The results showed that there was no statistical
online argumentation performance
effects for the time, F1,27 = 0.273, P = 0.606, the time
and group interaction, F2,27 = 0.7151, P = 0.573, the The students’ epistemological beliefs include five factor
time and problem-solving process, F3,88 = 0.808, P = scores from the EBI (simple knowledge, certain knowl-
0.503, or any three-way interaction between time, edge, omniscient authority, quick learning, and fixed
problem-solving process, and group, F3,88 = 0.147, ability). In order to identify relationships between
P = 0.943. This indicates that neither treatment had any students’ epistemological beliefs and their online
effect on the problem-solving process across the two argumentation performance, a multiple regression was
cases. conducted to predict the strength of the five factors in
The only significant main effect for the problem- epistemological beliefs that might affect students’
solving process, F2,74 = 24.654, P = 0.000, indicated a online argumentation performance. The five variables
statistically significant difference among the problem- were entered into the regression analysis using the step-
solving actions generated by the students. In the wise procedure. There was a significant linear relation-
follow-up analysis, a manova was employed to ship between simple knowledge and the score of online
compare the scores on the five problem-solving pro- argumentation performance, F1,28 = 6.234, P = 0.019.
cesses among the two groups at each case problem The multiple regression identified Simple Knowledge
solving. When the mean scores of the five problem- as the strongest predictor of the scores of online
solving processes were employed as dependent vari- argumentation performance. The multiple regression
ables to examine how the use of the constraint-based statistics indicated that the students’ tendency to believe
argumentation scaffolding affected students’ problem- simple knowledge accounted for 19% of the variance in
solving processes, the manova results were statistically the online argumentation performance.
in the performance of argumentation across the two solve problems quickly, rather than think over alterna-
discussions. Specifically, the treatments did not affect tive analyses and solutions (Schoenfeld 1983). Ill-
students’ performance in argumentation building over structured problem solving requires students to build
time. Follow-up tests showed that there were no signifi- problem space, generate multiple hypotheses, and
cant differences in the total scores of the argumentation represent personal arguments to justify their solutions.
performance between the first and the second However, students with a belief that solving problems
discussions. Perhaps the treatment period was too short are to give quickly one correct answer and to use correct
or the constraint-based argumentation tool was not used rules or principles might avoid multiple problem-
effectively. solving analyses and solutions. Further evidence came
The results of this study showed that there were dif- from a descriptive analysis of students’ epistemological
ferences in students’ problem-solving processes. Spe- beliefs. Students’ beliefs about simple knowledge were
cifically the scaffolded discussion group performed above average.
better than the control group in problem hypothesis In this study, epistemological beliefs predicted online
testing. While the scaffolded discussion group per- argumentation performance and frequency of evidence
formed significantly better than the threaded discussion generation during online discussions. Data from the
group in hypothesis testing, there were no significant scores on the online argumentation performance indi-
differences detected in the problem space construction, cated that simple knowledge had a positive relationship
hypothesis generation, and solution verification. This with the online argumentation performance. In addition,
investigation revealed a partial effect of the constraint- students with simple knowledge generated more evi-
based argumentation scaffold on hypothesis testing in dence to support their arguments during online discus-
solving ill-structured problems. Students in the scaf- sions. The more students believed that knowledge was
folded discussion group generated more alternative simple, the better the online argumentation perfor-
hypotheses and justified their arguments in the hypoth- mance. This finding indicates that students who view
eses than the other two groups. However, the difference knowledge as simple, requiring less time and effort to
was not sufficient to affect the total scores of problem- solve problems, are more likely to argue based on evi-
solving performance. In addition, following the treat- dence. This is inconsistent with earlier findings that epis-
ments, no significant difference in individual problem temological complexity and uncertainty were negatively
solving scores was found. Thus, based on the empirical related to avoiding arguments (Nussbaum & Bendixen
data, it is inconclusive whether the use of the constraint- 2002). There is the possibility that students with simple
based argumentation scaffolding could affect students’ knowledge belief see little point in argumentation and
performance in problem-solving processes. persist with their own argument, resulting in consistent
The current study provided equivocal support (P = arguments. However, more research is needed to address
0.056) for improvements in students’ performance in this inconsistent finding between students’ simple
solving ill-structured problems as a result of constraint- knowledge belief and their online argumentation types
based scaffolding, as did Cho and Jonassen (2002) and and performance. An individual who holds simple
Van Bruggen et al. (2003). Because the two treatment beliefs about the structure of knowledge seems more
groups received instruction from the same instructor likely to reduce most judgemental tasks to their most
regarding the classroom management problem solving, basic components, thereby eliminating the experience of
the effects of the constraint-based argumentation tool uncertainty (Nussbaum & Bendixen 2003).
on problem solving might not have been as strong. That So far, there has been no research on the relationship
is, through classroom discussions and activities, the of students’ simple knowledge belief in epistemological
students learned how to identify problems, generate beliefs and their online argumentation. Even if the result
hypotheses, and propose solutions for each classroom of this study did not manifest the strong relationship
management case study. Therefore, the students might between simple knowledge belief and online argumen-
have been familiar with the classroom management tation, the finding might suggest that online argumenta-
case studies, indicating that the effects of the constraint- tion patterns or performance could be related to
based argumentation scaffolding were not manifested. students’ epistemological beliefs. Further research on
Another possibility is that students have a tendency to online argumentation analysis might contribute to
investigate what role students’ epistemological beliefs problem-solving performance after the classroom man-
play in arguing through online discussions. agement case study. Individuals who believe in simple
Epistemological beliefs, as expected, had an effect on analysis and solutions to the ill-structured classroom
problem-solving performance. The results of this study management problems may be less inclined to explore
showed that omniscient authority was positively corre- more solution alternatives. The finding suggests that the
lated with hypothesis generation, hypothesis testing, stronger a student believes that knowledge is simple, the
and solution verification while students were involved lower the student performs on the ill-structured problem
in solving ill-structured classroom management solving.
problems. Also, belief in omniscient authority was the However, beliefs in fixed ability were positively
strongest predictor of the solution verification process. related to students’ problem-solving performance on
In particular, the more students believed that authorities the learning outcome measure. The more students
have access to knowledge that others do not, the better believed that ability to learn is innate and fixed, the
they did in solving classroom management problems. better they performed in solving problems. This finding
Unlike previous research, a belief in omniscient is also inconsistent with previous research that fixed
authority was associated with performance in solving ability is negatively related to problem solving (Schom-
problems. Current results contrast with previous mer 1990; Williams 2001). This finding did not repli-
research that has found a negative relationship between cate those of previous studies, the results of which had
a belief in omniscient authority and achievement in shown a negative relationship between the acceptance
problem solving. Previous studies on the relationship of fixed ability and problem-solving scores (Bendixen
between epistemological beliefs and problem solving et al. 1998; Williams 2001)
showed that students less likely to believe in omniscient Although some of the results were inconsistent with
authority performed better in terms of achievement in a previous research, the findings indicated that simple
hypermedia learning environment and text problem knowledge, omniscient authority, and fixed ability
solving (Schommer 1990; Bendixen & Hartley 2003). explained a significant proportion of variance over per-
One possible reason for the positive relationship formance on problem solving. This conclusion is con-
between students’ belief in omniscient authority and sistent with those of Schommer (1990), who found that
their problem solving is that students’ epistemological ill-defined problem solving was related to multiple
belief was not mature in terms of omniscient authority epistemological beliefs.
belief (M = 2.5). Even if the mean score of omniscient
belief was below the average score, students may have
Implications
expected the acceptance of their solution verification
from teachers. Given the time limitations, the belief Solving ill-structured problems is difficult for univer-
that teachers should approve problem solutions may sity students, in part because they require students to
have helped students in evaluating their own solutions justify or argue for their solutions and against others.
and generating more alternative solutions before they This study has supported other studies in suggesting
submitted their problem solving. It is also very likely that argumentation can be supported by using
that students still believe that all problems have right constraint-based discussion boards. However, even
answers that are determined by an omniscient when using those tools, students need to be coached on
authority. using alternative forms of discussion. In this study,
In the present study, simple knowledge and fixed rebuttals and elaborations, which are very important to
ability were revealed as predictors of individual argumentation supporting problem solving, were
problem-solving performance on the learning outcome underutilized. The nature of that coaching is an issue for
measure. Simple knowledge and fixed ability contrib- research.
uted significantly to the prediction of individual Another reason that argumentation and problems
problem-solving performance. Contrary to the positive solving are difficult, especially for American students,
relationship between simple knowledge and online are their epistemological dispositions towards simple
argumentation performance, there was a negative rela- knowledge and omniscient authority. Such dispositions,
tionship between simple knowledge and individual we fear, are artefacts of our educational system in which
teachers and professors attempt to convey information of this study add to a growing body of literature on the
in its simplest form and insist that students understand role of epistemological beliefs on argumentation and
their perspectives. For these students, argumentation problem solving (Schraw et al. 1995; Jonassen et al.
and ill-structured problems are especially difficult, 2002; Nussbaum & Bendixen 2002). Notable was the
because they require that students understand, generate, role of a belief in simple knowledge, which extends
and reply to multiple perspectives. Constraint-based prior research on solving ill-structured problems. The
discussion boards are likely to assist these students only current study has identified some key epistemological
if they are used consistently. beliefs that enter into important epistemological belief
studies, and has extended this work into technology-
Limitations enhanced learning environments.
Because of constraints in conducting classroom
research in the United States, this study exhibited a References
number of limitations. First, the sample size was rela- Andriessen J., Baker M. & Suthers D. (2003) Argumentation,
tively small. A larger sample would likely have demon- computer support, and the educational context of confront-
strated more obvious differences in performance. Also, ing cognitions. In Arguing to Learn: Confronting Cogni-
the effects of the constraint-based discussion board tions in Computer-Supported Collaborative Learning
were established by comparing the performance of stu- Environments (eds J. Andriessen, M. Baker & D. Suthers),
dents with those using a traditional threaded discussion pp. 1–25. Kluwer, Dordrecht, The Netherlands.
board. No control group comparisons were possible Baker M.J. (1999) Argumentation and constructive
because the control group did not participate in collabo- interaction. In Foundations of Argumentative Text Process-
rative activities. The contradictory nature of some of the ing (eds J. Andriessen & P. Coier), pp. 179–202. University
of Amsterdam Press, Amsterdam.
results is common in this kind of research. They may be
Bendixen L.D. & Hartley K. (2003) Successful learning with
due, in part, to the levels of motivation exhibited by the
Hypermedia: the role of epistemological beliefs and meta-
participants in the study. Finally, these results may not
cognitive awareness. Journal of Educational and Comput-
be generalizable beyond pre-service teachers. ing Research 28, 15–30.
Bendixen L.D. & Schraw G. (2001) Why do epistemological
Summary beliefs affect ill-defined problem solving? Paper presented
at the meeting of the American Educational Research Asso-
This study suggests that constraint-based discussion
ciation, Seattle, WA.
scaffolds can affect online argumentation types and
Bendixen L.D., Schraw G. & Dunkle M.E. (1998) Epistemic
performance when solving ill-structured diagnosis– beliefs and moral reasoning. Journal of Psychology 132,
solution problems. Argumentation scaffolding provided 187–200.
in the form of argument structure helps novice problem Boreham N.C. (1986) A model of efficiency in diagnostic
solvers to generate more evidence supporting their problem solving: implications for the education of
arguments. Computer-based scaffolding supports diagnosticians. Instructional Science 15, 191–211.
reflection of evidence-based explanation by prompting Bouwer A. (1998) ArgueTrack: the design of an argumenta-
students to negotiate or struggle with explaining the sig- tive dialogue interface. Paper presented at the 2nd Interna-
nificance of data (Zembal-Saul & Land 2002). Novice tional Workshop on Human-Computer Conversation,
problem solvers outperformed the other groups in Bellagio, Italy, 13–15, July 1998.
problem representation, justification, and monitoring Burnett R.E. (1993) Decision-making during the collaborative
planning of co-authors. In Hearing Ourselves Think: Cog-
and evaluation after being given guided question
nitive Research in the College Writing Classroom (eds A.
prompts (Ge & Land 2003). Therefore, the constraint-
Penrose & B. Sitko), pp. 125–146. Oxford University Press,
based argumentation scaffolding discussion board New York.
may be a useful way of stimulating reflective online Cho K.L. & Jonassen D.H. (2002) The effects of argumenta-
discussions. tion scaffolds on argumentation and problem solving. Edu-
Additionally, the study found that epistemological cational Technology Research and Development 50, 5–22.
beliefs were related to online argumentation perfor- Conklin J. (2003) Dialog mapping: reflections on an indus-
mance and problem-solving performance. The results trial strength case study. In Visualizing Argumentation:
Software Tools for Collaborative and Educational Sense- predicting argument-outcome links. Journal of Communi-
Making (eds P.A. Kirschner, S.J. Buckingham Shum, C.S. cation 50, 3–30.
Carr), pp. 117–136. Springer-Verlag, London. Nussbaum E.M. & Bendixen L.D. (2002) The effect of per-
Constantino-Gonzales M.A. & Suthers D. (2002) Coaching sonality, ability, & epistemological beliefs on students’
collaboration in a computer-mediated learning environ- argumentation behavior. Poster presented at the American
ment. In Proceedings of Computer Support for Collabora- Educational Research Association Annual Meeting, New
tive Learning 2002 (ed. G. Stahl), pp. 583–584. Lawrence Orleans, LA.
Erlbaum Associates, Hillsdale, NJ. Nussbaum E.M. & Bendixen L.D. (2003) How epistemology
Davis E.A. (1998) Scaffolding students’ reflection for science and personality affect students’ dispositions to argue. Paper
learning. Unpublished Doctoral Dissertation, University of presented at the annual meeting of the American Educa-
California, Berkeley, CA. tional Research Association, Chicago, IL.
Elstein A.S., Schulman L.S. & Sprafka S.A. (1978) Medical Nussbaum. E.M. & Sinatra G.M. (2002) On the opposite side:
Problem Solving: An Analysis of Clinical Reasoning. argument and conceptual engagement in physics. Paper
Harvard University Press, Cambridge, MA. presented at the meeting of the American Educational
Ge X. & Land S.M. (2003) Scaffolding students’ problem- Research Association, New Orleans, LA.
solving processes in an ill-structured task using question Nussbaum E.M., Hartley K., Sinatra G.M., Reynolds R.E. &
prompts and peer interactions. Educational Technology Bendixen L.D. (2002) Enhancing the quality of on-line
Research and Development 51, 21–38. discussions. Paper presented at the annual meeting of the
Jonassen D.H. (1996) Scaffolding diagnostic reasoning in American Educational Research Association, New Orleans,
case-based-learning environments. Journal of Computing LA.
in Higher Education 8, 48–68. Rittel H. (1984) Second generation design methods.
Jonassen D.H. (2000) Toward a design theory of problem Reprinted in Developments in Design Methodology (ed. N.
solving. Educational Technology: Research & Develop- Cross), pp. 317–327. John Wiley & Sons, New York.
ment 48, 63–85. Sanders J.A., Wiseman R.L. & Gass R.H. (1994) Does teach-
Jonassen D.H. & Remidez H. (2005) Mapping alternative dis- ing argumentation facilitate critical thinking? Communica-
course structures onto computer conferences. International tion Reports 7, 27–35.
Journal of Knowledge and Learning 1, 113–129. Scardamalia M. (2002) CSILE/Knowledge Forum®. In Edu-
Jonassen D.H., Marra R.M. & Palmer E. (2002) Epistemo- cation and Technology: An Encyclopedia, pp. 183–192.
logical development: an implicit entailment of constructiv- ABC-CLIO, Santa Barbara.
ist learning environments. In Curriculum, Plans, and Scardamalia M., Bereiter C. & Lamon M. (1994) The CSILE
Process of Instructional Design: International Perspec- project: trying to bring the classroom into world 3. In Class-
tives (eds N.M. Seel. & S. Dijkstra), pp. 75–88. Lawrence room Lessons: Integrating Cognitive Theory and Class-
Erlbaum, Mahwah, NJ. room Practice (ed. K. McGilly), pp. 201–228. MIT Press,
Kitchener K.S. (1983) Cognition, metacognition, and Cambridge, MA.
epistemic cognition. Human Development 26, 222– Schoenfeld A. (1983) Beyond the purely cognitive: belief
232. systems, social cognitions, and metacognitions as driving
Knudson R.E. (1991) Effects of instructional strategies, grade forces in intellectual performance. Cognitive Science 7,
and sex on students’ persuasive writing. Journal of Experi- 329–363.
mental Education 59, 141–152. Schommer M. (1990) Effects of beliefs about the nature of
Lombard M., Snyder-Duch J. & Bracken C.C. (2002) Content knowledge on comprehension. Journal of Educational Psy-
analysis in mass communication: assessment and reporting chology 82, 498–504.
of intercoder reliability. Human Communication Research Schommer M. & Dunnell P.A. (1997) Epistemological beliefs
28, 587–604. of gifted high school students. Roeper Review 19, 153–
Marttunen M. (1998) Electronic mail as a forum for argumen- 156.
tative interaction in higher education studies. Journal of Schommer-Aikins M., Duell O.K. & Hutter R. (2005) Episte-
Educational Computing Research 18, 397–405. mological beliefs, mathematical problem-solving beliefs,
de Mesquita P.B. (1992) Diagnostic problem solving of and academic performance of middle school students.
school psychologists: scientific method or guesswork? Elementary School Journal 105, 289–304.
Journal of School Psychology 30, 269–291. Schraw G., Dunkle M.E. & Bendixen L.D. (1995) Cognitive
Meyers R.A., Brashers D.E. & Haner J. (2000) Majority- processes in well-defined and ill-defined problem solving.
minority influence: identifying argumentative patterns and Applied Cognitive Psychology 9, 523–538.
Schraw G., Bendixen L.D. & Dunkle M.E. (2002) Develop- solving with argument visualization. In Visualizing Argu-
ment and validation of the Epistemic Belief Inventory mentation: Software Tools for Collaborative and Educa-
(EBI). In Personal Epistemology: The Psychology of tional Sense-Making (eds P.A. Kirschner, S. Buckingham
Beliefs about Knowledge and Knowing (eds B. Hofer & P. Shum & C.S. Carr), pp. 25–47. Springer-Verlag, London.
Pintrich), pp. 261–275. Erlbaum, Mahwah, NJ. Veerman A.L. & Treasure-Jones T. (1999) Software for
Suthers D. (1998) Representations for scaffolding collabora- problem solving through collaborative argumentation.
tive inquiry on ill-structured problems. Paper presented at In Foundations of Argumentative Text Processing (eds J.
the Annual Meeting of the American Educational Research Andriessen & P. Coirier), pp. 203–339. University Press,
Association, San Diego, CA. Amsterdam.
Suthers D. (2003) Representational guidance for collaborative Williams D. (2001) The impact of epistemological beliefs
inquiry. In Arguing to Learn: Confronting Cognitions in on middle school student’s knowledge acquisition and
Computer-Supported Collaborative Learning Environ- problem solving while working in a hypermedia-supported
ments (eds J. Andriessen, M. Baker & D. Suthers), problem-based learning environment. Paper presented at
pp. 27–46. Kluwer, Dordrecht, The Netherlands. the World Conference on Educational Multimedia, Hyper-
Tan S.C. (2000) Supporting collaborative problem solving media, & Educational Telecommunication, Tampere,
through computer-supported collaborative argumentation. Finland.
Unpublished Doctoral Dissertation. The Pennsylvania Zembal-Saul C. & Land S.M. (2002) Scaffolding the con-
State University, University Park, PA. struction of scientific arguments by prospective teachers
Toulmin S. (1958) The Uses of Argument. University Press, using inquiry-empowering technologies. Paper presented at
Cambridge, UK. the Annual Meeting of the American Educational Research
Van Bruggen J.M., Boshuizen H.P.A. & Kirschner P.A. Association, New Orleans, LA.
(2003) A cognitive framework for cooperative problem