Research On Collaborative Problem-Solving Using Multimodal Data

Collaborative problem-solving is an important way to promote personal cognitive development and social development in the information society,and it is also a basic core quality of personnel training.It promotes learners to develop a deep understanding of knowledge,thereby forming transferable collaborative problem-solving skills.How to provide learners with an effective collaborative problemsolving experience has become an important topic in the current educational research field.Currently,studies on learners’ collaborative problem-solving learning process mainly rely on questionnaires and diary data,which limits the validity of the findings.With the development of the Internet of Things and sensing technology,it has become possible to reveal the learning process of learners in a multidimensional and fine-grained manner.For example,the use of portable EEG devices to collect EEG data during learning can accurately measure learners’ cognitive load,engagement status,and so on.Learning is a dynamic activity,tracking,encoding,and processing multimodal data can reveal the complexities of the learning process at the micro-level,thus enabling researchers to gain a systematic and comprehensive understanding of the nuances among learners.Learning analytics based on multimodal data allows researchers to observe evidence on whether complex pedagogies are achieving desired outcomes.Therefore,this research is based on empirical research and driven by data.We collected multimodal data including EEG data,questionnaire data,knowledge testing,audio,and video recording,and online log data to systematically study the process of learners’ collaborative problem-solving.The characteristics and interaction of learners’ cognitive load,attention,and learning performance in collaborative problem-solving learning are studied,including the following four aspects.First,a study on learners’ cognitive load in collaborative problem-solving.To analyze the characteristics of learners’ cognitive load in the process of collaborative problem-solving,identify the key factors that affect the level of learners’ cognitive load,and explore how to guide learners to support complete teaching activities in the process of collaborative problem-solving.Results show there was a significant negative correlation between learners’ cognitive load and learning performance during collaborative problem-solving.The cognitive load level of learners in the problem conceptualization task was significantly higher than that in the problem-solving task.The most important variables among the decision tree for students’ cognitive load in problem conceptualization is their prior CPS skill,and the second is basic knowledge.The first two most important variables among the decision tree for students’ cognitive load in problem-solving are their cognitive load in the problem conceptualization tasks and CPS skills.Second,a study on learners’ attention in collaborative problem-solving.This paper probes into the collaboration pattern and explores the attention state transition and evolution rules of group members in different collaboration patterns.Results show in the process of collaborative problem-solving,there are significant differences in the attention levels of group members who exhibit "independent"collaborative patterns,"centralized" collaborative patterns,and "distributed" collaborative patterns in the process of collaborative problem-solving.In the "independent" collaborative patterns,students’attention can easily deviate from the work in progress,and it is difficult to return to the state of concentration.In the "distributed" collaborative patterns,highly attentive students tended to have a higher chance of transitioning to distraction in the "centered" collaborative patterns.Third,a study on learners’ interaction behavior in collaborative problem-solving.This paper probes into the internal adjustment mechanism of groups composed of different members,that is,how the composition of group members affects learners’ interaction paths.Results show there are significant differences in the distribution of learners’ interaction behaviors among the homogeneous high group,homogeneous low group,and heterogeneous group.The learners of the heterogeneous group are active and in-depth in the process of communication and understanding,and this in-depth communication can be maintained for a long time.Fourth,a study on learners’ cognitive engagement in collaborative problem-solving.This paper explores the differences in learners’ cognitive engagement in different collaborative problem-solving learning performances and identifies the key factors that affect learners’ cognitive engagement levels.Results show learners show higher cognitive participation in problem-solving tasks;For the successful group that has completed the collaborative problem-solving task,the level of learners’ cognitive engagement is related to their internal motivation,while for the failed group that has not completed the collaborative problem-solving task,the level of learners’ cognitive engagement level depends on their sense of self-efficacy.In general,driven by multimodal data,this research uses multimodal data collection and analysis technology,machine learning technology,and other related achievements to complete the identification of the explicit behavior and implicit learning state of learners and collaborative groups in the process of collaborative problem-solving,and realized the systematic research on the cognitive load,attention level,interaction sequence and cognitive engagement of learners in collaborative problem-solving.This research will further promote a comprehensive understanding of the state and evolution of individual learners and collaborative groups in the collaborative problem-solving learning process.