Showing papers by "Ana Paiva published in 2014"

FAtiMA Modular: Towards an Agent Architecture with a Generic Appraisal Framework

Abstract: This paper presents a generic and flexible architecture for emotional agents, with what we consider to be the minimum set of functionalities that allows us to implement and compare different appraisal theories in a given scenario. FAtiMA Modular, the architecture proposed is composed of a core algorithm and by a set of components that add particular functionality (either in terms of appraisal or behaviour) to the architecture, which makes the architecture more flexible and easier to extend.

Empathic Robots for Long-term Interaction

Abstract: As a great number of robotic products are entering people’s lives, the question of how can they behave in order to sustain long-term interactions with users becomes increasingly more relevant. In this paper, we present an empathic model for social robots that aim to interact with children for extended periods of time. The application of this model to a scenario where a social robot plays chess with children is described. To evaluate the proposed model, we conducted a long-term study in an elementary school and measured children’s perception of social presence, engagement and social support.

Teachers' views on the use of empathic robotic tutors in the classroom

Abstract: In this paper, we describe the results of an interview study conducted across several European countries on teachers' views on the use of empathic robotic tutors in the classroom. The main goals of the study were to elicit teachers' thoughts on the integration of the robotic tutors in the daily school practice, understanding the main roles that these robots could play and gather teachers' main concerns about this type of technology. Teachers' concerns were much related to the fairness of access to the technology, robustness of the robot in students' hands and disruption of other classroom activities. They saw a role for the tutor in acting as an engaging tool for all, preferably in groups, and gathering information about students' learning progress without taking over the teachers' responsibility for the actual assessment. The implications of these results are discussed in relation to teacher acceptance of ubiquitous technologies in general and robots in particular.

Improving social presence in human-agent interaction

Abstract: Humans have a tendency to consider media devices as social beings. Social agents and artificial opponents can be examined as one instance of this effect. With today's technology it is already possible to create artificial agents that are perceived as socially present. In this paper, we start by identifying the factors that influence perceptions of social presence in human-agent interactions. By taking these factors into account and by following previously defined guidelines for building socially present artificial opponents, a case study was created in which a social robot plays the Risk board game against three human players. An experiment was performed to ascertain whether the agent created in this case study is perceived as socially present. The experiment suggested that by following the guidelines for creating socially present artificial board game opponents, the perceived social presence of users towards the artificial agent improves.

Learning by appraising: an emotion-based approach to intrinsic reward design

Abstract: In this paper, we investigate the use of emotional information in the learning process of autonomous agents. Inspired by four dimensions that are commonly postulated by appraisal theories of emotions, we construct a set of reward features to guide the learning process and behaviour of a reinforcement learning (RL) agent that inhabits an environment of which it has only limited perception. Much like what occurs in biological agents, each reward feature evaluates a particular aspect of the (history of) interaction of the agent history with the environment, thereby, in a sense, replicating some aspects of appraisal processes observed in humans and other animals. Our experiments in several foraging scenarios demonstrate that by optimising the relative contributions of each reward feature, the resulting “emotional” RL agents perform better than standard goal-oriented agents, particularly in consideration of their inherent perceptual limitations. Our results support the claim that biological evolutionary adaptive mechanisms such as emotions can provide crucial clues in creating robust, general-purpose reward mechanisms for autonomous artificial agents, thereby allowing them to overcome some of the challenges imposed by their inherent limitations.

Guest Editorial: Emotion in Games

Abstract: The articles in this special section focus on the use of emotion technology and applications in computer games.

Context-Sensitive Affect Recognition for a Robotic Game Companion

Abstract: Social perception abilities are among the most important skills necessary for robots to engage humans in natural forms of interaction. Affect-sensitive robots are more likely to be able to establish and maintain believable interactions over extended periods of time. Nevertheless, the integration of affect recognition frameworks in real-time human-robot interaction scenarios is still underexplored. In this article, we propose and evaluate a context-sensitive affect recognition framework for a robotic game companion for children. The robot can automatically detect affective states experienced by children in an interactive chess game scenario. The affect recognition framework is based on the automatic extraction of task features and social interaction-based features. Vision-based indicators of the children’s nonverbal behaviour are merged with contextual features related to the game and the interaction and given as input to support vector machines to create a context-sensitive multimodal system for affect recognition. The affect recognition framework is fully integrated in an architecture for adaptive human-robot interaction. Experimental evaluation showed that children’s affect can be successfully predicted using a combination of behavioural and contextual data related to the game and the interaction with the robot. It was found that contextual data alone can be used to successfully predict a subset of affective dimensions, such as interest toward the robot. Experiments also showed that engagement with the robot can be predicted using information about the user’s valence, interest and anticipatory behaviour. These results provide evidence that social engagement can be modelled as a state consisting of affect and attention components in the context of the interaction.

Towards dialogue dimensions for a robotic tutor in collaborative learning scenarios

Abstract: There has been some studies in applying robots to education and recent research on socially intelligent robots show robots as partners that collaborate with people. On the other hand, serious games and interaction technologies have also proved to be important pedagogical tools, enhancing collaboration and interest in the learning process. This paper relates to the collaborative scenario in EMOTE EU FP7 project and its main goal is to develop and present the dialogue dimensions for a robotic tutor in a collaborative learning scenario grounded in human studies. Overall, seven dialogue dimensions between the teacher and students interaction were identified from data collected over 10 sessions of a collaborative serious game. Preliminary results regarding the teachers perspective of the students interaction suggest that student collaboration led to learning during the game. Besides, students seem to have learned a number of concepts as they played the game. We also present the protocol that was followed for the purposes of future data collection in human-human and human-robot interaction in similar scenarios.

Werewolves, cheats, and cultural sensitivity

Abstract: MIXER (Moderating Interactions for Cross-Cultural Empathic Relationships), which applies a novel approach to the education of children in cultural sensitivity. MIXER incorporates intelligent affective and interactive characters, including a model of a Theory of Mind mechanism, in a simulated virtual world. We discuss the relevant pedagogical approaches, related work, the underlying mind model used for MIXER agents as well as its innovative interaction interface utilising a tablet computer and a pictorial interaction language. We then consider the evaluation of the system, whether this shows it met its pedagogical objectives, and what can be learned from our results.

Disaster Prevention Social Awareness: The Stop Disasters! Case Study

Abstract: Serious games are increasingly being regarded as a valuable means to support educative processes and make people aware of important issues. Some of these address social awareness domains with the goal of promoting collective recognition of a given local or global issue as a first step towards its resolution. Even though nowadays widely used, serious games still require further study with regard to their impact. This gap in studies is especially strong in the domain of social awareness. The goal of this work is to address this gap by studying the impact of a highly disseminated serious game that had not yet been evaluated in any way: Stop Disasters!. We studied the impact of the game both in terms of the awareness and player experience it created. As a result we found statistical evidence to a positive impact of the serious game in player's awareness to wildfire prevention measures achieved in an overall positive and valuable game experience. This result provides further evidence to the positive impact of social awareness serious games in transmitting messages regarding social relevant issues in an overall positive and engaging experience.

Developing Interactive Embodied Characters Using the Thalamus Framework: A Collaborative Approach

Abstract: We address the situation of developing interactive scenarios featuring embodied characters that interact with users through various types of media easily presents as a challenge. Some of the problems that developers face are on collaborating while developing remotely, integrating all the independently developed components, and incrementally developing a system in such way that the developed components can be used since their incorporation, throughout the intermediate phases of development, and on to the final system. We describe how the Thalamus framework addresses these issues, and how it is being used on a large project that targets developing this type of scenarios. A case study is presented, illustrating actual development of such scenario which was then used for a Wizard-of-Oz study.

I'm the mayor: a robot tutor in enercities-2

Abstract: We are addressing the problem of creating empathic robot tutors to support school students studying geography topics on a multi-touch table. A multi-role serious game Enercities-2 has been developed from an earlier single-user version in which a Mayor, Economist and Environmentalist have control over differing resources. The game explores the tension between individual and collaborative success. A robot tutor, embodied as a NAO Torso robot, can play any one of these roles. This interactive demo using a large tablet and the NAO will allow attendees to play with the robot, which currently tries to maximize the collaborative score.

Creating a world for socio-cultural agents

Abstract: Creating agents that are capable of emulating similar socio-cultural dynamics to those found in human interaction remains as one of the hardest challenges of artificial intelligence. This problem becomes particularly important when considering embodied agents that are meant to interact with humans in a believable and empathic manner. In this article, we introduce a conceptual model for socio-cultural agents, and, based on this model, we present a set of requirements for these agents to be capable of showing appropriate socio-cultural behaviour. Our model differentiates between three levels of instantiation: the interaction level, consisting of elements that may change depending on the people involved, the group level, consisting of elements that may change depending on the group affiliation of the people involved, and the society level, consisting of elements that may change depending on the cultural background of those involved. As such, we are able to have culture alter agents’ social relationships rather than directly determining actions, allowing for virtual agents to act more appropriately in any social or cultural context.

Migration between two embodiments of an artificial pet

Abstract: Characters that cross dimensions have elicited an avid interest in literature and cinema. In analogy to these characters, we explore the concept of migration: Process by which an agent moves between embodiments, being active in only one at a time. We developed an autonomous artificial pet with two embodiments: A virtual within a smartphone and a physical robotic embodiment. Considering that owners' interactions with real pets lead to emotional attachment and potentially related health benefits, we conducted a user study with elementary school students to assess their attachment to the prototype and how natural they felt the interaction was. By the end of the experiment children felt closer to the artificial pet and 43.3% considered the two embodiments to correspond to the same entity, although migration was never explained to them. As a result, this paper presents a novel generic methodology that allows the evaluation of other implemented prototypes that support migration. Furthermore, we created a set of design guidelines for migrating agents.

Learning cost function and trajectory for robotic writing motion

Abstract: We present algorithms for inferring the cost function and reference trajectory from human demonstrations of hand-writing tasks. These two key elements are then used, through optimal control, to generate an impedance-based controller for a robotic hand. The key novelty lies in the flexibility of the feature design in the composition of the cost function, in contrast to the traditional approaches that consider linearly combined features. Cross-entropy-based methods form the core of our learning technique, resulting in sample-based stochastic algorithms for task encoding and decoding. The algorithms are validated using an anthropomorphic robot hand. We assess that the correct compliance is well encapsulated by subjecting the robot to perturbations during task reproduction.

Free Your Brain a Working Memory TrainingźGame

Abstract: Working memory training systems are designed to improve the user's working memory. However, current systems are frequently considered tedious deeply affecting the user's motivation and consequently the potential for training derived improvements. "Free Your Brain" is a brain training game combining insights from cognitive neuropsychological theories and flow theories. In this work we describe the game and its design process specifically establishing the link between the supporting theoretical background research and the developed solution.

Meet Me Halfway: Eye Behaviour as an Expression of Robot's Language.

Abstract: Eye contact is a crucial behaviour in human communication and therefore an essencial feature in humanrobot interaction. A study regarding the development of an eye behaviour model for a robotic tutor in a taskoriented environment is presented, along with a description of how our proposed model is being used to implement an autonomous robot in the EMOTE project. From Sci Fi movies and books, robots are becoming a real part of human lives. The field of robotics is widespreading and new forms of robots are being designed and created to interact with people everyday in our society, forcing to consider its social status. Nowadays, different types of robots are becoming real. According to a recent survey of The United Nations, robots can be grouped into three major categories: industrial robots, professional service robots, and personal service robots (Bartneck et al. 2007), accommodating several human requirements in various environments such as factories, battlefield, home, school, and so on. This way, according to the desired environment for the robot, different types of artificial intelligence (AI) are developed in order to express different types of behaviours and communication attending to the robot embodiment and perceptive capacities (e.g., industrial robots, contrary to social robots, do not interact with people and mostly serve to do heavy work in a factory). Since gaze is literally the first contact with another being, the exchange that takes place will often determine the first impression and play a role on the interaction flow (Collier 1985). In human-robot interaction (HRI) the head is implicitly the primary place for social interaction, and the eyes are an important communication feature, having a pragmatic design goal (DiSalvo et al. 2002). This has been leading most robot manufacturers to include eyes in the robots’ design. Thus, this paper focuses on eye behaviour as a feature of Robot Language and its implication in HRI. The eyes are the mirror of the soul Eye behaviour (EB) is one of the first most meaningful and critical non-verbal behaviour for social interaction (Cassell et al. 1994) and is indeed a crucial communication cue when Copyright c © 2014, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. Figure 1: (a)Mutual Eye Contact (b)Mutual Eye Gaze looking at HRI (Mutlu, Shiwa, and Kanda 2009). As humans and robots will share their physical space, EB emerges as a way of narrowing this communication, making the interaction seem more natural, intuitive, and engaging. In general terms, research shows that there are different types of EB (Lance and Marsella 2007) that comprise a set of properties (Lee et al. 2007) to serve different functions in the social interaction (Argyle et al. 1973). In this sense, the EB is interrelated with a context and has a role in cognitive, communicative and affective face-to-face interactions, where different EB patterns can emerge (Lee et al. 2007). Thereafter, EB can be of different types: there is a distinction between eye gaze (looking at the eyes of someone) and eye contact (establishing gaze that does not necessarily target eyes), that can also happen during mutual EB (Collier 1985) (see Fig.1a, 1b). On the other hand, when we intentionally redirect our gaze away from a face, we are performing gaze aversion which in the social context is related with cognitive, intimacy-modulating, and floor management functions (Andrist et al. 2014). Timing is also essential to interpret EB (Collier 1985). This literature serves to show the complexity of EB and the inherent challenges that the HRI field still needs to overcome to achieve a truly positive social interaction between a human and a robot. To achieve this stand, we propose a different perspective regarding robot’s communicative and expressive behaviour: since robots have different embodiments that constrain their expressive modalities, they should have a specific language that is adequate for their expression. We believe that by providing robots with their own communicative and expressive language, they will not be seen as ”limited humans”, but rather as different entities. Being EB the modality that most stands out in any type of social interaction, this type of behaviour has become our starting point. Figure 2: Eye Behaviour Study Setup (a b) Figure 3: Example of the gaze-glance-gaze dynamic

Seven Principles to Mine Flexible Behavior from Physiological Signals for Effective Emotion Recognition and Description in Affective Interactions

Abstract: Measuring affective interactions using physiological signals has become a critical step to understand engagements with human and artificial agents. However, traditional methods for signal analysis are not yet able to effectively deal with the differences of responses across individuals and with flexible sequential behavior. In this work, we rely on empirical results to define seven principles for a robust mining of physiological signals to recognize and characterize affective states. The majority of these principles are novel and driven from advanced pre-processing techniques and temporal data mining methods. A methodology that integrates these principles is proposed and validated using electrodermal signals collected during human-to-human and human-to-robot affective interactions.

"How do you imagine robots?" Childrens' expectations about robots

Abstract: Patŕicia Alves-Oliveira INESC-ID & Instituto Superior Tecnico, Universidade de Lisboa Porto Salvo, Portugal patricia.alves.oliveira@incesid.pt Helen Hastie Interaction Lab School of Mathematical and Computer Sciences Heriot-Watt University, Edinburgh h.hastie@hw.ac.uk Sofia Petisca INESC-ID & Instituto Superior Tecnico, Universidade de Lisboa Porto Salvo, Portugal sofia.petisca@inesc-id.pt Ana Paiva INESC-ID & Instituto Superior Tecnico, Universidade de Lisboa Porto Salvo, Portugal ana.paiva@inesc-id.pt Srinivasan Janarthanam Interaction Lab School of Mathematical and Computer Sciences Heriot-Watt University, Edinburgh sc445@hw.ac.uk

Make Way for the Robot Animators! Bringing Professional Animators and AI Programmers Together in the Quest for the Illusion of Life in Robotic Characters.

Abstract: We are looking at new ways of building algorithms for synthesizing and rendering animation in social robots that can keep them as interactive as necessary, while still following on principles and practices used by professional animators. We will be studying the animation process side by side with professional animators in order to understand how these algorithms and tools can be used by animators to achieve animation capable of correctly adapting to the environment and the artificial intelligence that controls the robot. Robotic characters are becoming widespread as useful tools in assistive, entertainment, tutoring and manufacturing applications. This is leading to the need of having better animation built into these robotic systems. Traditional character animation relies on having animators design animations which are then faithfully played back on the character. Now we pose the question: how can we have the animation process integrated with the artificial intelligence agent (AI) that drives the robot? We want the animation to be adaptable to both the agent’s internal state and its external environment, while keeping the quality of the animation in line with what an animator expects. The main question here is not just whether or not the users will like the robot or find it believable; instead, we want to have animators collaborating with the programmers in a way that the robot’s animation during interaction is satisfying to the actual animators. Motivation and Goals Animating an interactive robot is quite different from animating a virtual character. The biggest differences we point out from our experience are the fact that the robot shares the same physical space that humans do, and as such, follow on the same laws of physics. While animating a robot, animators must take into account gravity, inertia, and the actual torques that the robot’s motors can achieve. Because they are physical and can interact with users through various types of interfaces and sensors, robots also seem to lead to a much higher expectation from the users, and appeal more to some interactive non-verbal behaviours like gazing, rapport or proxemics than a virtual character would. Copyright c © 2014, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. Figure 1: Two example scenarios featuring a touch-based multimedia application, sensors, and different robots. On this issue of breaking out the traditional animation process into one that can be integrated with an AI, we highlight two different types of animation: Pre-designed character animation is when artists design an animation for a specific character topology using professional animation software, such that the animation can then be played back in a character with that same topology; Procedural character animation is when a software algorithm is able to synthesize the character’s animation in real-time, by having some higher-level control software tweaking the parameters of the algorithm. The purpose of our work is to merge both approaches by: 1. Finding out how to build character animation algorithms that can be tweaked by the AI that controls the character during interaction with users and the external physical world; 2. Having these algorithms make use of or blend with predesigned animation; 3. Bringing animation artists into the process of achieving the previous points. Our belief is that robot animation should be supported by principles and practices of professional character animation.

Computational Models of Cultural Behavior for Human-Agent Interaction (Dagstuhl Seminar 14131)

Abstract: From March 23 2014 to March 28, the seminar "Computational Models of Cultural Behavior for Human-Agent Interaction" held in Schloss Dagstuhl -- Leibniz Center for Informatics. During the seminar, an interdisciplinary group of researchers explored and discussed theories and techniques for computational models of culture as part of virtual human simulations. Culturally-sensitive agents do not only improve the acceptance of man-machine interfaces by adapting their verbal and non-verbal behavior to the user's assumed cultural background. They also bear enormous potential for a rapidly growing number of ICT-based language and cultural training scenarios that make use of role-play with virtual characters. The seminar brought together researchers with an interdisciplinary background that profited from each other's perspective and explored challenges for the future.

Non-Player Characters and Artificial Intelligence

Abstract: Serious Games rely on interactive systems to provide an efficient communication medium between the tutor and the user. Designing and implementing such medium is a multi-disciplinary task that aims at an environment that engages the user in a learning activity. User engagement is significantly related to the users’ sense of immersion or his willingness to accept the reality proposed by a game environment. This is a very relevant research topic for Artificial Intelligence (AI), since it requires computational systems to generate believable behaviors that can promote the users’ willingness to enter and engage in the game environment. In order to do this, AI research has been relying on social sciences, in particular psychology and sociology models, to ground the creation of computational models for non-player characters that behave according to the users’ expectations. In this chapter, the authors present some of the most relevant NPC research contributions following this approach. Gonçalo Pereira INESC-ID, Portugal & University of Lisbon, Portugal António Brisson INESC-ID, Portugal & University of Lisbon, Portugal João Dias INESC-ID, Portugal & University of Lisbon, Portugal André Carvalho INESC-ID, Portugal & University of Lisbon, Portugal Joana Dimas INESC-ID, Portugal & University of Lisbon, Portugal Samuel Mascarenhas INESC-ID, Portugal & University of Lisbon, Portugal Joana Campos INESC-ID, Portugal & University of Lisbon, Portugal Marco Vala INESC-ID, Portugal & University of Lisbon, Portugal Iolanda Leite INESC-ID, Portugal & University of Lisbon, Portugal Carlos Martinho INESC-ID, Portugal & University of Lisbon, Portugal Rui Prada INESC-ID, Portugal & University of Lisbon, Portugal Ana Paiva INESC-ID, Portugal & University of Lisbon, Portugal

An Embodied Empathic Tutor.

Abstract: This paper reports work being carried out in the EMOTE project (http://www.emote-project.eu/) towards the development of an empathic robot tutor at a multi-touch table for geography curriculum teaching for 11-14 year olds.

The influence of social display in competitive multiagent learning

Abstract: In this paper we analyze the impact of simple social signaling mechanisms in the performance of learning agents within competitive multiagent settings. In our framework, self-interested reinforcement learning agents interact and compete with each other for limited resources. The agents can exchange social signals that influence the total amount of reward received throughout time. In a series of experiments, we vary the amount resources available in the environment, the frequency of interactions and the importance each agent in the population gives to the social displays of others. We measure the combined performance of the population according to distinct selection paradigms based on the individual performances of each agent. The results of our study show that by focusing on the social displays of others, agents learn to collectively coordinate their feeding behavior by trading-off immediate benefit for long-term social welfare. Also, given populations where the impact of the social signal on the reward differs, the individuals with the highest fitness appear in the most socially-aware populations. The presence of social signaling gives also origin to more social inequalities where the more fit agents benefit from their higher status being appreciated by other members of the population.

Learning Effective Models of Emotions from Physiological Signals: The Seven Principles

Abstract: Learning effective models from emotion-elicited physiological responses for the classification and description of emotions is increasingly required to derive accurate analysis from affective interactions. Despite the relevance of this task, there is still lacking an integrative view of existing contributions. Additionally, there is no agreement on how to deal with the differences of physiological responses across individuals, and on how to learn from flexible sequential behavior and subtle but meaningful spontaneous variations of the signals. In this work, we rely on empirical evidence to define seven principles for a robust mining physiological signals to recognize and characterize affective states. These principles compose a coherent and complete roadmap for the development of new methods for the analysis of physiological signals. In particular, these principles address the current over-emphasis on feature-based models by including critical generative views derived from different streams of research, including multivariate data analysis and temporal data mining. Additionally, we explore how to use background knowledge related with the experimental setting and psychophysiological profiles from users to shape the learning of emotion-centered models. A methodology that integrates these principles is proposed and validated using signals collected during human-to-human and human-to-robot affective interactions.

Social Presence and Artificial Opponents

Abstract: In this paper, we argue that playing board games is a form of entertainment that provides participant’s with rich social interactions. However, when we try to replace one of the players with an artificial opponent, the social interaction between players is negatively affected by the social inability of nowadays artificial opponents. Currently, the social presence that human players attribute to artificial opponents is quite low. In order to tackle this problem, we investigate the topic of social presence, its definitions and which are its contributing factors. Also, we looked at nowadays social interactions with artificial agents and how these kind of agents deal with long term interactions. This related work along with some previous studies contributed to the development of a set of five guidelines intended for improving social presence in board game artificial opponents. Finally, in order to illustrate howone can implement such guidelines, we give an example of how we implemented them in a scenario where a digital table is used as an interface for a board game and a social robot plays Risk against three human opponents.

Child-Robot Interaction: Social Bonding, Learning and Ethics

Abstract: This workshop aims to exchange experiences with issues surrounding Child-Robot Interaction. More specifically, the main aims are to discuss how social bonding between children and robots can be evaluated, how robots can be used to aid children in their learning process, but also what ethical issues arise when children learn from and bond with a robot. Another aim is to discuss how teachers’ and caretakers’ perspectives on children’s use of robots should be taken into account when designing and evaluating robots for children.