This paper presents a dataset of a collaborative learning situation. Students were enrolled in two undergradute courses on computer networks where they were required to carry out a set of learning activities supported by Moodle and an online collaborative environment called CoTrackV2. The data collected includes logs of the writing process of shared documents, logs of the chat messages between the group members, and logs from Moodle with coarser-grained information about course-level interactions. This dataset has been generated with the aim of allowing researchers to study selfand socially-shared regulation in online environments.
Academic and work contexts are increasingly demanding the
competence of being able to collaborate with peers [
The literature has shown that students' motivation and
strategic regulation play a critical role in their success in Science,
Technology, Engineering and Mathematics (STEM) courses
[
Type of operation (>: writing,
<: deleting) The di erence in number of characters caused by the current action and source length
Have you contacted your Internet service provider, i.e. your operator? [...] The learning situation took place at two undergraduate courses on Computer Networks during 4 days in the spring semester of the academic year 2021 in a European University. There were 33 students, that were grouped into 8 di erent groups of 4-5 people to carry out an introductory learning situation aimed at challenging their previous knowledge and beliefs about certain computer network topics. Before starting the learning situation, students were asked to ll out an informed consent.
The situation was designed following the so-called pyramid or snowfall pattern, where the students had to rst carry out the proposed activities individually and then in groups (thus fostering the agreement among the group members in order to submit a common solution). The di erent activities were carried out during 4 two-hour face-to-face sessions. The learning situation was based on the following scenario: A hotel owner (role played by the teacher) goes to a team of telco engineers (role played by the students) to ask them to solve his problem: the internet connection is not working properly; the internet access is very slow and sometimes does not work at all. The hotel owner and the telco engineers agree to an interview in a few days. In order for the telco engineers to think about the problem, the hotel owner sends them a diagram of the current network. The di erent activities that students needed to complete were: Questions ind (individual): Thinking of questions to ask the hotel owner to nd out more about his network. Questions group (in groups of 4-5 students): Agreeing on 7 nal questions to ask the hotel owner.
Questions class (whole class): Asking the hotel owner about his network. For this task, there was a spokesperson in each group. The teacher, playing the role of the hotel owner, answered those questions posed by the groups.
Diagnosis ind (individual): Proposing a solution to the hotel's Internet access problem.
Diagnosis group (in groups of 4-5 students): Agreeing on a nal proposal with the rest of the group members. Diagnosis class (whole class): Creating a concept map of the technical concepts that emerged during the whole situation.
Students had to work through an online collaborative environment called CoTrackV21. This environment o ered the possibility to write documents collaboratively and had a built-in chat so that the di erent members of the group could communicate. In addition, students used Moodle to submit individual assignments, to visit subject-related content and to access the link to the CoTrackV2 sessions, so we were able to obtain traces of the content visited by the students, the writing process and the chat messages. Besides these traces, at the end of the learning situation, the students answered a questionnaire related to group regulation.
The dataset collected2 is based on the logs of two di erent tools: CoTrackV2 and Moodle. The data obtained by CoTrackV2 is divided into 2 les: 1) document_logs.csv, with actions from the writing process of the shared documents; and 2) chat_logs.csv, that contains the logs of the communication between the group members. The attributes in the two les are presented in Table 1, and Table 2, respectively. Regarding the data obtained through Moodle, we have 2 les: 1) moodle_logs.csv, with the logs of the contents visited by the students. Details are given in Table 33; and 2) individual_submissions.csv, where the individual submissions for activities 1 and 4 are collected, containing the timestamp of the submission, the id of the student submitting the solution and the solution itself. Besides these les, we have two others: 1) a le containing the learning design, including the start time and the name of the tasks; and 2) a le containing the students' answers to the nal questionnaire. All les provided have been properly anonymized. 1CotrackV2 website: https://www.cotrack.website/ 2The dataset will be available at https://zenodo.org/record/5033198#.YNsQv-gzaUk 3The examples presented in the di erent tables have been translated into English for a better understanding, but the dataset is in Spanish.
The dataset we have generated may allow researchers to answer di erent research questions related to Socially-Shared Regulation of Learning (SSRL). For example, the research questions that have guided the design of this learning situation are the following: 1) How do self- and socially-shared regulation processes occur in groups that complete group activities with di erent levels of success?; 2) Are there di erent patterns of regulation associated with the performance of groups when solving activities? To answer these questions, we want to analyse the data from a temporal perspective using di erent techniques, like process mining (e.g.: Heuristic Miner or Fuzzy Miner algorithms), Markov models (e.g.: pMiner algorithm), social network analysis (temporal networks) and epistemic network analysis. Beforehand, we want to identify SSRL features that allow us to map lowlevel data to higher-level constructs. After that, we could make use of the techniques mentioned above and compare the results of the di erent approaches. Beyond detecting the di erent processes, we would like to build predictive models with the identi ed features. However, at this stage of the research, it would be very bene cial to get feedback from the community to better guide the analysis.
This research is partially funded by the European Regional Development Fund and the Spanish National Research Agency under project grant TIN2017-85179-C3-2-R.