Creative Education
2012. Vol.3, No.5, 636-642
Published Online September 2012 in SciRes (http://www.SciRP.org/journal/ce) http://dx.doi.org/10.4236/ce.2012.35093
Copyright © 2012 SciR e s . 636
Towards an On-Line Student Assistant within
a CSCL Framework
Fernando Ramos-Quintana, M. Dolores Vargas-Cerdán, J. Rafael Rojano-Cáceres
Tecnológico de Monterrey, Campus Cuernavaca Autopis ta del Sol, Xochitepec, México
Received July 20th, 2012; revised August 18th, 2012; accepted August 30th, 2012
The CSCL framework of this work aims at reinforcing knowledge by rebuilding and representing it syn-
thetically in form of a network of concepts, which is built through exchanged messages using a chat and a
graphical tool. We have found that during the collaborative learning sessions important troubles bring
about deadlocks affecting the learning process. The hypothesis of this work states that significant troubles
are expressed by students in form of assistance requests originating behavior patterns associated with cer-
tain phases defined within the collaborative learning sessions. Analysis of dialogs of real collaborative
learning sessions confirmed this hypothesis, whose results point towards the development of an on-line
assistance system that aims at breaking the deadlocks caused by troubles expressed in the form of assis-
tance requests. Derived of an analysis of 55 dialogs of a course of databases, we defined three phases
characterized by particular assistance requests occurring during the collaborative learning sessions: a) a
bootstrap phase characterized by requests about doubts of the use of tools; b) a work phase, wherein re-
quests are related with the construction of the network concepts; c) the goodbye phase wherein requests
reflect doubts about the final deliverable network.
Keywords: Computer-Supported Collaborative Learning; Dialogues; Behavior Patterns
Introduction
This work is situated within a CSCL environment that uses
dialogs to build knowledge synthetized and represented by a
network of concepts (NoC). This CSCL environment aims at
encouraging collaborative learning attitudes and the develop-
ment of cognitive skills to retain and build knowledge. The
knowledge about the topic under study is synthetized in the
form of a NoC represented by a graph similar to Petri Nets.
The communication between students to collaborate in the
construction of the network of concepts is established through a
dialog, which can be formally defined as follows:
12 1nn
mmm m

 Dialog 
where, the symbol “<” represents the temporal term “before”;
mi represents any message, and i and n are integer numbers.
Therefore, the dialog can be seen as a totally ordered sequence
of exchanged messages whose goal is to build, in a collabora-
tive way, the knowledge of a topic expressed synthetically by a
NoC.
The exchanged messages contain not only drawing actions
that aim at linking concepts to form the network, but also
doubts about the use of the tool, the way of linking the concepts
and general questions about the final product to be delivered.
The doubts about the use of the tool and the way of linking
concepts represent the most important causes of deadlocks that
affect the performance of the collaborative learning process.
We have observed that as the dialog evolves, some important
behaviors associated with requests of assistance are related with
the causes of deadlocks mentioned before.
In this article, we are focused on the analysis of messages of
dialogs to determine a set of behavior patterns that can aid to
identify the required assistance. These results are being used to
develop an on-line assistant system, currently under develop-
ment, that can provide students with aids to break the deadlocks.
This assistant system is out of the scope of this article.
We have discovered patterns that reveal behaviors by ana-
lyzing 55 dialogs of 55 pairs of students. These 55 dialogs,
which constitute a total of 4910 exchanged messages, are de-
rived from three different semesters, two groups by semester, of
a database course of an undergraduate program in Computer
Science, based on this analysis we have defined different types
of assistance requests which cause deadlocks and consequently
damage the performance of the collaborative learning process.
Thereby, the types of assistance requests were defined within
three phases related with behaviors of students occurring during
the construction of the network: a) the initial phase is charac-
terized by assistance requests mainly related with the use of the
tool; b) in the intermediate or work phase, the students re-
quested mostly assistance about the construction of the NoC; c)
the final phase is characterized by requests about details of the
built network and the product to be delivered.
The remaining sections of this paper are organized as follows:
the second section reviews relevant work related with our work
in the domain of CSCL systems; the third section describes the
CSCL system within which this investigation is carried-out.
The construction of network of concepts is specially treated in
this section; the fourth section presents an analysis and discus-
sion of qualitative and quantitative aspects of the messages,
which were used for the analysis that aimed at determining
behavior patterns that affect the collaborative learning proc-
esses within the context of a course of databases; finally, the
fifth section shows the conclusions.

F. RAMOS-QUINTANA ET AL.
Relevant Related Work
CSCL Systems
Dillenburg and other people coincide in defining collabora-
tive learning as a situation within which two or more people
learn or attempt to learn something together (Dillenburg et al.,
1996; Dillenburg, 1999; Collazos, 2001). It improves the indi-
vidual learning and provides the students with the opportunity
to develop cognitive skills meanwhile they develop social abili-
ties (Barnes, 1977). Roschelle (Roschelle et al., 1995) affirms
that collaboration is “the mutual engagement of participants in
a coordinated effort to solve a problem together”.
The benefit of the collaborative approach for learning lies in
the processes of articulation, conflict and co-construction of
ideas occurring when working closely with a peer. Participants
in problem-solving situations have to make their ideas explicit
to other collaborators, disagreements, justifications and nego-
tiations, helping students to coverage to a common object of
shared understanding (Crook, 1996).
One of the most important joint projects of computer science
and education research is determining how to devise and deploy
software tools to support classroom instruction. One approach
is related with computer supported collaborative learning
(CSCL) systems, which advocate group collaboration to pro-
mote learning among the students.
However, the benefits of collaborative learning are only
achieved by active teams that function well (Soller, 2001). In
addition, adequate technology tools and methodologies should
be constantly improved to enrich the collaborative learning
environments: in a review by Soller (Soller et al., 2005) they
studied tools and methodologies that support collaborative
learning interactions; argumentation in collaborative learning
systems can help to facilitate the communication between par-
ticipants as argued in a review of computer supported argu-
mentation made by Scheuer et al. (Scheuer, 2010); an essay of
the relationship between technology and collaborative learning
performance is carried-out in Stahl (Stah, 2006); such im-
provements can be related with the development of network
architectures (Bote, 2004).
The analysis of collaborative interactions aims at monitoring
student’s activity and assessing the performance of learning:
Beatriz Barros and M. Felisa Verdejo (Barros and Verdejo,
2000) analyzed student’s interactions to improve collaborations;
Juan (Juan 2008) developed an information system to monitor
the student’s activity in online collaborative learning; Putam-
bekar (Putambekar, 2006) aimed at knowing whether as the
collaborative interactions evolve new ideas raise or the built
knowledge is enriched; in a study made by Yang and Chang
(Yang & Chang, 2012) it has been found that interactive blogs
are associated with positive attitudes towards academic
achievement.
Assistance systems to help students in case of stuck situa-
tions have been one of the most important concerns in CSCL.
The work presented in this paper belongs to these kinds of sys-
tems. Dialogs generated in collaborative learning systems have
attracted an important number of researchers, who have been
interested in their analysis as source of data to gain insights into
the quality and performance of the collaboration process. Other
researchers like Pilkington (Pilkington, 2001) consider that the
analysis of dialogs can help bridge the gap between the empiri-
cal evaluation of interaction and the design of Intelligent Edu-
cational Systems (IESs) capable of interacting with their users;
Woojin Paik et al. (Paik, 2004) aimed at developing interven-
tion techniques to identify and remove obstacles of online
learning groups. The final goal is to build an automatic system
able to monitor the activities of the online learning group
members to alert the instructors when the members encounter
barriers; In web-enhanced courses, instructors would like to
review the discussions of students to understand the kind of
contributions, monitor their progress or verify if they need as-
sistance or guidance (Painter et al., 2003).
Some works consider the intervention of agents to support
the collaborators by providing aids in stuck situations or im-
proving the participation of students. In (Kim, 2007) speech act
analysis was carried-out to assess the effect of instructor inter-
vention on student participation, resulting in an increase of the
number of exchanged messages made by the students. However,
one of the most complex problems to assist the participants in
on-line mode is to be able of detecting the different categories
of messages, and thus defining the kind of assistance that could
correspond to that one requested by a particular message. The
following approaches, most of them based on the analysis of
speech acts derived from dialogs under study, could potentially
be adapted and applied to contribute to solve several aspects of
this problem: the Cohen’s work (Cohen, 2004) aimed at detect-
ing many categories of messages with high precision and mod-
erate recall by using text-classification learning methods; the
Feng’s work (Feng, 2006) is centered on detecting which mes-
sages in a thread contains the focus of the conversation, local-
izing what messages are related with the subject, recovering
past conversations and using them to solve doubts. This work
integrates studies of conversational speech acts, an analysis of
message values based on poster trustworthiness and an analysis
of lexical similarity; in (Ravi & Kim, 2007), speech act classi-
fiers were developed which were able to identify whether a
message contains questions or answers; the purpose of a sub-
sequent Kim’s work (Kim, 2008) was to develop tools that
could automatically assess student participation and promote
interactions by sending responses to student messages. These
software tools apply data mining and information retrieval
techniques for guiding student discussions. Thus, past student
discussions of related courses are used and the retrieved infor-
mation is presented on the discussion board; in (Kim et al.,
2005) was developed a course ontology that represents generic
components of distance education courses and a query ontology
that describes types of student’s queries and requests. The sys-
tem can map student queries to relevant course materials and
the results are sent to the students. In case of non-appropriate
mappings or the sent materials do not satisfy the student re-
quests, the case is sent to the instructor’s attention; in (Caballé,
2008) is proposed a model based on data analysis from online
collaborative interactions towards an automatic assessment in
real time. The use of machine learning approaches is suggested
to label automatically the messages; in an interesting approach
by Seo (Seo et al., 2011) it was examined whether interaction
patterns could be classified automatically by using a state tran-
sition model to identify successful versus unsuccessful student
Q & A discussions, and classification of threads as success-
ful/unsuccessful using the state information.
The analysis of messages to determine what assistance has
been requested could be derived from only one message or
from a sequence of them. Nevertheless, to precise the requested
assistance based only on one message, and moreover only con-
sidering key words belonging to it, could be result erroneous
because a correct interpretation of a request needs contextual
Copyright © 2012 SciRe s . 637

F. RAMOS-QUINTANA ET AL.
Copyright © 2012 SciRe s .
638
information, which is usually obtained by considering a se-
quence of messages but not only one. This approach has asso-
ciated an important complexity, because one isolated message
could have several interpretations, thus two or more messages
together result in the product of individual interpretations of the
messages being considered. For instance, if one message is
associated with three illocutive acts, in case of an analysis
based on speech acts, then this message would have three po-
tential interpretations. Such problem could require dealing with
more than one message to facilitate the interpretation. In order
to simplify the exercise, let’s suppose now that three messages
are being considered to give an adequate contextual interpreta-
tion; and let’s also suppose that each one of the messages has
three possible interpretations. Thus, the product of these three
messages results in at least nine possible interpretations and
thus the interpretation ambiguity starts to be a real problem.
Moreover, this exercise supposes that the participants write the
messages with a correct syntax and semantic way, which is not
commonly the case, at least in our Mexican reality. The modern
technologies of communications, such as the ones using in col-
laborative learning systems supported by computers, have con-
siderable modified the way of writing in such a way that is very
hard to associate a formal linguistic structure with dialogs es-
tablished by students during a collaborative learning process.
Therefore, the interpretation ambiguity increases significantly
and then the application of a formal approach, such as the
analysis based on speech acts, aiming at interpreting the inten-
tion of messages becomes a very complex task to be accom-
plished. In addition, as exposed by Kim (Kim, 2008), the dis-
cussion data from students, especially in undergraduate pro-
grams, is highly incoherent and noisy. The raw data includes
humorous messages and personal announcement as well as
technical questions and answers.
Owing to the problems exposed before, we have adopted an-
other approach, different than those based on the interpretation
of speech acts, towards the on-line intervention in the collabo-
rative process when the students request assistance in stuck
situations. The study presented in this paper deals with one
relevant step of the whole objective, which is to assist student
requests in on-line mode to unstuck deadlock situations. This
step aims at characterizing the kind of assistance requests asso-
ciated with different phases of the learning process during a
work session. We have proven, in a previous work (Ramos-
Quintana et al., 2008), the existence of a pattern composed of
three phases during a collaborative work session: a bootstrap
phase, a work phase and a goodbye phase. This issue is dis-
cussed later in the forth section.
The Computer Supported Collaborative
Learning (CSCL) System
The CSCL model is supported by a collaborative computer
tool called Collaborative Distributed Tool (CDT) (Rojano-
Caceres et al., 2010) developed by the research group on Com-
puter Supported Collaborative Learning from the Tecnológico
de Monterrey Campus Cuernavaca and the Universidad Ve-
racruzana. This tool, which can be accessed at
http://collaborativelearningframework.net/, is integrated by
three elements: 1) A chat tool that serves to build the dialog
whose objective is to generate the drawing actions to build the
NoC; 2) A graphical working space, where the network of con-
cepts is built, and; 3) A bottom-up methodology to build the
NoC.
Figure 1 illustrates the interface of CDT. As we can see, two
Chat Toll Gr aphic Tool
Collborative
Distributed
Tool
(CDR)
Keren: estas aperiendao ami
Marco: como ves asi?
Marco:
Keren: haber el cuadr ito Amarillo es la entidad
Keren: por ejemplo
Keren: seria
Marco: elamarillo es la relacion
Keren: aaa ok
Marco: los rosas ovalados son las entidades
Keren: aaaa
Marco: como vez?
Keren: si, pero el atributo tmb puede ser la llave primaria
Marco: si eso estaba pensando tambien..
Marco: :D
Keren: :)
Marc o: propon alguno···
puede ser como empleado tiene el atributo
Keren: numbe
r
, id···
Figure 1.
The CSCL framework.

F. RAMOS-QUINTANA ET AL.
students communicate through the chat to build the dialog. A
shared area allows students to build the NoC by drawing geo-
metric figures denoting nodes and directed arcs (represented by
arrows). Two kinds of nodes are available: nodes in form of
circles represent concepts to be linked; nodes in form of rectan-
gles represent relations between concepts. Directed arcs link
circle-nodes with rectangle-nodes and vice-versa. Below we
explain how a NoC is built, as well as the role of nodes and
directed arcs in this construction.
The Network of Concepts: One of the main abilities when
the retention or acquisition of knowledge process is taking
place is to synthesize and represent such knowledge in a
structured way by linking the concepts involved in a correct
semantic way. For this purpose, the development of the skill of
analysis will allow to extract the most relevant underlying
concepts and establish relations between them in order to get a
structural representation of knowledge being learned. This
structural representation is called Network of Concepts (NoC)
in this work. This way of working leads students to build
knowledge from the isolated underlying concepts to structural
concepts by developing at the same time different abilities such
as analysis, synthesis, construction of structures, abstraction
and generalization, which are very important in the process of
solving real problems.
The NoC starts as follows to be built: underlying concepts
are represented by circle-nodes and rectangle-nodes serve to
link two circle-nodes through directed arcs represented by ar-
rows. Two kinds of rectangle-nodes are available: black rectan-
gle-nodes represent dynamic links and white rectangle-nodes
represent static links. We mean by dynamic links those that
imply actions of ‘execution’, ‘processing’ or ‘transformation’;
in the case of transformation a concept c1 is transformed into a
concept c2 through a dynamic relationship r1, as expressed by
the example of the rule (1) shown below.
1metamorphosis
12
tadpole frog
r
CC

  (1)
Semantically, this rule means: if a tadpole undergoes a
metamorphosis, then it will be transformed into a frog.
Meanwhile, static relationships denote some reference or
characteristic to another concept that does not imply a trans-
formation. For example, it can be used to stay the fact that ‘the
sun is a star’, such representation is straightforward, which is
built by linking the concept of “Sun” with the concept of “Star”
by the relationship “is a”, in Figure 2 we present graphically
this example.
The rest of the network is built following the same procedure.
For example, in the theme of Relational Model from the Data-
base subject we can get important concepts such as relations,
tuple or rows and grouping, those can be represented as shown
in Figure 3.
The Methodology for the Construction
of the NoC
We exemplify this methodology for the reinforcement mo-
ment, wherein a topic has been learned previously and we aim
at improving the retention of isolated underlying concepts and
Sun Is a Star
Figure 2.
Example of two linked concept s.
Figure 3.
Example of a whol e NoC of the topic of databases.
Copyright © 2012 SciRe s . 639

F. RAMOS-QUINTANA ET AL.
how they are structurally linked. At the beginning of the proc-
ess, the teacher provides the student with a topic to be rein-
forced, after carrying out an important effort of analysis the
students try to list the underlying concepts of the topic (the first
level). Thus, at this level, the development of the analysis skill
is prioritized; at the next level, underlying concepts are linking
by pairs through dynamic or static rectangle-nodes. The skills
to be developed at this level are, once again the analysis, then
synthesis and the construction of structures composed of two
linked underlying concepts; finally, they link the pairs until
getting the whole NoC. At the last level, the skills to be devel-
oped are: the analysis, the synthesis, the construction of
structures, the abstraction and the generalization.
Qualitative and Quan ti tative An al y si s: A Case Study
Description: This case of study took place at the Statistics
and Computer Science Faculty of the Universidad Veracruzana
in Mexico. This investigation has been made by analyzing the
contents of dialogs of 55 pairs of students in the field of a data-
base course in the undergraduate level. The 55 dialogs are de-
rived from several courses of three different semesters. In a
previous work (Ramos-Quintana et al., 2008), we have reported
a general pattern composed of three sequentially ordered phases:
1) the handshake phase (now renamed “the bootstrap phase”,
because the students exchanged messages not only to say hello,
but also messages that denote a kind of setup to start working in
the construction of the NoC); 2) the work phase, where the
construction of the network takes properly place; 3) the closing
of the dialog, where the students say good bye.
In this work, we have characterized those messages belong-
ing to each phase that reveal requests, implicit and explicit, of
assistance. The following messages are examples of implicit or
explicit requests of assistance:
We mean by explicit messages of assistance requests those
messages containing a question that reveals an explicit trouble
that entails deadlocks in the learning process.
Type of explicit messages: 1) How do I do to write inside
the circles? 2) But, how do I put arrows? 3) What is the purpose
of the arrows?
Meanwhile, in the implicit messages, there are not questions
explicitly expressed, but the messages express also implicitly
troubles that deadlock the learning process.
Type of implicit messages: 4) I’m trying to find out how to
link nodes!!! 5) I guess that the direction of the arrows of the
items inside the tables should point to the opposite direction.
The purpose of this work is not to interpret automatically the
meaning of messages. Instead, we make a not automatic analy-
sis to extract those messages containing implicit or explicit
request of assistance to characterize and associate them with the
different phases of a learning session described before.
Type of Mess ag es of the Boo tstrap Phase
During the bootstrap phase, the exchanged messages mainly
concern some salutations, questions to know whether the cou-
ples have enough information to start and questions, highlight-
ing doubts about the use of the computer tool and the objective
of the task to be carried-out. Some examples of messages rep-
resenting the bootstrap phase are the following: Hi, do we insert
the list of concepts??? I’am ready; Do you have some con-
cepts??? What’s up; What concepts you got; Do you know how
to insert arrows to link two nodes?? What do we have to do??
Ok, let’s start to work; why should we work with the most im-
portant concepts?? I’m trying to test the communication…; Tell
me how many concepts you have, I got five.
Type of Messages of th e Work Phase
During the work phase the troubles are related with the con-
struction of correct structures, for instance, why a node X
should be linked with a node Y and not with the node Z, the
sense of the arrow, etc. Some examples of typical messages
associated with the work phase are the following: Could we put
the transition?? But, we have to build the whole network??? We
have to get more concepts; Is it ready now??? Do you have any
another idea???? Do you know which is the sense of this ar-
row??? Let me see if I remember something else; The tuples
form a database?? Do you know why the node X should be
linked with the node Y…??
Type of Messages of the Goodb ye Phase
In the goodbye phase the students have doubts about whether
the NoC is already finished or about the destination of the final
deliverable product, etc. Some examples of messages associ-
ated with the goodbye phase are the following: Here it go…¡¡¡¡
I guess the network is ready; I’m not completely convinced…
But… is it ok like that??? Can we already leave; So, what else;
Do you think it is ready to be sent…??? What’s the e-mail ad-
dress to be sent?? But, what you think about this? Let’s make
the last touch.
Quantitative Aspects
The next analysis illustrates how much time has the couples
taken for each phase. The analysis has been done by consider-
ing 55 dialogs, for a total of 4910 exchanged messages. Each
dialog corresponds to a couple. For instance, Table 1 below
shows ranges of time consumed by students during the boot-
strap phase. Thus, most of the students, 45 couples from 55
(81.8% of couples), fell in the range between 30 seconds to 6
minutes in the bootstrap phase. Meanwhile, 655
(10.9%) fell
in the range between 7 to 13 minutes, and finally 455 (7.2%)
in the range between 14 to 22 minutes.
Table 2 shows ranges of time consumed by students during
the work phase. We can see that most of the couples 31 55
Table 1.
Bootstrap phase.
Ranges of time in minutes Number of couples %
From0.5 to 6 45 82
From7 to 13 6 11
From14 to 22 4 7
Table 2.
Work phase.
Ranges of time in minutes Number of couples %
From15 to 28 14 26
From29 to 46 31 56
From47 to 60 10 18
Copyright © 2012 SciRe s.
640

F. RAMOS-QUINTANA ET AL.
(56%) during the work phase fell in the range between 29 to 46
minutes to complete this phase. Meanwhile, 1455 (25%) fell
in the range between 15 to 28 minutes. And 1055 (18%)
couples took between 47 to 60 minutes.
Table 3 shows ranges of time consumed by students during
the goodbye phase. We can verify that most of the couples
4455 (80%) fell in the range from 30 seconds to 6.5 minutes
to complete this phase. Meanwhile, 655 (10.9%) took be-
tween 6.31 to 10.3 minutes and 555 (9%) fell in the range
between 10:31 to 14:31 minutes.
Qualitative Aspects
As mentioned before, a total of 55 dialogs were analyzed,
which totalized 4910 messages. 1558 messages from a total of
4910 were messages that requested assistance, which represents
31% of these messages, this amount is significantly important
because it represents 13 of the whole set of messages. The
total of messages denoting a request of assistance for the three
phases is qualitatively shared as follows: 138 (bootstrapt phase),
1240 (work phase), 180 (goodbye phase). Thus, from this 31%,
25% belongs to the work phase (12404910 ), that is, 80.6 % of
the total of messages that reveals assistance request. Due to the
fact expose before, we can conclude that most of the request
requiring a help is concentrated in the work phase, that is, in
doubts about the semantic construction of the network of con-
cepts, as expected.
Table 3.
Goodbye phase.
Ranges of time in minutes Number of couples %
From 0.3 to 6.5 44 80
From 6.6 to 10 6 11
From 10 to 14 5 9
Figure 4 shows an average curve denoting behavior patterns
for each phase. We can observe a certain increase of the curve
at the beginning of the bootstrap phase, which is characterized
by messages of assistance requests related with the use of the
tool, the preparation of materials to start with the work phase
and some expressions denoting salutations. At the end of this
bootstrap phase these kinds of messages tend to disappear.
However, the curve tends to increase right after the point where
the work phase begins. The work phase has its highest peak
within the period of time between 32 to 35 min. In the
neighborhood of this period of time the students suffered most
of the troubles related with the semantic construction of the
network. After the neighborhood of this interval of time (32 to
35), the curve tends at decreasing until the end of the phase. At
this point, the goodbye phase starts to increase but the peak is
not enough relevant to be considered as a serious cause to pro-
voke a real deadlock. The kinds of messages in this phase are
related with doubts about the finished work and also doubts
about the destination of the deliverable work, among the most
important.
Conclusion
Some efforts have been carried out to provide CSCL systems
with automatic capacity of assistance or supervision of stu-
dent’s performance during the learning process interaction. We
are currently working towards building such computer assistant
system. However, the development of this system needs some
important steps to be achieved. One of them aims at determin-
ing the adequate moment to assist the students along with the
kind of assistance they need: this work showed the results con-
cerning the study of both problems.
During the collaborative learning process an important num-
ber of troubles can bring about deadlocks, due to several rea-
sons, which put at risk the performance of the learning process.
Working under the collaborative learning framework described
in this article, we have confirmed that students request repeti-
Work phase Goodbye Phase
Bootstrap Pha s e
50
45
40
35
30
25
20
15
10
5
0
Number of demanda for assistance
1 3 5 7 9
11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97 99
101 103 105 107 109
111
113 115 117 119 121 123
Time
Figure 4.
The three phases denoting behavior patterns related with assistance requests per phase.
Copyright © 2012 SciRe s . 641

F. RAMOS-QUINTANA ET AL.
tively for assistance to solve some troubles, varying the fre-
quency of such requests in depending on the phases, described
before, of the learning process. In this work, we aimed at de-
termining quantitatively the number of requests for assistance
and discovering qualitatively the type of request of assistance,
which can reveal the causes of deadlocks. From a quantitative
point of view, this study revealed that about 30% of exchanged
messages expressed request of assistance, which represents a
very significant percentage, which can affect seriously the per-
formance of the learning process. Meanwhile, from the qualita-
tive point of view, the most important assistance requests were
related with doubts about the use of the computer tool and with
doubts about links, semantically correct, between underlying
concepts, which showed the relevance of the development of
skills to build correct structures.
Derived from these results, we have determined behavior
patterns associated with the phases (the bootstrap phase, the
work phase and the goodbye phase) of the learning process
defined in this work. Based on these behavior patterns, we can
count with the following elements to offer the assistance closer
to the student’s needs: the phase during which the assistance
request occurs and the kind of help that the student is requested.
As mentioned previously, this work is part of our final objec-
tive towards the development of an automatic computer system
capable to assist students when they get stuck during the learn-
ing process in on-line mode.
REFERENCES
Barnes, D., & Todd, F. (1977). Communication and learning in small
groups. London: Routledge & Kegan Paul Ltd.
Barros, B., & Verdejo, M. F. (2000). Analysing student interaction
processes in order to improve collaboration. The DEGREE approach.
International Journal of Artificial Intelligence in Education, 11, 221-
241.
Bote, M. L., Vaquero, L. M., Vega, G., Asensio, J. I., Gómez, E., &
Dimitriadis, Y. (2004). GRIDCOLE: A grid collaborative learning
environment. Proceedings of the 1st International Workshop on Col-
laborative Learning Applications of Grid Technology, Chicago, 19-
22 April 2004.
Caballé, S., Xhafa, F., & Abraham, A. (2008). Towards an automatic
real-time assessment of online discussions in CSCL practices. Pro-
ceedings of the 3rd International Conference on Digital Information
Management (pp. 470-475), London, 13-16 November 2008.
Cohen, W., Carvalho V., & Mitchell, T. (2004). Learning to classify
Email into “Speech Acts”. Proceedings of Empirical Methods on
Natural Language Processing (pp. 309-316), Barcelona, 27-31 July
2004.
Collazos, C., Guerrero, L., Pino, J., & Ochoa, S . (2002). Evaluating col-
laborative learning processes. Proceedings of the 8th International
Workshop on Groupware, La Serena, 1-4 September 2002, 203-221.
Crook, C. (1996). Computers and the collaborative experience of learn-
ing. New York: Routledge.
Dillenbourg, P. (1999). What do you mean by collaborative learning?
In P. Dillenbourg (Ed.), Collaborative-learning: Cognitive and com-
putational approache s (pp. 1-19). Oxford: Elsevier.
Dillenburg, P., Baker M., Blake A. et al. (1996). The evolution of re-
search on collaborative learning. In E. Spada, & P. Reimann (Eds.),
learning in humans and mac hi ne s (pp. 189-2 11). Oxford: Pergamon.
Feng, D., Shaw, E., Kim, J., & Hovy, E. (2006). Learning to detect
conversation focus of threaded discussions. Proceedings of the Joint
Human Language Technology Conference/Annual Meeting of the
North American (pp. 19-26), New Yor k, 4-9 June 2006.
Juan, A., Daradoumis, T., Faulin, J., & Xhafa, F. (2008). Developing an
information system for monitoring student’s activity in online col-
laborative learning. In Proceedings of the Complex, Intelligent and
Software Intensive Systems (pp. 270-275), Catalonia, 4-7 March
2008.
Kim, J., Beal, C., & Maqbool, Z. (2005). Developing teaching aids for
distance education. Proceedings of the 2005 conference on Artificial
Intelligence in Education: Supporting Learning through Intelligent
and Socially Informed Technology (pp. 848-850), Amsterdam, 18-22
July 2005.
Kim, J., Shaw, E., & Chern, G. (2007). Novel tools for assessing stu-
dent discussions: Modeling threads and participant roles through
speech act and topic analysis. Proceedings of the 13th International
Conference on Artificial Intelligence in Education, Marina Del Rey,
9-13 July 2007.
Kim, J., Shaw, E., Tavano, E., Arromratana, A., Sarda P., & Beal C.
(2008). Towards automatic scaffolding of on-line discussions in en-
gineering courses. Annual Meeting of the American Educational Re-
search Association, New York, 24-28 March 2008.
Paik, W., Lee, J. Y., & McMahan, E. (2004). Facilitating collaborative
learning in virtual (and sometimes mobile) environment. Lecture
Notes in Computer Science, 3307, 161-166.
Painter, C., Coffin, C., & Hewings, A. (2003). Impacts of directed
tutorial activities in computer conferencing: A case study. Distance
Education, 24, 159-174. doi:10.1080/0158791032000127455
Pilkington, R. (2001). Analysing educational dialogue interaction:
Towards models that support learning. Introduction to the IJAIED
special Issue on analysing educational dialogue interaction. Interna-
tional Journal of Artificial Intelligence in Education, 12, 1-7.
Puntambekar, S. (2006). Analyzing collaborative interactions: Diver-
gence, shared understanding and construction of knowledge. Com-
puters & Education, 47, 332-351.
Ramos-Quintana, F., Sámano-Galindo, J., & Zárate-Silva, V. Z. (2008).
Collaborative environments through dialogues and PBL to encourage
the self-directed learning in computational sciences. Lecture Notes in
Computer Science, 5102, 725-734.
Ravi, S., & Kim, J. (2007). Profiling student interactions in threaded
discussions with speech act classifiers. Proceedings of the AI in
Education Conference, Marina Del Rey, 9-13 July 2007.
Rojano, R., Ramos, F., & García, G. (2010). Proposal for a collabora-
tive distributed tool for building network of concepts. Puerto Val-
larta: Alfa-Omega, 383-388.
Roschelle, J., & Teasley, S. D. (1995). The construction of shared
knowledge in collaborative problem solving. In C. E. O’Malley (Ed.),
Computer-Supported Collaborative Learning (pp. 69-197). Berlin:
Springer-Verlag. doi:10.1007/978-3-642-85098-1_5
Scheuer, O., Loll, F., Pinkwart, N., & McLaren, B. (2010). Computer-
supported argumentation: A review of the state of the art. Interna-
tional Journal of Computer-Suported Collaborative Learning, 5, 43-
102.
Seo, S., Kang, J., Drummond, J., & Kim. J. (2011). Using graphical
models to classify dialogue transition in online Q & A discussions.
Lecture Notes in Computer Science, 6738, 550- 553.
doi:10.1007/978-3-642-21869-9_98
Soller, A. (2001). Supporting social interaction in an intelligent col-
laborative learning system. International Journal of Artificial Intel-
ligence in Education, 12 , 40-62.
Soller, A., Martínez Monés, A., Jermann, P., & Muehlenbrock, M.
(2005). From mirroring to guiding: A review of state of the art tech-
nology for supporting collaborative learning. International Journal
of AI in Education, 15, 261-290.
Stahl, G., Koschmann, T., & Suthers, D. (2006). Computer supported
collaborative learning: An historical perspective. In R. K. Sawyer
(Ed.), Cambridge handbook of the learning sciences (pp. 409-426).
Cambridge: Cambridge University Press.
Yang, C., & Chang, Y. S. (2010) Assessing the effects of interactive
blogging on student attitudes towards peer instruction, learning mo-
tivation and academic achievements. Journal of Compouter Assisted
Learning, 28, 126-135. doi:10.1111/j.1365-2729.2011.00423.x
Copyright © 2012 SciRe s.
642