Abstract:

In our research we have explored how a teacher and his group of teacher students interact during their group discussions. The purpose was to investigate what kind of interaction students experience as being the most dialogical. By dialogicality we mean the intuitive concept that means to a great extent the reciprocity of interaction, responding to the other as a whole person and giving space to the other to tell his thoughts using his own language, concepts, and interpretive schemes. In this study, dialogicality is based on the subjective assesments of the informants. Thus, the concept had to be defined in an intuitive way. Dialogicality as well as cognitively high-level discussion and knowledge co-creation are typically part of productive interaction, which in turn is often regarded as a prerequisite for successful collaborative learning.

The detailed research questions were:
a) How do emotions, the network structure of the conversation and the quantity and topics of utterances reflect dialogicality? Or do they play any role at all?
b) Can we discover in the data other issues that may co-occur with a high level of dialogicality?

To gain insight into the above research questions, we gathered roughly nine hours of video data from three groups and 11 group discussions involving 16 people.

The data collection was conducted in 2019 and 2020 by the School of Vocational Teacher Education of t Haaga-Helia University of Applied Sciences. This data has been analyzed through:
a) network analysis,
b) natural language processing technologies such as sentiment analysis, and
c) facial expression analysis using the AFFDEX algorithm that is implemented in the iMotions software.

In this paper, we present in detail the findings concerning one team and one interaction video. We concentrate on two conversation snippets: the most and the least dialogical ones.

For the sentiment analysis, a lexicon-based method was used. In the network analysis, standard metrics, such as degree, closeness, betweenness and eigenvector centrality were used. To analyze the emotional reactions of the participants, we used the iMotions software and, more precisely, the AFFDEX algorithm. AFFDEX is based on discrete emotion theory, which assumes seven basic emotions: joy, anger, sadness, fear, contempt, disgust, and surprise. In addition, AFFDEX measures the engagement of the participants.

The study shows that there is a relation between dialogicality as perceived by a participant and engagement and amount of speech produced in the discussion as well as emotional reactions as measured from facial expressions and linguistic features in transcribed speech. The participants ranked the most dialogical phases the segments that were all-involving, interactionally active, often emotionally loaded, and focused more on shared phenomena than on the individual oneself. An interaction where all students participate and where a teacher is speaking less is perceived as being the most dialogical. In addition, a correlation between the use of positive words and perceived dialogicality was found. Future research on a larger and more complete dataset will show if these findings can be generalized. If so, the results will be useful for any teacher or team leader who wishes to foster successful group interaction.

Keywords:

Group interaction, dialogicality, facial emotion recognition, sentiment analysis, network analysis, collaborative learning.