Ginevra Castellano is a Professor in Intelligent Interactive Systems at the Department of Information Technology, Uppsala University, where she leads the Uppsala Social Robotics Lab. Her research interests are in the areas of social robotics and affective computing, and include social learning, personalized adaptive robots, multimodal behaviours and uncanny valley effect in robots and virtual agents. She has published more than 100 research papers on these topics, receiving over 3000 citations. She was the coordinator of the EU FP7 EMOTE (EMbOdied-perceptive Tutors for Empathy-based learning) project (2012–2016). She is the recipient of a Swedish Research Council starting grant (2016–2020) and PI for Uppsala University of the EU Horizon 2020 ANIMATAS (Advancing intuitive human-machine interaction with human-like social capabilities for education in schools; 2018-2021) project, the COIN (Co-adaptive human-robot interactive systems) project, funded by the Swedish Foundation for Strategic Research (2016–2021), and the project "The ethics and social consequences of AI & caring robots. Learning trust, empathy, and accountability" (2020-2024), supported by the Marianne and Marcus Wallenberg Foundation, Sweden. Castellano was a general co-chair at IVA 2017. She is an Associate Editor of Frontiers in Robotics and AI, Human-Robot Interaction section, and IEEE Transactions in Affective Computing. Castellano is a recipient of the 10-Year Technical Impact Award at the ACM International Conference on Multimodal Interaction 2019.

 
Talk Title: Investigating People’s Affective Responses to Robots
 
The recently published EU guidelines for trustworthy AI warn of the need to study
the social impact of AI systems integrated in all areas of our lives.
AI systems such as robots in socially assistive roles can impact people’s social relationships and attachment,
so one important area of investigation is the study of how people emotionally respond to robots.
One question that I address in my research is how a robot’s human-likeness in appearance and behaviour
influences people’s affective perceptions and affective responses to robots.
I present two studies investigating people’s affective perceptions and affective responses to robots displaying emotional and empathic behaviour.
I show that, in experiments with a mixed embodiment robot, human likeness seems to affect high and low level affective responses
in different ways. I also show that people are able to understand human-like and robot-specific empathic traits in a robot.
This work contributes to addressing a key question for ethical and trustworthy AI and robotics: when do people empathise and socially connect with robots?
 

Ellen Souza is a Professor at the Federal Rural University of Pernambuco, in Brazil and a postdoctoral researcher at the Institute of Mathematics and Computer Sciences – ICMC/USP. She has a PhD in Computer Science from the Federal University of Pernambuco. Her research interests include digital inclusion, socially assistive robotics, text mining, and artificial intelligence.

Talk Title: Assistive Robot for Training Emotional Facial Expressions to People with the Symptoms of Alexithymia.

Alexithymia is a disorder characterized by the inability to identify and describe emotions in the self and by the difficulty to distinguish and appreciate the emotions of others. Alexithymia is prevalent in approximately 10% of the general population and is known to be comorbid with a number of psychiatric conditions. It also affects more than 50% of the population with Autism Spectrum Disorder (ASD). A common treatment for this problem is the use of behavioral analysis, coupled with occupational and speech therapies. Therapists often concentrate on building a foundation of naming emotions and appreciating a range of feelings through games, drawings, and stories. In this talk, I’ll present a different approach to training emotional facial expressions to people with the symptoms of Alexithymia using assistive robots. The solution comprises an artificial intelligence that uses a Deep Neural Network to capture the face of the person and classify it into one of the six universal facial emotions. More than 65 robot-assisted therapy sessions were performed by a multidisciplinary team composed of a psychologist, an occupational therapist, and a speech therapist.

Yukie Nagai is a Project Professor at the International Research Center for Neurointelligence, the University of Tokyo. She received her Ph.D. in Engineering from Osaka University in 2004 and worked as a Postdoc Researcher at the National Institute of Information and Communications Technology (NICT) and at Bielefeld University for five years. She then became a Specially Appointed Associate Professor at Osaka University in 2009 and a Senior Researcher at NICT in 2017. Since April 2019, she leads Cognitive Developmental Robotics Lab at the University of Tokyo.
Yukie Nagai has been investigating underlying neural mechanisms for social cognitive development by means of computational approaches. She designs neural network models for robots to learn to acquire cognitive functions such as self-other cognition, estimation of others’ intention and emotion, altruism, and so on. Her computational studies suggest that the theory of predictive coding provides a unified account for both the continuity and diversity of cognitive development.

Talk Title: Development of Perception and Learning Based on Predictive Coding

A theoretical framework called predictive coding has been attracting increasing attention in developmental robotics as well as in neuroscience. It suggests that the human brain works as a predictive machine that tries to minimize prediction errors. We have been investigating how children acquire the ability of predictive processing and what causes individual diversities in the development. This talk presents two studies that examined the development of perception and learning in representational drawing.

he first experiments employed a computational neural network based on predictive coding. Our experiment revealed that aberrant precisions in prior knowledge caused difficulties in perception and thus learning. Neural networks with hypo- or hyper-priors exhibited poor generalization ability or weak adaptability whereas networks with moderate priors successfully acquired drawing behavior. The second experiment analyzed children’s drawings to verify our findings from the computational study. Our results demonstrated that young children had undifferentiated prior representations and/or limited capability to adapt to input stimuli. The ability to draw target objects gradually improved with age. The immaturity observed in younger children was analogous to the difficulties caused by aberrant priors in neural networks. I discuss how this finding can be extended to account for affective perception based on predictive coding.