21
RESEARCH ARTICLE Examining the effect of problem type in a synchronous computer-supported collaborative learning (CSCL) environment Manu Kapur Charles K. Kinzer Published online: 23 May 2007 Ó Association for Educational Communications and Technology 2007 Abstract This study investigated the effect of well- vs. ill-structured problem types on: (a) group interactional activity, (b) evolution of group participation inequities, (c) group discussion quality, and (d) group performance in a synchronous, computer-supported collaborative learning (CSCL) environment. Participants were 60 11th-grade science students working in three-member groups (triads) who were randomly assigned to solve a well- or an ill-structured problem scenario on Newtonian Kinematics. Although groups solving ill-structured problems generated more problem-centered interactional activity (a positive effect), they also exhibited participation patterns that were more inequitable (a negative effect) than groups solving well-structured problems. Interestingly, inequities in member participation patterns exhibited a high sensitivity to initial exchange and tended to get ‘‘locked-in’’ early in the discussion, ultimately lowering the quality of discussion and, in turn, the group performance. Findings and their implications for theory, method- ology, and scaffolding of CSCL groups are discussed. Keywords Collaborative problem solving Ill-structured problems Participation inequity Sensitivity to initial exchange Well-structured problems M. Kapur (&) National Institute of Education, Nanyang Technological University, 1 Nanyang Walk, Singapore 637616, Singapore e-mail: [email protected] C. K. Kinzer Teachers College, Columbia University, 525 W 120th Street, New York 10027, NY, USA e-mail: [email protected] 123 Education Tech Research Dev (2007) 55:439–459 DOI 10.1007/s11423-007-9045-6

Examining the effect of problem type in a synchronous computer-supported collaborative learning (CSCL) environment

Embed Size (px)

Citation preview

RESEARCH ARTICLE

Examining the effect of problem type in a synchronouscomputer-supported collaborative learning (CSCL)environment

Manu Kapur Æ Charles K. Kinzer

Published online: 23 May 2007� Association for Educational Communications and Technology 2007

Abstract This study investigated the effect of well- vs. ill-structured problem types on:

(a) group interactional activity, (b) evolution of group participation inequities, (c) group

discussion quality, and (d) group performance in a synchronous, computer-supported

collaborative learning (CSCL) environment. Participants were 60 11th-grade science

students working in three-member groups (triads) who were randomly assigned to solve a

well- or an ill-structured problem scenario on Newtonian Kinematics. Although groups

solving ill-structured problems generated more problem-centered interactional activity

(a positive effect), they also exhibited participation patterns that were more inequitable

(a negative effect) than groups solving well-structured problems. Interestingly, inequities

in member participation patterns exhibited a high sensitivity to initial exchange and tended

to get ‘‘locked-in’’ early in the discussion, ultimately lowering the quality of discussion

and, in turn, the group performance. Findings and their implications for theory, method-

ology, and scaffolding of CSCL groups are discussed.

Keywords Collaborative problem solving � Ill-structured problems �Participation inequity � Sensitivity to initial exchange � Well-structured problems

M. Kapur (&)National Institute of Education, Nanyang Technological University, 1 Nanyang Walk,Singapore 637616, Singaporee-mail: [email protected]

C. K. KinzerTeachers College, Columbia University, 525 W 120th Street, New York 10027, NY, USAe-mail: [email protected]

123

Education Tech Research Dev (2007) 55:439–459DOI 10.1007/s11423-007-9045-6

Introduction

Situative, socio-constructivist theories of learning emphasize the importance of having

learners engage in contextualized, authentic, ill-structured activities for any meaningful

learning to take place (Brown, Collins, & Duguid, 1989; Scardamalia & Bereiter, 2003).

However, typical problem solving taught in schools often tends to be well-defined and

well-structured with an overemphasis on the procedures required to solve the problem

(Hmelo-Silver, 2004; Jonassen, 2000). Unfortunately, students skilled in automatically

applying step-wise procedures are often not adequately prepared when they encounter

problems in which they need to transfer their learning to new domains, a skill required to

function effectively in society (Scardamalia & Bereiter, 2003).

Jonassen (2000) articulated a distinction between well- and ill-structured problems.

He argues that well-structured (WS) problems:

• present all elements of the problem to the learners,

• require the application of a limited number of regular and well-structured rules and

principles that are organized in predictive and prescriptive ways, and

• have knowable, comprehensible solutions where the relationship between decision

choices and all problem states is known or probabilistic.

On the other hand, ill-structured (IS) problems:

• possess problem elements that are unknown or not known with any degree of

confidence,

• possess multiple solutions, solution paths, or no solutions at all,

• possess multiple criteria for evaluating solutions, and

• often require learners to make judgments and express personal opinions or beliefs.

An extensive amount of research in face-to-face settings has empirically examined the

advantages of IS over WS problem solving.1 Such research generally favors IS problem

solving on both cognitive and affective measures (see Albanese & Mitchell, 1993; Gallagher,

Stepien, & Rosenthal, 1992; Hmelo, 1998). More importantly, the research in face-to-face

settings that has evaluated IS problem solving favorably over WS problem solving has in

turn resulted in an assumption of a transfer of these positive effects to computer-supported

collaborative learning (CSCL) settings. However, this assumption has not been clearly

supported by research (Jonassen & Kwon, 2001; Stahl, 2005; Suthers, 2006).

On the one hand, researchers theorize that group tasks need to be complex and

ill-structured to maximize group interaction (Cohen, 1994; Erkens, Andriessen, Peters,

2003). The argument here is simple: if the task is highly structured, it may not only dampen

the productive agency that drives collaboration (Schwartz, 1999) but also leave the group

without much to discuss, argue, and negotiate in the first place. On the other hand, recent

CSCL research suggests quite the opposite, supporting the use of clearer and greater

structure in collaborative tasks and problems (Schellens, Van Keer, Valcke, & De Wever,

2005; Weinberger, Stegman, & Fischer, 2005). The argument here is equally simple: the

demands (e.g., explaining, negotiating, elaborating, arguing, etc.) of collaboration already

place a high socio-cognitive load on the collaborators (Dillenbourg, 1999); making the

problem or task driving the collaboration highly ill-structured may add to the socio-

cognitive load even more, and, in turn, overwhelm learners (Kirschner, Sweller, & Clark,

1 Ill- and well-structured problem solving refer to groups solving ill- and well-structured problemsrespectively.

440 M. Kapur, C. K. Kinzer

123

2006). Finding a middle ground are researchers who keep the group task ill-structured, and

focus on scaffolding and supporting the collaborative process (e.g., Cho & Jonassen, 2002;

Ge & Land, 2003; Hmelo-Silver, 2004; Suthers & Hundhausen, 2003).

We believe this debate on the optimality of structuredness of group tasks and problems

to be necessary and healthy (Dillenbourg, 2002; Suthers, 2006). Yet, empirical accounts

specifically examining the effects of problem structuredness on group interactional

processes and outcomes remain surprisingly scarce (e.g., Kapur, 2006). Therein lies the

need for more research investigating the effects problem structuredness on collaborative

problem-solving interactions and outcomes. Once we come to understand what these

differences are, we may be better positioned to design process scaffolds and supports for

problem-solving CSCL groups (Palincsar & Brown, 1984).

This study was designed to address the foregoing need through a mixed-method analysis

(Suthers, 2006) of how differences between problem types (WS vs. IS) influence group

interactional activity, and how interactional activity, in turn, affects group performance

(Lam, 1997). Specifically, our research questions and methodology, detailed later, explore

(a) the effect of problem type on the amount and type of interactional activity in

small-group problem-solving tasks, (b) the effect of problem type on the level and

evolution of participation inequity by group members, and (c) how problem-centeredness

of interactional activity and participation inequity affect the quality of group discussion,

and, in turn, group performance.

Collaborative interactional activity

Because learners in collaborative groups interact and influence each other in the process of

problem solving, problem-solving interactions have been extensively used in investigating

productivity conditions of small, collaborative groups (e.g., Barron, 2003; Cohen, Lotan,

Abram, Scarloss, & Schultz, 2002; Jonassen & Kwon, 2001; Schellens et al., 2005; Spada,

Meier, Rummel, & Hauser, 2005; Suthers, 2006). Previous research supports the position

that greater problem-centered interactional activity results, on average, in better group

performance (Cohen et al., 2002; Jonassen & Kwon, 2001; Lam, 1997). Problem-centered

interactional activity typically involves defining the problem, identifying relevant parame-

ters, brainstorming solutions, evaluating and elaborating those suggestions, selecting solu-

tions, and negotiating a final decision. For example, Cohen et al. (2002) showed that

groups with high problem-centered interactional activity not only produced better group

performance but also registered higher individual achievement gains subsequently, whereas

off-task interactional activity had an adverse effect on group performance.

The strong effect problem-centered interactional activity is commonly attributed to the

socio-cognitive conflict in collaborative interactions (Erkens et al., 2003; Littleton &

Hakkinen, 1999). The idea being that not only are groups on average more productive than

individuals in generating ideas and insights but are also more effective in confronting

and discussing sub-optimal ideas and misconceptions. Thus, the relationship between

individual domain knowledge and socio-cognitive conflict is dialectical (Cohen, 1994):

socio-cognitive conflict not only arises from differences between individual knowledge

structures but also perpetuates the very collaborative interactions from which it arises. It

follows that the greater the intensity of problem-centered interactional activity, the higher

the likelihood of socio-cognitive conflict, and the greater the effort a group spends in

discussing, confronting and counter-suggesting, evaluating, elaborating, and analyzing

member contributions. Empirical research suggests that such effort, and thus the existence

Synchronous collabortive problem solving 441

123

of socio-cognitive conflict, is in fact a significant predictor of group performance (Howe,

Tolmie, Anderson, & MacKenzie, 1992; Light & Glachan, 1985). This evidence is con-

sistent with the strong theoretical support for the use of ill-structured problems.

At the same time, it is reasonable to argue that the purported positive peer-facilitation

effect on group performance through socio-cognitive conflict is contingent on some level

of equity in member participation (Cohen, 1994; Cohen et al., 2002; Schwartz, 1999). For

instance, one would expect this effect to vary with the level of participation inequity in a

group; the higher the inequity, the lower the effect on average. Inequitable member par-

ticipation implies that group performance is primarily influenced by one or two dominant

members. This activity leaves little opportunity for multiple perspectives, strategies, and

solutions to be shared and discussed, in turn, reducing opportunities for a group to benefit

from the peer facilitation effect due to socio-cognitive conflict. Further, previous research

indicates that individual participation is a strong predictor of individual learning gains;

lower participation results in lower achievement gain (Schellens et al., 2005). So even

though both theory and research advocate the use of ill-structured problems for maxi-

mizing problem-centered interactional activity, whether this interactional activity is

equitable across group members, and the effect of participation inequity on problem-

solving performance, remains relatively unexplored. In this study, we examined the effects

of interactional activity and participation inequity because both are critical components of

the underlying socio-cognitive conflict mechanism.

Based on the above rationale, our research questions were:

1. What is the effect of problem type (well- vs. ill-structured) on the amount and type of

interactional activity?

2. What is the effect of problem type on the level of participation inequity?

3. How do problem-centeredness of interactional activity and participation inequity

affect the quality of group discussion, and, in turn, group performance? Findings

resulting from these questions led us into an a posteriori, mixed-method analysis of

participation inequities in group discussions. Consequently, we added a fourth research

question to our study:

4. How do inequities in member participation evolve in groups solving well- and

ill-structured problems?

The constructs—amount and type of interactional activity, participation inequity, quality

of group discussion, and group performance—are operationalized in the next section.

Method

Participants and design

Sixty 11th-grade students (46 male, 14 female) from the science stream of a co-educa-

tional, English-medium high school in Ghaziabad, India participated in this study. Students

in the science stream typically study mathematics, physics, chemistry, and English as their

main academic subjects. The proportion of males to females in this sample is considered

normal for the science stream in the senior secondary years (11th and 12th grades) in India.

These participants were of average to above-average ability; their 10th grade All India

Central Board of Secondary Education (CBSE) mathematics and science scores were in the

70–95 percentile range. Participants typically came from middle to upper-middle class

families. The study was aligned with the school’s physics curriculum.

442 M. Kapur, C. K. Kinzer

123

A randomized experimental design was used: participants were randomly assigned to

three-member groups (or triads), which were then equally and randomly assigned to either

the well-structured (WS) or the ill-structured (IS) problem-type treatment. This resulted in

10 WS groups and 10 IS groups. A post-randomization check was carried out to confirm if

participants were homogeneously distributed across the two experimental conditions in

terms of academic ability in mathematics and science, as evidenced by their 10th-grade

CBSE mathematics and science scores. A MANOVA comparing these CBSE mathematics

and science scores between WS and IS groups did not reach significance, F(2,57) = .680,

p = .510, at the .05 level.

Procedure

The study was carried out in the school’s computer laboratory where a substantial amount

of curricular problem-solving activities takes place for these participants. The online

synchronous collaborative environment was a java-based, text-only chat application run-

ning on the Internet. These participants were technologically savvy in using online chat

though mainly in informal, social settings. Therefore, they were familiarized in the use of

the synchronous text-only chat application prior to the study because this was the first time

they had experienced it in a formal, instructional setting. Each group’s discussion and

solution were automatically archived as a text file to be used for analysis. Group members

could only interact within their group. A seating arrangement ensured that participants of a

given group were not proximally located so that the only means of communication between

group members was synchronous, text-only chat. The presence of teachers in the computer

laboratory also helped ensure that students communicated only through the online chat

environment.

To mitigate status effects (Cohen, 1994; Rosenholtz, 1985), we ensured that participants

were not cognizant of their group members’ identities; the chat environment was

configured so that each participant was identifiable only by an alpha-numeric code.

Cross-checking their interactions revealed that the participants followed the instruction not

to use their names and instead used the codes when referring to each other. No help

regarding the problem-solving task was given to any participant or group during the study.

Furthermore, no external member roles or division of labor were suggested to any of the

groups. The procedures described above were identical for both WS and IS groups. All

groups were given a sufficient amount of time; they were able to complete the tasks and the

time stamp in the chat environment indicated that all groups made full use of the allotted

time of 2 h. Also, we did not receive any requests for extra time from any of the groups.

Instrumentation

Problem scenarios

Consistent with Jonassen’s design theory typology for problems (2000) mentioned earlier,

a WS and an IS problem scenario were developed by the authors. Both problem scenarios

dealt with a car accident scenario that required participants to apply Laws of Newtonian

Kinematics and Friction to solve them (see Appendix). These laws govern the motion of an

object in a straight line under constant acceleration. The equations representing these laws

involve the following parameters: velocity, time, displacement, acceleration (including

Synchronous collabortive problem solving 443

123

gravitational acceleration), and coefficient of friction. In our design, we carefully noted

that any equations or formulas required in solving the problem scenarios did not require

anything more than what could not be typed using the keyboard.

The contrast between the WS and IS problem scenarios is central to the study and,

therefore, noteworthy. The WS problem scenario presented all the relevant parameters

explicitly (e.g., size of the dent, range of values of the collision force, etc.). In contrast, the IS

problem scenario contained many more parameters; the relevance of these parameters

varying from low (e.g., driver’s age, weight, etc.) to high (e.g., size of dent, magnitude of the

collision force, etc.). Further, the parameters in the IS problem scenario could not be ascer-

tained to a high degree of certainty (e.g., the presence of skid marks make the magnitude of the

collision force difficult to ascertain to a high degree of certainty), in turn, requiring partici-

pants to rely on deductions, assumptions, opinions, or beliefs. Consequently, relative to the

WS problem scenario, the IS problem scenario admitted many more solutions and solution

paths as well as criteria for evaluating those solutions and solution paths.

Content validation of the two problem scenarios was achieved with the help of two

physics teachers from the school with experience in teaching those subjects at the senior

secondary levels. The teachers also assessed the time students might need to solve the

problems. Feedback from the teachers resulted in minor modifications to the problem

scenarios. Problem classification validation was then undertaken by having the top three

tenth-grade science students and the two teachers classify the two problems. Instead of

providing teachers and students with the top-down typology (Jonassen, 2000) we used to

design the WS and IS problems, we presented them with the two problems and asked them

articulate in writing if and how the two were different. From their responses (e.g.,

‘‘problem 1 has fewer unknowns,’’ ‘‘problem 2 is more complex and can be solved in manyways,’’ etc.; problems 1 and 2 refer to the WS and IS problems respectively), we could

infer in a bottom-up manner that their classifications were consistent with ours. The same

three students were also asked to solve the problems to confirm that the 2 h were sufficient

time for the task. All of them completed the problems and submitted their work in about

1 h. However, to avoid insufficient time for group work, we decided to give each group 2 h

to allow adequate time for group interaction and discussion.

Group interactional activity

The raw data unit within each problem-solving discussion was an utterance. In online text-

chat, an utterance includes everything a participant types before hitting the ‘‘Enter’’ button

on the keyboard. Hence, each problem-solving discussion is a sequence of utterances.

Quantitative Content Analysis (QCA) (Chi, 1997) was then used to segment and code each

utterance into one or more interaction units. The unit of analysis was semantically defined

as the function(s) that an intentional utterance served in the problem-solving process.

Bransford and Nitsch (1978) support the case for semantically-defined units by viewing

meaning-making and understanding as functions of the interdependence between interac-

tion and context. They argued that to fully comprehend a given interaction, one must not

only understand its words and the sentences (syntactic features), but also how it is situated

in a discussion context.

Thus, every utterance was segmented into one or more interaction unit(s), and coded

into categories adapted from the Functional Category System (FCS)—an interaction

coding scheme developed by Poole and Holmes (1995) specifically for the purpose of

studying small-group collaborative interactions in problem-solving contexts. The interac-

tion behaviors that FCS categories represent have been systematically identified, content

444 M. Kapur, C. K. Kinzer

123

validated, and tested in several research studies (e.g., Jonassen & Kwon, 2001; Kapur,

2006; Poole & Holmes, 1995), in turn, adding to the validity and reliability of using the

FCS (Rourke & Anderson, 2004). Accordingly, each interaction unit was coded into one of

seven categories:

1. Problem Analysis (PA): Statements that define or state the causes behind a problem

(e.g., ‘‘I think the man was driving too fast’’),2. Problem Critique (PC): Statements that evaluate problem analysis statements (e.g.,

‘‘how can you be sure that the man was driving fast’’),3. Orientation (OO): Statements that attempt to orient or guide the group’s process,

including simple repetitions of others’ statements or clarifications; Statements that

reflect on or evaluate the group’s process or progress (e.g., ‘‘let’s take turns giving ouropinions’’),

4. Criteria Development (CD): Statements that concern criteria for decision making or

general parameters for solutions (e.g., ‘‘we need to find the initial speed of the car’’),

5. Solution Development (SD): Suggestions of alternatives, ideas, proposals for solving

the problem; Statements that provide detail or elaborate on a previously stated

alternative. They are neutral in character and provide ideas or further information

about alternatives (e.g., ‘‘use the 2nd equation of motion’’),

6. Solution Evaluation (SE): Statements that evaluate alternatives and give reasons,

explicit or implicit, for the evaluations; this category also included statements that

simply agreed or disagreed with criteria development or solution suggestion

statements; Statements that state the decision in its final form or ask for final group

confirmation of the decision. (e.g., ‘‘yes, but how do we get acceleration?’’), or

7. Non-Task (NT): Statements that do not have anything to do with the decision task.

They include off-topic jokes and tangents (e.g., ‘‘let’s take a break!’’).

The frequency of interaction units in the seven FCS categories formed the seven measures

of group interactional activity (Jonassen & Kwon, 2001). Furthermore, the sum of

frequencies of interaction units in PA, PC, CD, SD, and SE formed a measure of problem-

centered activity (PCA) (Cohen et al., 2002).

Group discussion quality (DQ)

Employing a teleological approach consistent with the conception of problem solving as a

goal-directed activity, quality of an interaction unit was rated as high, low, or neutral by two

independent raters (inter-rater reliability reported in the Results section). Quality of an

interaction unit was rated as high (assigned an impact of +1) if its impact on the group progress

towards the goal of solving the problem was assessed to be positive. Similarly, quality was

rated as low (assigned an impact of�1) if the opposite were true. Interaction units that were

assessed to have maintained the status quo were rated neutral, i.e., assigned an impact of 0

(Kapur, Voiklis, Kinzer, & Black, 2006). The measure of the overall quality of a group’s

discussion was defined as the difference in the numbers of high and low quality interaction

units as an impact-weighted proportion of the total number of interaction units produced by

the group. For example, if there were 38 interactional units of which 23 had an impact of + 1,

15 had an impact of �1, and 5 had an impact of 0, then the discussion quality was equal to

23� 1ð Þ þ 15� �1ð Þ þ 5� 0ð Þ23þ 15þ 5

¼ 8

43¼ :19

Synchronous collabortive problem solving 445

123

Note that this measure is bounded between �1 and 1, and simulation studies (e.g., Kapur

et al., 2006) indicate that it is arguably a better measure of discussion quality than the

commonly used simple proportion of high-quality interaction units, because it adjusts for

the proportion of low quality interaction units as well.

Group participation inequity (PI)

First, participation proportion for every group member was calculated by dividing the total

number of intentional interactions by the member over the total number of interactions by

the whole group. Then, the standard deviation (SD) of the resulting three participation

proportions within each group—a measure of dispersion—was taken as a measure of the

group’s participation inequity. For example, the SD of the participation proportions .4, .3,

and .3 equals .06. Thus, a low SD implies closely clustered participation proportions within

a group—an equitable participation pattern. In contrast, the SD of participation proportions

.8, .15, and .05 equals .41. Thus, a high SD implied a discussion dominated by one or two

members within the group—an inequitable participation pattern.

Group performance (GP)

Group performance was operationalized as the quality of solution produced by the group.

Because there was no right or wrong answer to the problem scenarios, the strategy adopted

was to focus on the extent to which groups were able to support their decisions through a

synthesis of both qualitative and quantitative arguments, and ground them in justifiable

assumptions. Given this focus, a holistic rubric (e.g., Lee, Chan, & Van Aalst, 2006) for

rating group solutions was developed (see Table 1) by the authors in consultation with the

students’ physics teachers. The quality of group solutions was then rated by two inde-

pendent raters (inter-rater reliability reported in the Results section) on a scale from 0 to 4

points in units of 0.5 using the rubric shown in Table 1.

Covariates

Two covariates were used in this study. The first was a measure of a group’s prior

knowledge, operationalized as the mean score of the group members on the standardized

Table 1 Rubric for Coding Quality of Group Solution

Quality Description

0 Solution is weakly supported, if at all

1 Solution supported in a limited way relying either on a purely quantitative or a qualitativeargument with little, if any, discussion and justification of the assumptions made

2 Solution is only partially supported by a mix of both qualitative and quantitative arguments with;assumptions made are not mentioned, adequately discussed, or justified to support the decision

3 Solution synthesizes both qualitative and quantitative arguments; assumptions made are notadequately discussed and justified to support the decision

4 Solution synthesizes both qualitative and quantitative arguments; assumptions made are adequatelydiscussed and justified to support the decision

Mid-point scores of .5, 1.5, 2.5, and 3.5 were assigned when the quality of solution was assessed to bebetween the major units 0, 1, 2, 3, and 4

446 M. Kapur, C. K. Kinzer

123

CBSE mathematics and science test (e.g., Cohen et al., 2002). The second was a measure

of prior knowledge inequity within a group. The SD of the three individual scores on the

CBSE test formed a measure of prior knowledge inequity within a group; the higher the

SD, the greater the prior knowledge inequity (or heterogeneity) in a group.

Results

This section is organized into five subsections. The first section reports the inter-rater

reliabilities for this study. The four sections thereafter describe the analytical procedures

and report the results for each of the four research questions respectively. SPSS 14.0 was

used for the various statistical procedures and results reported throughout this section.

Inter-rater reliabilities

The lead researcher (first author) and two research assistants were first trained in the use of

the FCS and then independently segmented and coded the problem-solving interactions.

The resulting inter-rater reliabilities—Krippendorff’s alphas—were .87, .85, and .82.

Similarly, the resulting Krippendorff’s alphas for rating group discussion quality and group

solutions were .92 and .95, respectively. All differences were reconciled through discus-

sions between the coders and the researcher.

Effect of problem type on interactional activity

To answer the first research question, a MANCOVA was used to investigate differences

between WS and IS groups in their interactional activity in the seven FCS categories, with

group prior knowledge as a covariate (Stevens, 2002). A significant multivariate effect

was found between WS and IS groups, F(7,11) = 2.969, p = .042; partial g2 = .652,

power = .74. Group prior knowledge did not have any significant effect on group inter-

actional activity, F(7,11) = .920, p = .526. Box’s test for violations against homosce-

dasticity was not significant, p = .147. Table 2 presents the mean frequencies and standard

deviations of interactional activity in each FCS category, for each problem type.

Follow-up univariate analysis (univariate Fs) for each dependent variable indicated that

IS groups had significantly greater amounts of interactional activity than WS groups in five

of the seven FCS categories:

(i) PA: problem analysis, F(1,17) = 5.218, p = .035, partial g2 = .24,

(ii) OO: orientation, F(1,17) = 5.258, p = .035, partial g2 = .24,

(iii) CD: criteria development, F(1,17) = 11.027, p = .004, partial g2 = .39,

(iv) SD: solution development, F(1,17) = 10.550, p = .005, partial g2 = .38, and

(v) SE: solution evaluation, F(1,17) = 8.974, p = .008, partial g2 = .35.

2 Note that partial g2 is an estimate of the effect size. As a rule of thumb, partial g2 = .01 is consideredsmall, .06 medium, and .14 large (Cohen, 1977). Further note that a strong multivariate effect can have alarge effect size.

Synchronous collabortive problem solving 447

123

Effect of problem type on participation inequity

To answer the second research question, two ANCOVAs were carried out: one at the group

level and another at the individual level. To control for Type I error inflation, a Bonferroni-

corrected alpha level of .05/2 = .025 was used.

At the group level, an ANCOVA was carried out to investigate the effect of problem

type on group participation inequity (PI), with group prior knowledge and group prior

knowledge inequity as the covariates. The effect was significant, F(1,16) = 8.479,

p = .010, partial g2 = .35, power = .67. Results suggested that IS groups (M = .136,

SD = .034) exhibited significantly greater PI than WS groups (M = .075, SD = .051).

Neither group prior knowledge, F(1,16) = .276, p = .607, nor group prior knowledge

inequity, F(1,16) = 1.689, p = .212, had a significant effect on PI. Levene’s test of

homogeneity of error variances was not significant, F(1,18) = .757, p = .396.

At the individual level, an ANCOVA was carried out to rule out gender and individual

prior knowledge as predictors of an individual’s participation proportion during group

discussions. Gender was the between-subjects factor and individual prior knowledge

formed the covariate. Neither gender, F(1,57) = .022, p = .884, nor individual prior

knowledge, F(1,57) = .777, p = .382, had any significant effect on individual participation

proportion. Levene’s test of homogeneity of error variances was not significant, F(1,58) =

1.449, p = .234.

Thus, the group- and individual-level analysis of participation inequity, taken together,

revealed a potentially negative effect of IS problem solving: member participation in IS

groups was significantly more inequitable than in WS groups, and that such inequity was

not a function of pre-existing factors such as group prior knowledge, group prior knowl-

edge inequity, individual prior knowledge, and gender.

Explaining group performance: a statistical path model

To answer the third research question, statistical path analysis (e.g., Cohen et al., 2002)

was used. Path analysis uses a set of linear regression models to relate independent vari-

ables with the intermediary and dependent variables (Kline, 1998). Standardized beta

coefficients in the regression models form the path coefficients (Kline, 1998). In this study,

path analysis was used to model how problem type (IS coded 1, WS coded 0) influences

Table 2 Mean Frequencies and Standard Deviations of the Functional Content of Group InteractionalActivity across Well- and Ill-Structured Problem Types

Functional Category Well-Structured Ill-Structured

M SD M SD

PA: Problem Analysis 4.70 3.65 13.40 9.96

PC: Problem Critique 3.10 2.08 6.00 5.93

OO: Orientation 32.80 15.22 58.90 37.36

CD: Criteria Development 10.40 5.64 18.70 6.65

SD: Solution Development 25.00 12.74 48.70 19.44

SE: Solution Evaluation 45.50 17.79 89.80 46.48

NT: Non-Task 3.50 4.33 2.10 2.77

There were n = 10 groups each in the well- and ill-structured problem types

448 M. Kapur, C. K. Kinzer

123

problem-centered activity (PCA) and PI; how PCA and PI in turn affect discussion quality

(DQ); and, finally, how DQ influences group performance (GP). Figure 1 presents the path

model. With the exception of the effect of PCA on DQ, all other paths were significant at

the .05 level or lower. Model validation and sensitivity analysis did not reveal substantive

violations of the assumptions of the regression models.

The path coefficients for the model shown in Fig. 1 were obtained in three steps:

(i) The Pearson’s correlation between problem type (IS coded 1, WS coded 0) and PCA,

r(20) = .634, p = .003, formed the path coefficient between the two.3 Similarly, the

path coefficient between problem type and PI was the Pearson’s correlation between

the two, r(20) = .601, p = .002. These path coefficients merely confirmed our findings

from the first two research questions; i.e., even though interactional activity in IS

groups was more problem-centered, there was also greater participation inequity in

their discussions.

(ii) Multiple linear regression was then used to obtain path coefficients for the influence of

PI and PCA on DQ.4 The multiple regression model for DQ was significant, R2 = .348,

F(2, 17) = 4.535, p = .026. PI was the only significant predictor, t = �2.396, p = .028; b= �.482. The negative path coefficient, b = �.482, indicated that higher PI resulted in

lower quality of group discussion. This effect held true for WS and IS groups separately

as well, i.e., the effect of PI on DQ was not spurious. The effect of PCA was not

significant, t = �1.236, p = .233. This analysis essentially revealed that despite high

PCA in IS groups, it was the negative effect of problem type on PI that lowered the

quality of their group discussions. The DQ of IS groups was, on average, of a lower

quality (M = .416, SD = .127) than that of WS groups (M = .633, SD = .189).

(iii) Finally, as in step (i), the path coefficient for the effect of DQ on group performance

(GP) was the Pearson’s correlation between the two, r(20) = .809, p < .001. The

positive and significant path coefficient of .809 indicates that the higher the quality

of discussion, the better the group performance. This effect held true for WS and IS

groups separately as well, i.e., the effect of DQ on GP was not spurious. Continuing

the argument from earlier, higher PI resulted in lower DQ and lower DQ, in turn,

resulted in lower GP. Thus, IS groups produced, on average, solutions of a lower

quality (M = 1.000, SD = .745) than those produced by WS groups (M = 2.250,

SD = 1.317).

* p < .05 ** p < .01 *** p < .001

PCA

Problem Type

PI

GP

.601 ** -.482*

.634 **

DQ.809***

Fig. 1 A statistical path model of synchronous collaborative problem solving: Influence of problem type(ill-structured coded 1, well-structured coded 0) on problem-centered activity (PCA) and participationinequity (PI); influence of PCA and PI on group discussion quality (DQ); and finally, influence of DQ ongroup performance (GP)

3 Note that one could also regress PCA on problem type to get b (standardized regression coefficient) as thepath coefficient. This is because the value of b in simple linear regression models with only one predictorequals the Pearson’s correlation between the predictor and the dependent variable (Kline, 1998).4 Note that the pair-wise correlation between PCA and PI was low and statistically not significant, r(20) =.226, p = .169, suggesting that multicollinearity among the two predictors was not an issue.

Synchronous collabortive problem solving 449

123

Explaining the evolution of participation inequity

Explaining the evolution of PI in group discussions formed the thrust of the fourth research

question. Before proceeding further, recall the results reported earlier (for research ques-

tion 2) that PI was not a function of pre-existing factors such as group prior knowledge,

group prior knowledge inequity, individual prior knowledge, and gender. Further recall

that our research design mitigated status effects because participants were not cognizant of

their group members’ identities. Therefore, in the absence of any externally imposed

division of labor or member roles, these findings suggest a natural evolution of partici-

pation patterns that tends to be inequitable in IS groups in synchronous environments – an

unexpected adverse effect of IS problem solving.

So we delved deeper into examining the evolution of participation inequity in WS vs. IS

groups. We started by calculating PIs after each utterance in a discussion, giving the level

of PI in the discussion up to any given utterance. For example, if participation proportions

(defined earlier) up to and including the 10th utterance were .6, .3, and .1, then PI up to that

utterance equaled .25 (the SD of the participation proportions). Thus, calculating the PIs

after each utterance resulted in a temporal sequence of PI values for each group discussion.

Plotting the sequence of PI values over time (defined notionally; each utterance, not the

actual time, being a tick on the evolutionary clock) for the 20 groups allowed us to

compare the evolution of PI between WS and IS groups. Using Fig. 2, we present an

analysis of the salient patterns by contrasting the typical trajectories of WS with IS groups.

What was surprising was how sensitive the evolution of these trajectories was to the

initial exchange between group members in both WS and IS groups. This can be seen in the

sharp fluctuations in the PI trajectories in Fig. 2 before they quickly settled into an inequity

plateau, i.e., the trajectories tended to flatten out after the initial fluctuations. A part of these

sharp fluctuations is to be expected by design especially in the first 3–5 utterances. What we

are referring to is the observation that these fluctuations were persistent in the initial

exchange even after everyone had started contributing to the discussion. Also, recall that any

point on a group’s trajectory corresponds to the level of PI in the group’s discussion up to

that point. Figure 2 suggests that once the trajectories settled into inequity plateaus (and this

happened fairly quickly), groups tended to maintain the corresponding PI level for the rest

0

0.1

0.2

0.3

0.4

0.5

0.6

Notional Time

Par

tici

pat

ion

Ineq

uit

y

Well Structured

Ill Structured

Fig. 2 Evolution of participationinequities across problem types

450 M. Kapur, C. K. Kinzer

123

of their discussion. The main difference between WS and IS groups seemed to be that the

former typically settled into a lower plateau (i.e., higher equity) whereas the latter into a

higher plateau (i.e., lower equity). The evolution of PI was not a gradual process but one that

was highly sensitive to initial exchange. Critically, this sensitivity to initial exchange

seemed to get ‘‘locked-in’’ for the rest of the discussion, i.e., once settled into an inequity

plateau early in the discussion, groups were not able to get out of it.

Results of the path analysis presented earlier indicated that this final level of PI sig-

nificantly predicted discussion quality and, in turn, group performance. This finding makes

the early lock-in of PI a significant finding because it suggests that the impact of initial

exchange on group performance is much greater than what comes later; the seeds of

eventual group performance seem to be sown fairly early in a group’s discussion. Thus,

we next focused our attention on the nature of these initial exchanges across the WS and

IS groups.

A qualitative analysis of initial exchanges

A qualitative examination of the initial exchange across the 20 groups suggested that

groups (not surprisingly) started by attempting to organize the discussion or propose ideas

for solving the problem. But what was surprising was how differently this exchange

evolved in WS and IS groups. In IS groups, the first idea put forth tended to be taken up

with little debate on its merits. As the discussion evolved, this framework gained

momentum and was difficult to break down. Invariably, the group member who proposed

the idea ended up dominating the discussion. This can be seen from the following initial

exchange in an IS group (I01, I02, I03 are the three members of the group):

1. I01 > hello everyone!

2. I02 > let’s start

3. I01 > ok I feel its a clear case of sheer negligence on Mr Rahul’s part

4. I01 > he must be driving pretty fast

5. I03 > I don’t understand!

6. I01 > physician says it was a considerable impact ranging b/w 20 g to 25 g

7. I03 > yes, me too! what are we supposed to do?

8. I02 > car has been severely struck

9. I03 > he was not able to control the car I think

10. I01 > Yes this proves that the impact was pretty hard and he was driving fast

11. I01 > we have to submit a report with our analysis and recommendations

12. I01 > we can put it like this that it was a blind turn, Mr Rahul was driving his car

around 40–45 km/h.....neither of them saw each other

13. I03 > sounds ok to me

14. I02 > sure, work out the details now ...

In the above excerpt, I01 started by stating his claim (utterance 3) and goes immediately

into supporting it (utterance 4, 6, and so on). Meanwhile, utterances 5 and 7 suggest that

I02 and I03 were struggling with the ambiguity and ill-structuredness of the problem

scenario and had little idea of how to make sense of the problem or what was expected of

them. In this ambiguity, the idea offered by I01 seemed to offer some sense of a direction

for the group. I01 continued with supporting his claim, which was eventually accepted. The

subsequent discussion (not shown here) continued with I01 taking the lead as the dominant

Synchronous collabortive problem solving 451

123

member. Also note how inequitable this early exchange is with I01 dominating the dis-

cussion, and once this early exchange was accepted as a frame for the rest of the

discussion, these participation patterns tended to get locked in. Consequently, the resulting

discussion showed that I01’s arguments were not sufficiently challenged and, in turn, the

problem and solution spaces were not explored in a manner demanded by the problem.

This is an important point. In an ill-structured situation, the problem and solution spaces

are highly complex and ambiguous without there being any clear right or a wrong frame-

work for getting to a solution. Thus, it is difficult to know in the initial stages if an idea is

going to be productive or not. Therefore, it is less important what the idea is or for that matter

the prior knowledge and ability of the member proposing it; what is critically important is

that ideas are put forth and engaged sufficiently with participation from all members.

Examining the data further, we found out that even when the most able (in terms of

prior knowledge) member was the proposer and ended up dominating the subsequent

discussion, it was not a guarantee that that would lead to a productive group outcome.

In fact, only 20% of the time (i.e., in 2 out of 10 groups), the dominant person was the most

able in terms of prior knowledge in mathematics and science; 20% of the time this person

was the least able; 50% of the time this person was the one in the middle, and finally 10%

of the time the discussion was jointly dominated by the middle and lowest-ability mem-

bers. Interestingly, among the IS groups, the higher-performing groups were dominated by

a person in the middle range of prior knowledge.

Contrastingly, in WS groups, ideas were more likely to be subjected to closer scrutiny,

which resulted in a more equitable participation pattern. This can be seen from the fol-

lowing excerpt from a WS group (W01, W02, W03 are the three members of the group):

1. W03 > hi!

2. W02 > hello

3. W01 > hi what do u think ... should we fight the case in court?

4. W02 > wait let me analyze

5. W03 > Yes! I would fight the fine in the court with a reason that the boy would have

come under my car if I would have not applied the brakes.

6. W02 > I don’t think that’s good enough

7. W01 > I agree

8. W01 > I think we should calculate the speed of the car.

9. W02 > disp = 15 m

10. W01 > I have made some calculations and found that driver’s speed should not

exceed 48 km/h acc. to given data

11. W03 > can u explain it?

12. W02 > I think the policeman did not measure his speed correctly and we can

definitely fight fine in the court

13. W03 > but I think that we will have to explain the logic that we have used.

14. W01 > yes yes

In this excerpt, W01 seeks others’ opinions (utterance 3) and was asked to wait by W02.

Meanwhile, W03 makes the claim that the fine ought to be fought (in utterance 5) and

gives a reason for it (also in utterance 5). The idea is immediately rejected by W02 and

W01 for not having sufficient weight (utterances 6 and 7 respectively). W01 proposes an

alternative idea of calculating the speed of the car (utterance 8). W02 builds on it by

proposing a value for the displacement of the car. W01 responds after having made some

calculations. Again, instead of accepting it, W03 asks W01 to explain his work. Finally,

452 M. Kapur, C. K. Kinzer

123

W02 introduces another idea (that the policeman did not measure the speed correctly; in

utterance 12) to support his claim for fighting the fine. He was asked to explain the logic

underlying his reasoning by W03 (utterance 13), something that resonated with W01 as

well (utterance 14).

Several features of this exchange in the WS group contrast with those of the IS group

presented earlier. First, participation in the initial exchange was more equitable in the WS

group. Second, the proposed ideas were less likely to be taken up on their face value in the

WS group. Third, there was little, if any, confusion expressed by the WS group members in

terms of what the problem demanded; the problem and solution spaces were less ambig-

uous. This made it relatively easier to evaluate the efficacy and viability of ideas proposed

at the outset.

These contrasts underscore the point made earlier. As all members tended to propose

and counter-argue ideas, the ability of the members mattered little. What was more

important was that ideas were proposed and subjected to argumentation by all members of

the group, ensuring a more equitable participation pattern. Once this pattern was estab-

lished in the initial exchange, it tended to get locked in for the rest of the discussion. For

WS groups, this was a positive event because they settled into lower inequity plateaus

ensuring participation from all members. However, as our analyses show, this was not the

case for IS groups.

Discussion

This study affirms the importance of problem type as a key variable influencing

synchronous, collaborative problem solving. The main findings of this study are:

(i) Consistent with theory and previous research, IS groups engaged in significantly

greater problem-centered interactional activity;

(ii) However, participation in this activity was significantly more inequitable in IS

groups, i.e., one or two members tended to dominate the discussions;

(iii) The evolution of participation patterns revealed a high sensitivity to initial exchange

that locked groups into their final inequity levels early on in their discussions; and

(iv) These final participation inequity levels (which were locked-in early in the

discussion) adversely affected discussion quality, and, in turn, group performance.

The finding that IS groups engaged in significantly greater problem-centered interactional

activity is, in many ways, a positive effect of solving IS problems but one that is not surprising.

The lower levels of problem-centered activity in WS groups can be explained by the fact that

WS problems, by design, provide a clearer problem definition, are predictive and prescriptive

with regard to possible solution paths, and therefore may not require elaborate group dis-

cussions on problem definition and solution development (Shin, Jonassen, & McGee, 2003).

On the other hand, IS problems do not clearly define the problem task and thus allow for non-

routine and multiple solution paths. This, in turn, may be the driving force for more open

exchanges and elaborated discussion in the problem-centered functional categories (Hmelo,

1998; Hmelo-Silver, 2004). It is also consistent with situative, socio-constructivist theories

and models for instructional practices that posit the use of IS problems for maximizing

problem-centered interaction (Spiro & Jehng, 1990; Suthers, 2006).

The second finding that despite the greater problem-centered interactional activity, IS

groups also exhibited significantly greater participation inequity in their discussions, i.e.,

member participation in IS groups tended to be dominated by one or two members.

Synchronous collabortive problem solving 453

123

Inequitable member participation implies that group performance is primarily influenced

by one or two dominant members. This leaves little opportunity for multiple perspectives,

strategies, and solutions to be shared and discussed, in turn, reducing opportunities for a

group to benefit from the peer facilitation effect due to socio-cognitive conflict. Path

analysis further demonstrated that high participation inequity resulted in low quality of

discussion leading, in turn, to poor group performance. In groups where participation was

more equitable, the quality of discussion tended to be better, possibly benefiting maximally

from the multiple perspectives and inputs of all members in the group instead of just one or

two. Thus, the effect of IS problems on participation inequity in group discussion is a

potentially negative effect that remains largely unexplored in CSCL research.

Still, the most important finding we believe emerges from the sensitivity to early exchange

that tends to lock-in participation levels; a lock-in that affects the quality of the discussion,

and, eventually determines how successful a group is in solving the problem. Because a lock-

in of participation levels also implies a lock-in to the dominant members’ proposals, this

sensitivity only underscores the quality of such proposals made early in the discussion,

especially in an ill-structured problem space. This is not to say that the quality of proposals

that come later in a discussion are not important. What the data suggest is that once a

discussion gets locked in, it gathers inertia and it becomes increasingly difficult for proposals

to make an impact proportional to their quality. High-quality proposals made earlier in a

group discussion, on average, do more good than if they were made later. Likewise, low-

quality ones, on average, do more harm if they come earlier than later in the discussion (Kapur

et al., 2006). This sensitivity to initial exchange is particularly significant given that the

groups were not scaffolded in their problem-solving process. Thus, scaffolding CSCL groups

in their early exchange may in fact be more important than the entire process.

We believe that this implication is an important contribution to the growing knowledge

about scaffolding CSCL groups (e.g., Dillenbourg, 2002; Fischer & Mandl, 2005; Suthers

& Hundhausen, 2003). For example, instead of scaffolding the entire process of collabo-

ration through role assignments (e.g., Schellens et al., 2005), it may be more critical to

scaffold participation in the early exchange. Likewise, instead of scaffolding the entire

process of problem solving using process scaffolds, it may only be necessary to scaffold

how a group analyzes and frames the problem and then fade the scaffolds (Palincsar &

Brown, 1984). These are testable hypotheses that emerge from this study and we invite the

field to test and extend this line of inquiry (Kapur & Kinzer, 2007).

A methodological implication from this finding is that CSCL research needs to pay

particular attention to the temporal aspects of interactional dynamics (Stahl, 2005). As this

study demonstrates, studying the evolution of interactional patterns can be insightful,

presenting counterintuitive departures from assumptions of linearity in the problem-solving

process. Interestingly, sensitivity to initial exchange exhibited by CSCL groups in our

study seems analogous to sensitivity to initial conditions exhibited by many complex

systems (Bar-Yam, 2003); the idea being that small changes initially can lead to vastly

different outcomes over time, which is what we found in our study. Furthermore, the

locking-in mechanism is analogous to attractors in the phase space of complex systems

(Bar-Yam, 2003). In fact, a realization of the inherent complexity of group interactional

dynamics is giving way to a more temporal and emergent view of how groups function and

perform (Arrow, McGrath, & Berdahl, 2000; Stahl, 2005). A complexity-grounded,

emergent view also presents a unique challenge to traditional analytical measures and

methods for analyzing the processes of collaborative problem-solving groups, and in turn,

how these processes relate to the success (or failure) of groups. Yet, no reliable analysis of

this relationship has been undertaken, in part due to a lack of methods and measures for

454 M. Kapur, C. K. Kinzer

123

characterizing the evolutionary interactional dynamics of collaborative groups (Barab,

Hay, & Yamagata-Lynch, 2001; Kapur et al., 2006). This study presents a first step in

examining the temporality of participation patterns and how they evolve, and leverages

them to explain how and why certain groups are more successful than others.

Conclusion

Situated and socio-constructivist theories of learning and collaboration strongly advocate

problem solving to be authentic and ill-structured for meaningful learning to take place

(Brown et al., 1989; Scardamalia & Bereiter, 2003). Inasmuch as researchers strongly

argue for incorporating ill-structured problem solving into the curriculum (Hmelo-Silver,

2004; Jonassen, 2000; Spiro & Jehng, 1990), this study provides evidence that we need to

be more mindful of the tenuous balance between maximizing problem-centered interac-

tional activity and reducing participation inequity when using ill-structured problems for

collaborative work. Because participation during collaboration is a consistent predictor of

individual learning gains (Cohen, 1994), the purported benefits of ill-structured problem

solving need to be balanced with a focus on facilitating equity in collaboration. Of course,

it is conceivable that some level of scaffolding through structuring the task may be a way

forward (Dillenbourg, 2002). But this approach arguably runs against the grain of having

students engage in ill-structured problem solving because scaffolds, by definition, reduce

the very complexity and structuredness that is argued for by theory. Perhaps the trick lies in

structuring the task in a manner that makes it neither overly simple so as to lose a group’s

interest nor too complex lest the group should simply give up trying. In short, it may be

that task structuring that stretches the ‘zone of proximal development’ (Vygotsky, 1978)

for the group may lead to a maximally productive collaboration. Needless to say, it is much

too early to suggest any generalization of these findings from this sample to other contexts,

technology-mediated environments, content domains, age-groups, cultures, etc. What we

do have, however, are sufficient warrants to articulate some implications and invite the

field to examine them further.

Appendix – problem scenarios

The well-structured problem used in the study

You are an inspector for a car insurance company. You’ve been assigned the following case:

A car was involved in a head-on collision with a delivery truck on a straight road. The

collision caused a 67 cm deep dent in the front of the car and a 5 cm deep dent in the front

of the truck. Witnesses say that it all happened so fast that neither driver had any time to

brake. Based on the injuries sustained by the drivers, the doctor estimated the force of the

collision to be between 120,000 N and 150,000 N. According to the car’s information

manual, its mass is 540 kg. The driver of the car has a mass of 60 kg. The speed limit on

the road is 60 kmph.

The driver of the car, insured by your company, is claiming insurance to cover the cost of

repair due to the accident. However, if he was speeding, your company has the right to reject

his claim. Would you accept or reject the driver’s claim? Present your case as best you can.

Synchronous collabortive problem solving 455

123

The ill-structured problem used in the study

You have recently been hired as an inspector for the India Insurance Company. On your

first day you are sent to the site of an accident between a small car and a delivery truck. As

you arrive, the ambulance is carrying away the driver of the small car who seems conscious

but bruised and shaken up. Opening your work file, you find your assignment:

Dear new inspector,

At 7:30 am this morning, a driver insured by our company (policy #241-575-374B) collided with a deliverytruck in a small alley in downtown Ghaziabad. Although the accepted speed limit in the alley is 25 kmph,the damage seems rather large. Please determine whether we can apply clause 315-6 to the policy holder.Note that doing this requires a solid body of evidence. Although I don’t recall your first name, I do recallbeing told good things about the quality and thoroughness of your work. Please submit your report tome with your analysis and recommendation by today.

Sincerely,

Amit ‘‘the Boss’’

P.S.: Since this is your first day, I have attached clause 315-6 to this letter.

Clause 315–6: - The policy will cover the cost of repair for collisions involving the policy holder. In theeventuality where the policy holder is found criminally responsible1, or reckless2 in his or her driving, theinsurance company will assume 50% of the repair costs and reserves the right to increase the premiumover the following 5 years. In order for the company to pay any amount, the holder agrees to yield accessto any medical files related to the accidents.

1The term criminally responsible refers to driving under the influence of substances such as alcohol, or illicitsubstances such as heroin or cocaine.

2The term reckless refers to driving without respecting the driving code - such as cutting through more thantwo lanes in less than 100 m or driving more than 35kmph above the prescribed speed limit.

Customer File

Policy No: 241-575-374B

Name: Mr. Rahul Singh

Age: 52 yrs

Driving Experience: 24 yrs

Previous Claims: 1993 – 20,000 Rupees; 1981 – 5,000 Rupees

Policy Type: 2 way insurance, including: Fire, Theft, & Vandalism (Max 100,000 Rupees)

Civil Responsibility: 1,000,000 Rupees

Deductible: 1000 Rupees

Insured car: Zen

To carry out your investigation, you go through a number of steps such as (a) inter-

viewing eye-witnesses, (b) analyzing the accident scene, (c) accessing the driver’s medical

file, and (d) interviewing the treating Emergency Room (ER) physician.

Eye witness’ account

‘‘I saw the car coming into the alley. I’m not too sure how fast it was going. I heard a big

BANG! It all happened so fast. It looked like the driver didn’t see the truck. I am not sure

but I don’t even think the car had time to brake.’’

456 M. Kapur, C. K. Kinzer

123

Accident Scene

• Mass of the car (from the car’s information manual) = 540 kg• Head-on collision between the car and the truck.

• Front end of car collapsed: 1700 (43 cm) remaining between front license plate andcentre of front wheel.

• Front end of truck slightly dented: about 200(5 cm) in depth.

• Slight evidence of skid marks: only about 600 (15 cm).• Original distance between front license plate and centre of front wheel (from the car

manual): about 43.300 (110 cm)

References

Arrow, H., Mcgrath, J. E., & Berdahl, J. L. (2000). Small groups as complex systems. Thousand Oaks, CA:Sage Publications Inc.

Albanese, M., & Mitchell, S. (1993). Problem-based learning: A review of the literature on its outcomes andimplementation issues. Academic Medicine, 68(1), 52–81.

Barab, S. A., Hay, K. E., & Yamagata-Lynch, L. C. (2001). Constructing networks of action-relevantepisodes: An in-situ research methodology. The Journal of the Learning Sciences, 10(1&2), 63–112.

Barron, B. (2003). When smart groups fail. The Journal of the Learning Sciences, 12(3), 307–359.Bar-Yam, Y. (2003). Dynamics of complex systems. Reading, MA: Addison Wesley.Bransford, J.D., & Nitsch, K. E. (1978). Coming to understand things we could not previously understand.

In J. F. Kavanaugh & W. Strange (Eds.), Speech and language in the laboratory, school, and clinic.Harvard, MA: MIT Press.

Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learning. EducationalResearcher, 18(1), 32–42.

Chi, M. T. H. (1997). Quantifying qualitative analyses of verbal data: A practical guide. The Journal of theLearning Sciences, 6(3), 271–315.

Cho, K. L., & Jonassen, D. H. (2002). The effects of argumentation scaffolds on argumentation and problemsolving. Educational Technology, Research and Development, 50(3), 5–22.

Medical Chart

BP (Blood Pressure): 105/65

HR (Heart Rate): 100

Weight: 60 kg

Notes: Blue-black bump on forehead; Major belt laceration on neck, and chest.

Drug/Alcohol Screen: Negative

Treating ER Physician

Dr: That seat belt saved his life. This was a considerable impact.

You: How could you tell?

Dr: Well, from experience I could tell you that the depth of the wound from the seat belt corresponds toan impact ranging between 20 g and 25 g.

You: Wow! 20 to 25 times the gravitational acceleration, that’s enormous. How confident are you of thisvalue?

Dr: Well it certainly is more than 20 g but not profound enough for 25 g. Well, I have to run now, I’mbeing paged.

You: OK. Thank you for your time.

Synchronous collabortive problem solving 457

123

Cohen, E. G. (1994). Designing groupwork: Strategies for heterogeneous classrooms. (Eds.), New York:Teachers College Press.

Cohen, E. G., Lotan, R. A., Abram, P. L., Scarloss, B. A., & Schultz, S. E. (2002). Can groups learn?Teachers College Record, 104(6), 1045–1068.

Cohen, J. (1977). Statistical power analysis for the behavioral sciences. NY: Academic Press.Dillenbourg, P. (1999). Collaborative learning: Cognitive and computational approaches. NY: Elsevier

Science.Dillenbourg, P. (2002). Over-scripting CSCL: The risks of blending collaborative learning with instructional

design. In P. A. Kirschner (Eds.), Three Worlds of CSCL: Can we support CSCL? (pp. 61–91).Heerlen: Open Universiteit Nederland.

Erkens, G., Andriessen, J., & Peters, N. (2003). Interaction and performance in computer-supported col-laborative tasks. In H. van Oostendorp (Eds.), Cognition in a Digital World(pp. 225–252). Mahwah,NJ: Erlbaum.

Fischer, F., & Mandl, H. (2005). Knowledge convergence in computer-supported collaborative learning:The role of external representation tools. The Journal of the Learning Sciences, 14(3), 405–441.

Gallagher, S. A., Stepien, W. J., & Rosenthal, H. (1992). The effects of problem-based learning on problemsolving. Gifted Child Quarterly, 36(4), 195–200.

Ge, X., & Land, S. M. (2003). Scaffolding students’ problem-solving processes in an ill-structured taskusing question prompts and peer interactions. Educational Technology, Research and Development,51(1), 21–38.

Hmelo, C. E. (1998). Problem-based learning: Effects on the early acquisition of cognitive skill in medi-cine.The Journal of the Learning Sciences, 7, 173–208.

Hmelo-Silver, C. E. (2004). Problem-based learning: What and how do students learn? EducationalPsychology Review, 235–266.

Howe, C., Tolmie, A., Anderson, A., & MacKenzie, M. (1992). Conceptual knowledge in physics: The roleof group interaction in computer supported learning. Learning and Instruction, 2, 161–183.

Jonassen, D. H. (2000). Towards a design theory of problem solving. Educational Technology, Research andDevelopment, 48(4), 63–85.

Jonassen, D. H. & Kwon, H. I. (2001). Communication patterns in computer-mediated vs. face-to-face groupproblem solving. Educational Technology, Research and Development, 49(1), 35–52.

Kapur, M. (2006). Productive failure. In S. Barab, K. Hay, & D. Hickey (Eds.), Proceedings of theInternational Conference on the Learning Sciences (pp. 307–313). Mahwah, NJ: Erlbaum.

Kapur, M., & Kinzer, C. (2007). Sensitivities to early exchange in synchronous computer-supported col-laborative learning (CSCL) groups. Paper presented at the annual meeting of the American EducationalResearch Association, Chicago, IL, USA.

Kapur, M., Voiklis, J., Kinzer, C., & Black, J. (2006). Insights into the emergence of convergence in groupdiscussions. In S. Barab, K. Hay, & D. Hickey (Eds.), Proceedings of the International Conference onthe Learning Sciences (pp. 300–306). Mahwah, NJ: Erlbaum.

Kirschner, P. A., Sweller, J., & Clark, R. E. (2006). Why minimal guidance during instruction does notwork: An analysis of the failure of constructivist, discovery, problem-based, experiential, and inquiry-based teaching. Educational Psychologist, 41(2), 75–86.

Kline, R. B. (1998). Principles and practice of structural equation modeling. Guilford: New York andLondon.

Lam, S. K. (1997). The effects of group decision support systems and task structures on group communi-cation and decision quality. Journal of Management Information Systems, 13(4), 193–215.

Lee, E. Y. C., Chan, C. K. K., & van Aalst., J. (2006). Students assessing their own collaborative knowledgebuilding. International Journal of Computer-Supported Collaborative Learning, 1(1), 57–87.

Light, P., & Glachan, M. (1985). Facilitating of problem solving through peer interaction. EducationalPsychology, 5, 217–225.

Littleton, K., & Hakkinen, P. (1999). Learning together: Understanding the processes of computer-supportedcollaborative learning. In P. Dillenbourg (Eds.), Collaborative learning: Cognitive and computationalapproaches (pp. 20–30).Oxford: Elsevier.

Palincsar, A. S., & Brown, A. (1984). Reciprocal teaching of comprehension-fostering and comprehensionmonitoring activities. Cognition and Instruction, 1(2), 117–175.

Poole, M. S., & Holmes, M. E. (1995). Decision development in computer-assisted group decision making.Human Communications Research, 22(1), 90–127.

Rosenholtz, S. J. (1985). Treating problems of academic status. In J. Berger & M. Zelditch, Jr. (Eds.), Status,rewards, and influence (pp. 445–470). San Francisco: Jossey-Bass.

Rourke, L., & Anderson, T. (2004). Validity in quantitative content analysis. Educational Technology,Research and Development, 52(1), 5–18.

458 M. Kapur, C. K. Kinzer

123

Scardamalia, M., & Bereiter, C. (2003). Knowledge building. In J. W. Guthrie (Eds.), Encyclopedia ofEducation. New York, USA: Macmillan Reference.

Schellens, T., Van Keer, H., Valcke, M., & De Wever, B. (2005). The impact of role assignment as ascripting tool on knowledge construction in asynchronous discussion groups. In T. Koschmann, D.Suthers, & T. W. Chan (Eds.), Proceedings of the International Conference on Computer SupportedCollaborative Learning 2005 (pp. 557–566). Mahwah, NJ: Erlbaum.

Schwartz, D. L. (1999). The productive agency that drives collaborative learning. In P. Dillenbourg (Eds.),Collaborative learning: Cognitive and computational approaches (pp. 197–218). NY: ElsevierScience.

Shin, N., Jonassen, D. H., & McGee, S. (2003). Predictors of well-structured and ill-structured problemsolving in an astronomy simulation. Journal of Research in Science Teaching, 40(1), 6–33.

Spada, H., Meier, A., Rummel, N., & Hauser, S. (2005). A new method to assess the quality of collaborativeprocess in CSCL. In T. Koschmann, D. Suthers, & T. W. Chan (Eds.), Proceedings of the InternationalConference on Computer Supported Collaborative Learning 2005 (pp. 622–631). Mahwah, NJ:Erlbaum.

Spiro, R. J., & Jehng, J. (1990). Cognitive flexibility and hypertext: Theory and technology for the non-linear and multi-dimensional traversal of complex subject matter. In D. Nix & R. Spiro (Eds.),Cognition, Education, and Multimedia. Hillsdale, NJ: Erlbaum.

Stahl, G. (2005). Group cognition in computer-assisted collaborative learning. Journal of Computer AssistedLearning, 21, 79–90.

Stevens, J. P. (2002). Applied multivariate statistics for the social sciences. Mahwah, NJ: Erlbaum.Suthers, D. D. (2006). Technology affordances for intersubjective meaning making: A research agenda for

CSCL. International Journal of Computer-Supported Collaborative Learning, 1(3), 315–337.Suthers, D., & Hundhausen, C. (2003). An empirical study of the effects of representational guidance on

collaborative learning. The Journal of the Learning Sciences, 12(2), 183–219.Vygotsky, L. S. (1978). Mind in society. Cambridge, MA: Harvard University Press.Weinberger, A., Stegman, K., & Fischer, F. (2005). Computer-supported collaborative learning in higher

education: Scripts for argumentative knowledge construction in distributed groups. In T. Koschmann,D. Suthers, & T. W. Chan (Eds.), Proceedings of the International Conference on Computer SupportedCollaborative Learning 2005 (pp. 717–726). Mahwah, NJ: Erlbaum.

Manu Kapur is an Assistant Professor of Learning Sciences and Technologies at the National Institute ofEducation of Nanyang Technological University of Singapore.

Charles K. Kinzer is a Professor of Education and Technologies, and Coordinator of the program inCommunication, Computing and Technology in Education at Teachers College, Columbia University inNew York.

Synchronous collabortive problem solving 459

123