Poster #2 - Robotese and Motherese: Examining Patterns of Attention Preference in Infants and Toddlers in Eye-Tracking Paradigms

• In Event: Poster Session 02

Abstract

Introduction:
Infants and toddlers innately respond to infant-directed speech (IDS), or “motherese,” supporting early responsiveness to joint attention, emotions, and language input (Filipe et al., 2018; Saint-Georges et al., 2013). Young children typically demonstrate an early preference for IDS, however children with autism spectrum disorder risk often do not (Filipe et al., 2018; Kuhl et al., 2005). Rather, autistic preferential attention is consistently allocated to non-social stimuli and non-human social partners such as robots (Chevalier et al., 2020; Zheng et al., 2018). Relatedly, interactions with robots have been characterized as eliciting increased turn-taking, more frequent joint attention, and increased language use (David et al., 2020; Chevalier et al., 2020). However, specific aspects of robots that elicit heightened social interest remain to be understood. Thus, directly comparing child social attention to motherese versus robotese (e.g., the speech and gestural qualities of robots) may yield novel insight regarding such preferences.

Objective:
The current study utilized eye-tracking paradigms to compare visual attention patterns of infants and toddlers to contrasting, dynamic social stimuli that demonstrated both motherese and robotese conditions.

Methods:
Participants included a sample of typically developing infants and toddlers (n=16) age 7-17 months (+/-2 weeks). Consistent with prior eye-tracking empirical studies with young children, we measured gaze fixation, duration, and total looking time in the context of novel video stimuli. Child developmental profiles were obtained using the Mullen Scales of Early Learning (Mullen, 1995) and the Early Social Communication Scales (Mundy et al., 2013).

Results:
A paired sample Wilcoxon test revealed no significant difference in initial fixation duration between robotese and motherese conditions (z= -.025, p=.959). Paired sample t-testing also indicated no significant between conditions difference for total fixation duration (r = .198, p=.440). See Figure 1 for still frames from exemplar stimuli including both motherese and robotese cues; gaze plot trends with simultaneous social agents (i.e., dump truck condition) confirmed child allocation of visual attention in a triadic manner. Two-way ANOVAs confirmed significant interaction effects (see Figure 2) between variable agent type (robotese or motherese) and areas of interest (social or non-social) across all clips: Brown Bear (F(1, 15) =9.88, p=.007), Itsy-Bitsy–Spider (F(1, 15) =10.83, p=.005), and Peek-A-Boo (F(1, 15) =31.56, p<.001).

Conclusion:
Findings from the current study provide insight regarding infant attention and social gaze pattern preferences in the context of motherese and novel robotese. More specifically, our results elucidate the feasibility and social relevance of implementing robotic paradigms with populations of very young typically developing children. Evaluation of differences in gaze allocation across robotese and motherese conditions did not yield significant preferential patterns, however trends indicated preferential attention towards robotese when specific activities (i.e., Peek-A-Boo, Itsy-Bitsy-Spider, Dump Truck) were observed. Further, these findings corroborate previous research studies indicating that young children may view the robot as a social agent, similarly to how they perceive human agents. We urge future studies investigate conditional responses to changes in vocal, gestural, and physical qualities of robots, as well as how preferential and conditional responses vary with child age and developmental trajectories.

Authors

• Gabriella M Garcia, Texas Tech University
  Presenting Author

• Miranda Cox, Texas Tech University

• Jessica Elaine Blume, Texas Tech University

• Marianne Garcia, Texas Tech University

• Ann M Mastergeorge, Texas Tech University