Backchannel behavior in child-caregiver conversations

Keywords
Loading...
Thumbnail Image
Issue Date
2022-08-06
Language
en
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Conversation is a coordinative activity (Clark, 1996) that requires cooperative social interaction between interlocutors. The coordination nature of conversations has been the hallmark of children’s socio-cognitive development as it involves the sophisticated ability to manage the flow of conversation through backchanneling i.e., signaling listener’s attention through verbal (short responses like Yeah) and non-verbal behaviors (e.g. smiling, nodding). Previous studies on middle childhood children’s backchannel behaviors were scarce and either conducted in highly controlled experimental settings (Hess and Johnson, 1988) or qualitative observation (Bodur et al., 2022). This thesis investigated middle-childhood children’s production and responses of multimodal cues to elicit backchannels in naturalistic childcaregiver conversations (ChiCo: Bodur et al., 2021). In order to quantitatively infer the potential combinations of multimodal cues, two backchannel opportunity prediction models(Support Vector Machine model: SVM; Long short-term memory model: LSTM) were first trained respectively on children and adults’ responses to the speaker cues in child-caregiver and adult-adult conversations and then tested on different input feature modalities. The comparable model performance between children and adults indicated that children between the ages of 6 and 12 can produce and respond to backchannel inviting cues as consistently as adults. The comparison of model input modalities suggested that features from vocal modality contributed most to backchannel occurrences, which may be caused by characteristics of child-directed speech and caregivers’ linguistic alignment to scaffold children’s language processing. In a word, although school-age children are still at the stage of developing sociocognitive competencies, their performance of producing and responding to BC inviting cues are strikingly close to adult-level mastery. The broader impact of this thesis lies in the application of machine learning models in balancing the needs of ecologically valid settings and quantitative analysis.
Description
Citation
Faculty
Faculteit der Letteren