Soutenance de thèse Christiana Tsiourti

 Artificial Agents as Social Companions: Design Guidelines for Emotional Interactions
Extended Abstract:
As the average age of the population in many countries increases, in the next
decades, we will see growing numbers of care-dependent older people suffering
from illness and disability. At the same time, there are growing concerns of a
severe global shortage in the supply of formally trained caregivers. The use of
robots and virtual characters, endowed with social and emotional intelligence,
presents an exciting opportunity to bridge the elderly-care gap. Socially
intelligent robots and virtual characters have the potential to play a major role as
“artificial companions” for the elderly, providing social and emotional support,
helping their owners to perform daily life activities, or assisting them to sustain
their fitness and health. Additionally, they may be able to provide additional
functionalities (i.e., safeguarding) to support formal and informal caregivers.
Much research has already been done on the design of artificial companions,
and existing results from the fields of Affective Computing, Human-Computer
Interaction (HCI) and Human-Robot Interaction (HRI) have already
demonstrated the beneficial role of this technology with respect to the health and
psychological well-being of the elderly. Nevertheless, so far, existing prototypes
of artificial companions still lack important social abilities and emotional skills
that enable them to establish “truly” natural , engaging and long-lasting
relationships with older adults. Moreover, there are no definitive guidelines to aid
designers and researchers in the process of creating robust, user-friendly,
empathetic, artificial companions.

This dissertation builds upon previous work and contributes to the design of
artificial companions endowed with social and emotional intelligence and
communication skills. This work is based on 5 scientific articles and is structured
in three parts. The aim of the first part is to contextualise the use of artificial
companions within elderly-care. Article 1 reports a user-centered design study,
conducted to clarify several aspects related to the perception and acceptance of
artificial companions by healthy elderly individuals and formal caregivers.
Different questions related to user characteristics, care needs and potential
applications of an artificial companion, the companion’s appearance and
 communication skills, as well as ethical and privacy considerations, were
addressed in a mixed-method design study. A fully autonomous artificial
companion prototype was designed and implemented according to these
findings. Article 2 reports a longitudinal evaluation study, carried out in two
different European countries (i.e., Switzerland and the Netherlands) to examine
how healthy elderly individuals interact with an artificial companion integrated
into their domestic environment. This study explored how older adults evaluate
the companion in terms of acceptance, perceived usability, and usefulness in
realistic scenarios covering various types of interaction and tasks that occur in
daily life.

 The second and third part of this dissertation is specifically focused on two
important building blocks of artificial companions: affect sensitivity , i.e., the
capability of a companion to sense, process and interpret information concerning
the user’s mental and affective states , and affect expressivity , i.e., the capability
of a companion to communicate its internal emotional state and intentions in a
believable and clear way. Article 3, investigates the design of an affect sensing
and recognition framework for artificial companions and proposes an approach
based on the assessment of psychophysiological changes (i.e., cardiac activity,
electrodermal activity, respiration) that reflect changes in the autonomic nervous
system (ANS) functioning. To this end, we surveyed 173 publications reporting
research studies on ANS activity in emotion and identified 15 wearable sensor
systems used to monitor physiological signals for the purpose of affect
assessment. Articles 4 and 5 are dedicated to the design of an affect expression
framework for artificial companions. Article 4 explores whether people can
recognize nonverbal emotional cues expressed by humanoid robots and
whether the robot’s embodiment affects this recognition. We designed facial,
vocal, and body emotional expressions for Pepper, a highly humanlike robot,
and Hobbit, a robot with abstract humanlike features. In an online survey, we
asked participants to watch videos with the robots and evaluate their emotional
expressions qualitatively and quantitatively. Article 5 examines how people
recognize and respond to emotions displayed multimodally, by the body and
voice of humanoid robots, with a particular emphasis on the effects of
incongruence. In a social HRI laboratory experiment, we investigated contextual
incongruence (i.e., the conflict situation where there is no match between the
social situation and the emotional behaviour of the robot) and cross-modal
incongruence (i.e., the conflict situation in which there is no match between the
emotional information expressed from a robot’s body and the robot’s voice).

 Overall, the empirical and theoretical body of work conducted within this
dissertation offers new perspectives on the design of artificial companions and
makes several research contributions that are of interest to the greater HCI, HRI,
and Affective Computing communities. Firstly, our findings contribute to a better
understanding of older adults' needs and requirements for the development of
artificial companions and reveal factors that shape the acceptance of this
technology in domestic elderly care environments. Secondly, we derived
methodological considerations, and practical guidelines to facilitate the use of
the psychophysiological approach for the design of affect sensing and
recognition frameworks for companion systems. Thirdly, and lastly, this work
contributes to the general understanding of how people respond to human-like
emotional cues enacted by robots within the scope of social interaction
scenarios and provides findings that have implications for the design of artificial
companions that use serval channels to communicate emotion and intentionality
in a clear and effective way.

 The findings of this work are summarized in the form of explicit guidelines and
recommendations that serve as a straightforward design tool-box for future HCI
and HRI researchers and designers of artificial companions. Although the work
described in this dissertation focuses on the domain of elderly care, the
development of artificial companions endowed social and emotional intelligence
and communication skills is relevant to a growing number of related fields which
require systems that can interact socially with humans, such as assistive
robotics in the care for disabled and chronically ill people, as well as virtual
agents and robots in education, and the general domains of home assistance
and entertainment.
Date: 22 Jun 2018
Lieu:  CUI, Battelle A-Rez (Route de Drize 7, 1227 Carouge), 10.00h
June 22, 2018