Computer-Supported Collaborative Learning (CSCL) has been shown to enhance learning by promoting peer interactions. With the support of collaborative learning flow patterns (CLFP), the collaboration process within the socalled CSCL macro-scripts can be further structured and constrained. However, social dynamics within tasks, even within scripts, have the potential to hinder the collaborative element as students might not participate sufficiently in conversations. To determine if participation in discussions between students was balanced within a CSCL script we analysed data from an implementation of the Pyramid CLFP. The data showed that student participation varied across different groups, with some groups and students not initiating conversation between them. Socially shared regulation is a social form of regulation where students regulate their learning as a group rather than as individuals through the means of discussion, negotiation, and perspective taking. As research has shown that for collaborative learning to be successful students need to engage in fruitful discussions and support each other to regulate their learning, we propose the implementation of a social awareness feature mirroring group participation through learning analytics in a CLFP tool (PyramidApp) to promote socially shared regulated learning.
This study focuses on motivating the need and presenting a design of an application of learning analytics to enhance social interactions in Computer Supported Collaborative Learning (CSCL). In particular, it proposes the design of a social awareness feature mirroring group participation to support student regulation in the context of scripted Pyramid-based collaborative learning flows.
Collaborative learning is an effective educational approach that has been shown to
enhance students' learning by allowing them to externalize their knowledge, monitor
their peers' learning and negotiate their opinions and ideas. However, when learners are
asked to collaborate without any instructions the results are not as effective as with
scaffolded collaboration [
Alternative or complementary approaches seeking effective collaborative learning
are related to supporting student’s regulation. Research has shown that collaborative
learning can be hindered by social dynamics as this type of learning situation poses
socio-emotional challenges and requires regulation of learning [
During individual learning, students internally regulate their thoughts and
behaviours to help them reach a goal. The theory of regulated learning addresses this process
and concerns how learners develop and effectively use learning skills [
Co-regulated learning describes a relationship between individuals where one person
is more knowledgeable or more skilled. Co-regulated learning is a dynamic
metacognitive process where individuals self-regulate their learning and share regulation of
their cognitions, emotions, behaviours, and motivations towards the aim of an academic
goal [
Group awareness was defined by Dourish and Bellotti (1992) [
There are three types of group awareness: Behavioural awareness refers to the
information about other members’ activities in the group, such as what they are doing
and what they will do later. Cognitive awareness refers to cognitive information,
including knowledge, beliefs or goals of the members in the group. Social awareness
refers to being aware of what other members of the group are doing or gathering
continuous information about them and acting accordingly [
One example of social awareness in CSCL is the Reflect table proposed by Bachour
et al (2008) [
Section 2 describes the need of supporting student regulation in the case of CSCL
scripts structured according to the Pyramid CLFP. Then, section 3 presents a design of
a social awareness feature aiming at facilitating socially shared regulation in Pyramid
App, an application enabling the design and enactment of Pyramid scripts [
PyramidApp is a web-based tool that facilitates the deployment of Pyramid pattern based CSCL activities where students can collaborate during several stages. Initially students are required to provide an individual answer to a given task. Then they are allocated into small groups where they can see the answers that were provided by their group members and rate them. In the next stage, the “answer improving” stage, students can discuss between them the given task and provide an improved answer. At the end of the small group level, students enter a second individual rating stage. Finally, the smaller groups are merged to formulate larger groups where they can further engage in discussions and vote for the best answers to reach a consensus at the end of the activity.
Log data from the PyramidApp tool was analysed to determine the discussion participation of students during Pyramid activities. In the following we first present students’ overall participation in the discussion in three Pyramid activities. Then in order to better understand the discussion participation differences of each student in each Pyramid activity we provide a detailed analysis. 2.1
The first column in Table 1 shows the activity ID. We considered three Pyramid activities conducted by a teacher online in a postgraduate course named “Learning Technologies”. During activity A, B, C teacher proposed open-ended knowledge sharing tasks for students. The third column represents the Pyramid group level, the fourth column provides an identifier for different groups and the fifth column shows the total number of students in each group. In the final column students’ chat participation is given which was calculated based on the percentage of students who wrote chat messages out of total students in a given group. As presented in Table 1, it can be seen that students’ discussion participation varied across different groups. Only GB2 and GC0 groups had 100% participation in the discussion meaning every student in the group participated in the discussion at least once. In the other groups however, the discussion participation varied across participants. It is noticeable that in GB1 and GC3 students had the least discussion participation.
The graphs under Fig. 1 present along the X-axis the number of chat messages posted by each student during the first and second rating levels of the Pyramid activities and the Y-axis represents students. As it can be seen in Fig.1(a) four students (S1-1 , S1-3, S1-6 and S1-7) only contributed to the first rating level and student S1-13 only participated in the second rating level. Noticeably S1-12 and S1-14 did not participate in discussion at any Pyramid level. In Activity B (see Fig. 1(b)) there is unequal participation in the discussion. For instance, it can be seen that seven students (S2-1, S2-2, S2-3, S24, S2-5, S2-6 and S2-7) mainly contributed and dominated the discussion while others have less participation or no participation at all in the discussion. In Activity C (see Fig. 1(c)) many students contributed to the discussion in the first rating level, whereas only two of them contributed to the discussion in the second rating level.
The analysis of the PyramidApp log data has shown that student’s participation varied
across different groups, reaching both extremes of 100% and 0%. As empirical
evidence indicates that SSRL has the potential of increasing group performance in
collaborative learning [
To calculate user participation, an algorithm counts the characters sent, edited, or deleted from the chat and the answer improvement text-editor. Then to compute the user’s participation the following algorithm has been proposed:
(User participation / (All participants average participation * 2)) * 100 This algorithm implies the following cases: at the beginning, all user’s participation will be 50% as the average will be 0, a user’s progress will be under 50% if their participation is under the average, and the user's progress will be at 100% if their participation is equal or greater than twice the average. See the notes of figure 2 for a detailed use case using the algorithm.
This study aimed to investigate students' participation in Pyramid pattern based scripted CSCL situations and explore regulated learning theories to help enhance participation in collaborative activities. Log data from PyramidApp were analysed to determine the distribution of students’ discussion participation within three classroom based CSCL situations. Results of the analysis showed that participation levels varied across different groups with a number of students not participating at all. The sample size that was used to analyze the data was small, however, as CSCL can be hindered by social dynamics and socio-emotional factors, we found it important to further investigate mechanisms that can reduce this effect. Thus, this paper further proposed the implementation of a social awareness feature within a pyramid CLFP to promote socially shared regulation by allowing group members to monitor each other’s contributions. However, its effectiveness has not yet been evaluated in practice and its future implementation could have some limitations. For instance, the design of the social awareness feature could potentially hinder its effectiveness as students might not become aware of its purpose. Additionally, the amount of information that appears on the page could possibly lead to an overload of their cognitive load, not allowing them to concentrate on the given task. Follow up research needs to be conducted to evaluate the design of the feature and its effectiveness in collaborative tasks. In contrast, students becoming aware of the tool’s functionality may cause the conversation to become a competition of who will send the most messages and derail the discussion from being productive. The current design of the platform prevents such behaviors from happening through a teacher monitoring system where teachers can intervene in the conversation. However, the feature’s limitation could be addressed in future work by implementing an algorithm to analyze the quality of contributions rather than word volume. Follow-up research will be conducted to determine if the implementation of the proposed feature enhances discussion participation within a pyramid CLFP. Further future work may also address how the feature affects dominant speakers' participation level and propose amendments to the design to help prevent dominant speakers from controlling and overpowering the conversation. The proposed design in this paper does not explicitly address this. Such a design needs to be carefully thought to not adversely affect the participation levels of dominant speakers. 5
We would like to acknowledge the help of Alexandre Argenté Pérez-Milá and Pascual Pablo Abenia Polo for assisting on the technical development of this study. This work has been partially funded by the Volkswagen Foundation in the framework of the project COURAGE (no. 95567) and by FEDER, the National Research Agency of the Spanish Ministry, TIN2017-85179-C3-3-R and MDM-2015-0502. D. Hernández-Leo also acknowledges the support of ICREA under the ICREA Academia programme.