soller-etal-its98
TRANSCRIPT
-
8/3/2019 Soller-etal-ITS98
1/11
In Proceedings of the 4th International Conference on Intelligent Tutoring Systems
(ITS 98). San Antonio, TX, 186-195.
Promoting Effective Peer Interaction in an Intelligent
Collaborative Learning System1
Amy Soller, Bradley Goodman, Frank Linton, and Robert Gaimari
The MITRE Corporation
202 Burlington Road, Bedford, MA, 01730-1420
Abstract. Placing students in a group and assigning them a task does notguarantee that the students will engage in effective collaborative learning
behavior. The collaborative learning model described in this paper identifies
the specific characteristics exhibited by effective collaborative learning teams,
and based on these characteristics, suggests strategies for promoting effective
peer interaction. The model is designed to help an intelligent collaborative
learning system recognize and target group interaction problem areas. Once
targeted, the system can take actions to help students collaborate more
effectively with their peers, maximizing individual student and group learning.
1 Introduction
Research has shown that classroom learning improves significantly when students
participate in learning activities with small groups of peers [2]. Students learning in
small groups encourage each other to ask questions, explain and justify their
opinions, articulate their reasoning, and elaborate and reflect upon their knowledge,
thereby motivating and improving learning. These benefits, however, are only
achieved by active and well-functioning learning teams. Placing students in a group
and assigning them a task does not guarantee that the students will engage in
effective collaborative learning behavior. While some peer groups seem to interact
naturally, others struggle to maintain a balance of participation, leadership,
understanding, and encouragement. Traditional lecture-oriented classrooms do not
teach students the social skills they need to interact effectively in a team, and few
students involved in team projects or exposed to integrated working environmentslearn these skills well [11].
The collaborative learning model (CL Model) described in this paper identifies the
characteristics exhibited by effective collaborative learning teams, and based on these
characteristics, suggests strategies for promoting more effective student interaction.
The model is designed to help an intelligent collaborative learning system recognize
and target group interaction problem areas. Once targeted, the system can take
1 This work was supported by the MITRE Sponsored Research Program.
-
8/3/2019 Soller-etal-ITS98
2/11
actions to help students collaborate more effectively with their peers, maximizing
individual student and group learning.
The next section describes the characteristics exhibited by effective collaborative
learning teams, and section 3 proposes strategies for helping groups achieve these
characteristics. Section 4 describes our ongoing efforts to implement the strategies in
an intelligent collaborative learning system, and section 5 summarizes the CL Model.
2 Characteristics of Effective Collaborative Learning Teams
This section describes the characteristics exhibited by effective collaborative learning
teams based on a review of research in educational psychology and computer-supported collaborative learning [2] [10] [11] [13] [18], and empirical data from a
study conducted by the authors of this paper [16]. The study was conducted during a
five day course in which students learned and used Object Modeling Technique to
collaboratively design software systems. The students were videotaped as they
worked in groups of four or five. The videotape transcriptions were coded with a
speech act based coding scheme, and studied using summary and sequential analysis
techniques.
The characteristics studied and seen to be exhibited during effective collaborative
learning interaction fall into five categories: participation, social grounding,
collaborative learning conversation skills, performance analysis and group
processing, and promotive interaction. The following five subsections describe these
categories and their corresponding characteristics.
2.1 Participation
A teams learning potential is maximized when all the students actively participate in
the groups discussions. Building involvement in group discussions increases the
amount of information available to the group, enhancing group decision making and
improving the students quality of thought during the learning process [10].
Encouraging active participation also increases the likelihood that all group members
will learn the subject matter, and decreases the likelihood that only a few students
will understand the material, leaving the others behind.
2.2 Social Grounding
Teams with social grounding skills establish and maintain a shared understanding of
meanings. The students take turns questioning, clarifying and rewording their peers
comments to ensure their own understanding of the teams interpretation of the
problem and the proposed solutions. In periods of successful collaborative activity,
students conversational turns build upon each other and the content contributes to
the joint problem solving activity [17].
-
8/3/2019 Soller-etal-ITS98
3/11
Analysis of the data collected from our study [16] revealed that students in
effective collaborative learning teams naturally take turns speaking by playing
characteristic roles [5] such as such as questioner, facilitator, and motivator.
2.3 Collaborative Learning Conversation Skills
An individuals learning achievement in a team can often be determined by the
quality of his communication in the group discussions [10]. Skill in learning
collaboratively means knowing when and how to question, inform, and motivate
ones teammates, knowing how to mediate and facilitate conversation, and knowing
how to deal with conflicting opinions.
The Collaborative Learning Skills Network (shown in part by Figure 1) illustratesthe conversation skills which are key to collaborative learning and problem solving,
based on our study [16]. This network is a modified version of McManus and
Aikens Collaborative Skills Network [13], which breaks down each cooperative
learning skill type (Communication, Trust, Leadership, and Creative Conflict) into its
corresponding subskills (e.g. Acknowledgment, Negotiation), and attributes (e.g.
Appreciation, Asking for information). Each attribute is assigned a sentence opener
which conveys the appropriate dialogue intention.
The goal of collaborative learning is for all the students to help each other learn
all the material together, whereas in cooperative learning each student individually
performs his or her assigned learning task, independently contributing to the groups
Collabo
rativeLearning
Skills
ActiveLearning
Motivate
Inform
Reques
t
C l a r i f ic a t i o n
J u s t i f i c a t i o n
O p i n i o n
E l a b o r a t i o n
R e i n f o r c e
R e p h r a s e
L e a d
S u g g e s t
E l a b o r a t e
E x p l a i n / C l a r i f y
J u s t i f y
A s s e r t
I n f o r m a t i o n
E n c o u r a g e
" T o e l a b o r a t e "
" I t h i n k "
" I t h i n k w e s h o u l d "
" In o t h e r w o r d s "
" T h a t ' s r i g h t "
" V e r y G o o d " , " G o o d P o i n t "
" W h y d o y o u t h i n k t h a t "
" C a n y o u e x p l a i n w h y / h o w "
" C a n y o u t e l l m e m o r e "
" D o y o u k n o w "
" I ' m r e a s o n a b l y s u r e "
" T o j u s t i f y "
" L e t m e e x p l a i n i t t h i s w a y "
" D o y o u t h i n k "
S k i l l S u b s k i l l A t t r i b u t e S e n t e n c e O p e n e r
Fig. 1. The Active Learning Skills section of the Collaborative Learning Skill Network
(adapted from McManus and Aikens Collaborative Skill Network, [13])
-
8/3/2019 Soller-etal-ITS98
4/11
final solution [8]. Effective collaborative learning interaction differs from
cooperative learning interaction in that students learning collaboratively utilize fewer
Trust and Leadership skills, and more Active Learning skills such as Justify,
Elaborate, Encourage, and Explain [16]. The Collaborative Learning Skills Network
reflects this difference.
2.4 Performance Analysis and Group Processing
Group processing exists when groups discuss their progress, and decide what
behaviors to continue or change [11]. Group processing can be facilitated by giving
students the opportunity to individually and collectively assess their performance.
During this self evaluation, each student learns individually how to collaborate moreeffectively with his teammates, and the group as a whole reflects on its performance.
2.5 Promotive Interaction
A group achievespromotive interdependence when the students in the group perceive
that their goals are positively correlated such that an individual can only attain his
goal if his team members also attain their goals [7]. In collaborative learning, these
goals correspond to each students need to understand his team members ideas,
questions, explanations, and problem solutions.
Students who are influenced by promotive interdependence engage in promotive
interaction; they verbally promote each others understanding through support, help,
and encouragement [11]. If a student does not understand the answer to a question or
solution to a problem, his teammates make special accommodations to address his
misunderstanding before the group moves on. Ensuring that each student receives the
help he needs from his peers is key to promoting effective collaborative interaction.
3 Strategies for Promoting Effective Peer Interaction
This section suggests implementation strategies for helping students attain the
collaborative learning skills required to excel in each of the five CL Model
categories.
3.1 Encouraging Participation
An intelligent collaborative learning system (ICLS) can encourage participation by
initiating and facilitating round-robin brainstorming sessions [10] at appropriate
times during learning activities. Consider the following scenario. An ICLS presents
an exercise to a group of students. After reading the problem description to himself
or herself, each group member individually formulates procedures for going about
-
8/3/2019 Soller-etal-ITS98
5/11
solving the problem. A student who is confident that he has the right procedure
may naturally speak up and suggest his ideas, whereas the student who is unsure (but
may actually have the best proposal) may remain quiet. During this phase of learning,
it is key that all students bring their suggestions and ideas into the group discussion,
especially since quiet students of lower ability have particular difficulty learning the
material [18]. The ICLS initiates and facilitates a round-robin brainstorming session a
few minutes after the students have read the problem description. Each student in the
group is required to openly state his rationale for solving the problem while the other
students listen. Round-robin brainstorming sessions establish an environment in
which each student in turn has the opportunity to express himself openly without his
teammates interrupting or evaluating his opinion. An ICLS can help ensure active
participation by engaging students in these sessions at appropriate times.
Personal Learning Assistants (PaLs), personified by animated computer agents,can be designed to partner with a student, building his confidence level and
encouraging him to participate. Providing a private channel of communication [12]
between a student and his personal learning assistant allows the student to openly
discuss his ideas with his PaL without worrying about his peers criticisms. A
students PaL could help him develop his ideas before he proposes them to the other
students. The personal learning assistant may also ask the student questions in order
to obtain a more accurate representation of his knowledge for input to the student
model. A more accurate student model allows coaching to better meet student needs.
3.2 Maintaining Social Grounding
An ICLS can model the turn-taking behavior that is characteristic of teams with
social grounding skills (section 2.2), by assigning the students roles [5] such as
questioner, clarifier, mediator, informer, and facilitator (among others), and rotating
these roles around the group for each consecutive dialogue segment. The beginning
of a new dialogue segment is identified by the start of a new context, often initiated
by sentence openers such as, OK, lets move on, or Now, what about this [16].
One or more critical roles, such as questioner or motivator, may be missing in a
group if there are too few students to fill all the necessary roles. A missing role can
be played by a simulated peer, or learning companion [6] [9]. The learning
companion can be dynamically adapted to best fit the needs of the group, playing the
role of critic during one dialogue segment, and facilitator during the next.
Identifying and characterizing the natural role switches that take place between
dialogue segments will aid in further developing advanced strategies for maintaining
social grounding through role playing.
3.3 Supporting Collaborative Learning Conversation Skills Practice
Students solving open-ended problems, in which an absolute answer or solution may
not exist, must explain their viewpoints to their peers, and justify their opinions.
-
8/3/2019 Soller-etal-ITS98
6/11
Assigning students open-ended activities encourages them to practice these essential
active learning conversation skills. A learning companion in an ICLS can also
encourage students to elaborate upon and justify their reasoning by playing the role
of devils advocate [10].
Providing students with collaborative learning skill usage statistics (e.g. 10%
Inform, 50% Request, 40% Argue) raises their awareness about the types of
contributions they are making to the group conversation. This capability, however,
requires either the ICLS to understand and code the students dialogue, or the
students to tell the system from which skill categories their utterances belong.
Section 4 describes an alternate method of obtaining collaborative learning skill
information.
3.4 Evaluating Student Performance and Promoting Group Processing
An ICLS can promote group processing by evaluating students individual and group
performance, and providing them with feedback. Students should receive individual
evaluations in private, along with suggestions for improving their individual
performance. The team should receive a group evaluation in public, along with
suggestions for improving group performance. The purpose of providing a group
evaluation is to inspire the students to openly discuss their effectiveness while they
are learning and determine how to improve their performance. This introspective
discussion may also be provoked by allowing the students to collaboratively view
and make comments on their student and group models [4].
3.5 Supporting Promotive Interaction
Webb [18] outlines five criteria for ensuring that students provide effective help to
their peers in a collaborative environment. These criteria are (1) help is timely, (2)
help is relevant to the students need, (3) the correct amount of elaboration or detail is
given, (4) the help is understood by the student, and (5) the student has an
opportunity to apply the help in solving the problem (and uses it!). The following
paragraphs suggest strategies to address each criteria.
When a student requests help, the ICLS can encourage his teammates to respond
in a timely manner. Assigning a mentor to each student provides them with a
personal support system. Student mentors feel responsible to ensure their mentees
understanding, and mentees know where to get help when they need it.In response to their questions, students must be provided with relevant
explanations containing an adequate level of elaboration. Their peers, however, may
not know how to compose high-quality, elaborated explanations, and may need
special training in using examples, analogies, and multiple representations in their
explanations [1]. To increase the frequency and quality of explanations, and ICLS
could strategically assign students roles such as Questioner and Explainer to help
them practice and improve these skills.
-
8/3/2019 Soller-etal-ITS98
7/11
Webbs fourth and fifth criteria can be met by an ICLS by observing and
analyzing a students actions in conjunction with his communicative actions to
determine whether or not a student understood and applied the help received.
4 Applying the CL Model to an ICLS
This section describes our ongoing research in the development of an intelligent
collaborative learning system based on the CL Model. Ideally, an ICLS would be
able to understand and interpret the group conversation, and could actively support
the group during their learning activities by applying the strategies described in
section 3. In working toward the realization of this vision, we have focused ourefforts on implementing group support strategies; later these strategies may draw
upon natural language understanding functionality.
Fig. 2. Sentence opener based communication interface
To enable student communication over a computer network, a chat style interface
(Figure 2), based on the Collaborative Learning Skill Network, was developed [15]
(see also [13] [14]). Students begin their statement with a sentence opener by
-
8/3/2019 Soller-etal-ITS98
8/11
selecting one of eight subskill categories: Request, Inform, Motivate, Argue,
Mediate, Task, Maintenance, or Acknowledge. The sentence openers represent
characteristic sentence starters from the subskill categories to which they belong.
Students view the group conversation as it progresses in the large window displaying
the students names and utterances. This sentence opener interface structures the
groups conversation, making the students actively aware of the dialogue focus and
discourse intent.
Each time a student makes a contribution to the group conversation, the
communication interface records the date, time, students name, and subskill applied
in the student action log (Figure 3). By dynamically performing summary analysis on
a groups logged interaction data, an ICLS can determine to what extent each student
is participating, and from which subskill categories he is contributing. This type of
data may allow the system to determine a students role within the group and what
skills he needs to practice. For example, although John has been following the group
conversation closely and interacting often, his summary CLC skill statistics show that
90% of his contributions are from the category Acknowledge.
The information in the student action log combined with knowledge of the student
actions taken on a shared workspace provides insight into the students competence
in the subject matter. For example, consider the sequence of actions shown in Figure
3. Brad asks Amy a question and Amy responds. Brad asks Amy a second question
and Amy responds again. Brad acknowledges Amys response and then creates a link
on the shared workspace, showing the relationship between his proposed hypothesis
and some new evidence he has discovered. Finally, Amy praises Brad for taking a
correct action. From this interaction sequence, the ICLS could deduce that Amy
helped Brad learn and apply the rule relating his hypothesis and the new evidence.This information could be used, for example, to enrich Amy and Brads student
models, enabling the ICLS to provide more effective coaching to them
Fig. 3. Example log generated from student communication and actions on a shared
workspace
The components of an ICLS, based on the CL Model, are responsible for using the
information obtained from the communication interface and shared tools to determine
when and how to apply the strategies described in section 3. For example, the
-
8/3/2019 Soller-etal-ITS98
9/11
instructional planning component may be responsible for not only selecting
appropriate exercises to present to the students, but also selecting roles to assign to
students. Table 1 summarizes the strategies that address each of the five CL Model
categories, and shows how the ICLS components can implement these strategies.
In the next stage of this research, a second study will be conducted in which
students from the first study will be asked to collaboratively solve problems similar
to those they solved before, however this time they will be required to communicate
using the sentence opener interface. The students will jointly design their software
system using a shared workspace. This study will test the usability of the
communication interface, and reveal any refinements needed to the collaborative
learning skill network. The ICLS components shown in Table 1 will then be
developed and tested to evaluate the CL Model.
5 Conclusion
Learning effective collaboration skills is essential to successfully learning course
Table 1. The CL Model support strategies that could be executed by each ICLS component
ICLS Component
CL Model Facet CL Skill CoachInstructionalPlanner
Student/ GroupModel
LearningCompanion
PersonalLearningAssistant
Participation
Facilitate round-
robin
brainstorming
sessions
Initiate round-
robin
brainstorming
sessions
Encourage
participation
SocialGrounding
Assign roles to
students, and
rotate roles
around group
Fill in missing
roles in
group
Ensure
students are
playing their
assigned roles
CollaborativeLearningConversationSkills
Provide
feedback on CL
skill usage
Assign tasks that
require students
to practice CL
skills
Store student/
group CL skill
usage statistics
Play devils
advocate to
encourage
active learning
PerformanceAnalysis &GroupProcessing
Provide
feedback on
group/
individual
performance
Allow students
to inspect and
comment on
their student/
group models
Promotive
Interaction
Ensure adequate
elaboration is
provided in
explanations
Assign mentors
or helpers to
students
Update
student/group
models when
students receive
help
Alert students
to their peers
requests for
help
-
8/3/2019 Soller-etal-ITS98
10/11
material with peers [3]. These skills are often not inherent, however they can be
learned through practice and support during structured group learning activities.
The CL Model described in this paper provides ICLS developers with a
framework and set of recommendations for helping groups acquire effective
collaborative learning skills. The model describes the characteristics exhibited by
effective learning teams, namely participation, social grounding, performance
analysis and group processing, application of collaborative learning conversation
skills, and promotive interaction.
References
1. Blumenfeld, P., Marx, R., Soloway, E., & Krajcik, J. (1996). Learning with peers: Fromsmall group cooperation to collaborative communities. Educational Researcher, 25(8),37-40.
2. Brown, A. & Palincsar, A. (1989). Guided, cooperative learning and individual knowledgeacquisition. In L. Resnick (Ed.), Knowledge, learning and instruction (pp. 307-336),Lawrence Erlbaum Associates.
3. Bruffee, K. (1993). Collaborative learning. Baltimore, MD: The Johns Hopkins UniversityPress.
4. Bull, S. & Broady, E. (1997). Spontaneous peer tutoring from sharing student models.Proceedings of the 8th World Conference on Artificial Intelligence in Education (AI-ED97), Kobe, Japan, 143-150.
5. Burton, M., Brna, P., & Treasure-Jones, T. (1997). Splitting the collaborative atom.Proceedings of the 8th World Conference on Artificial Intelligence in Education (AI-ED97), Kobe, Japan, 135-142.
6. Chan, T.W. & Baskin, A. (1988). Studying with the prince: The computer as a learningcompanion. Proceedings of the ITS 88 Conference, Montreal, Canada, 194-200.
7. Deutsch, M. (1962). Cooperation and trust: Some theoretical notes. In M. Jones (Ed.)Nebraska Symposium on Motivation (pp. 275-320). Lincoln: University of Nebraska Press.
8. Dillenbourg, P., Baker, M., Blaye, A., & OMalley, C. (1995). The evolution of research oncollaborative learning. In H. Spada and P. Reinmann (Eds.), Learning in Humans andMachines, Elsevier Science.
9. Goodman, B., Soller, A., Linton, F., & Gaimari, R. (1997). Encouraging student reflectionand articulation using a learning companion. Proceedings of the 8th World Conference onArtificial Intelligence in Education (AI-ED 97), Kobe, Japan, 151-158.
10. Jarboe, S. (1996). Procedures for enhancing group decision making. In B. Hirokawa andM. Poole (Eds.), Communication and Group Decision Making (pp. 345-383). ThousandOaks, CA: Sage Publications.
11. Johnson, D., Johnson, R., & Holubec, E. (1990). Circles of learning: Cooperation in theclassroom (3rd ed.). Edina, MN: Interaction Book Company.
12. Koschmann, T., Kelson, A., Feltovich, P., & Barrows, H. (1996). Computer-supportedproblem-based learning. In T. Koschmann (Ed.), CSCL: Theory and Practice of anEmerging Paradigm (pp. 83-124). Mahwah, NJ: Lawrence Erlbaum Associates.
13. McManus, M, & Aiken, R. (1995). Monitoring computer-based problem solving. Journalof Artificial Intelligence in Education, 6(4), 307-336.
14. Robertson, J. (1997). BetterBlether: An educational communication tool. Unpublishedundergraduate honours dissertation, Departments of Computer Science and ArtificialIntelligence, University of Edinburgh.
15. Soller, A. (1997). Toward an intelligent CSCL communication interface. Proceedings ofAI-ED 97 Workshop IV, Kobe, Japan, 94-95.
-
8/3/2019 Soller-etal-ITS98
11/11
16. Soller, A., Linton, F., Goodman, B., & Gaimari, R. (1996). [Videotaped study: 3 groups of4-5 students each solving software system design problems using Object ModelingTechnique during a one week course at The MITRE Institute]. Unpublished raw data.
17. Teasley, S. & Roschelle, J. (1993). Constructing a joint problem space. In S. Lajoie & S.Derry (Eds.), Computers as cognitive tools (pp. 229-257). Hillsdale, NJ: LawrenceErlbaum.
18. Webb, N. (1992). Testing a theoretical model of student interaction and learning in smallgroups. In R. Hertz-Lazarowitz and N. Miller (Eds.), Interaction in Cooperative Groups:The Theoretical Anatomy of Group Learning (pp. 102-119). New York: CambridgeUniversity Press.
Acknowledgments
Special thanks to the students in the OMT class we studied, and to the instructor for beingexceptionally cooperative. Thanks also to Brant Cheikes and Mark Burton for their support.