Humanoid-based protocols to study social cognition

Kompatsiari, Kyveli (2020): Humanoid-based protocols to study social cognition: on the role of eye contact in joint attention. Dissertation, LMU München: Graduate School of Systemic Neurosciences (GSN)

Abstract

Social cognition is broadly defined as the way humans understand and process their interactions with other humans. In recent years, humans have become more and more used to interact with non-human agents, such as technological artifacts. Although these interactions have been restricted to human-controlled artifacts, they will soon include interactions with embodied and autonomous mechanical agents, i.e., robots. This challenge has motivated an area of research related to the investigation of human reactions towards robots, widely referred to as Human-Robot Interaction (HRI). Classical HRI protocols often rely on explicit measures, e.g., subjective reports. Therefore, they cannot address the quantification of the crucial implicit social cognitive processes that are evoked during an interaction. This thesis aims to develop a link between cognitive neuroscience and human-robot interaction (HRI) to study social cognition. This approach overcomes methodological constraints of both fields, allowing to trigger and capture the mechanisms of real-life social interactions while ensuring high experimental control. The present PhD work demonstrates this through the systematic study of the effect of online eye contact on gaze-mediated orienting of attention. The study presented in Publication I aims to adapt the gaze-cueing paradigm from cognitive science to an objective neuroscientific HRI protocol. Furthermore, it investigates whether the gaze-mediated orienting of attention is sensitive to the establishment of eye contact. The study replicates classic screen-based findings of attentional orienting mediated by gaze both at behavioral and neural levels, highlighting the feasibility and the scientific value of adding neuroscientific methods to HRI protocols. The aim of the study presented in Publication II is to examine whether and how real-time eye contact affects the dual-component model of joint attention orienting. To this end, cue validity and stimulus-to-onset asynchrony are also manipulated. The results show an interactive effect of strategic (cue validity) and social (eye contact) top-down components on the botton-up reflexive component of gaze-mediated orienting of attention. The study presented in Publication III aims to examine the subjective engagement and attribution of human likeness towards the robot depending on established eye contact or not during a joint attention task. Subjective reports show that eye contact increases human likeness attribution and feelings of engagement with the robot compared to a no-eye contact condition. The aim of the study presented in Publication IV is to investigate whether eye contact established by a humanoid robot affects objective measures of engagement (i.e. joint attention and fixation durations), and subjective feelings of engagement with the robot during a joint attention task. Results show that eye contact modulates attentional engagement, with longer fixations at the robot’s face and cueing effect when the robot establishes eye contact. In contrast, subjective reports show that the feeling of being engaged with the robot in an HRI protocol is not modulated by real-time eye contact. This study further supports the necessity for adding objective methods to HRI. Overall, this PhD work shows that embodied artificial agents can advance the theoretical knowledge of social cognitive mechanisms by serving as sophisticated interactive stimuli of high ecological validity and excellent experimental control. Moreover, humanoid-based protocols grounded in cognitive science can advance the HRI community by informing about the exact cognitive mechanisms that are present during HRI.

• Dokumententyp: Dissertationen (Dissertation, LMU München)
• Themengebiete: 500 Naturwissenschaften und Mathematik ; 500 Naturwissenschaften und Mathematik > 570 Biowissenschaften, Biologie
• Fakultäten: Graduate School of Systemic Neurosciences (GSN)
• Sprache der Hochschulschrift: Englisch
• Datum der mündlichen Prüfung: 28. Februar 2020
• 1. Bericht­erstatter:in: Wykowska, Agnieszka