Elisabetta Zibetti

Predicting Extraversion from Non-verbal Features During a Face-to-Face Human-Robot Interaction

In this paper we present a system for automatic prediction of extraversion during the first thin slices of human-robot interaction (HRI). This work is based on the hypothesis that personality traits and attitude towards robot appear in the behavioural response of humans during HRI. We propose a set of four non-verbal movement features that characterize human behavior during the interaction. We focus our study on predicting Extraversion using these features extracted from a dataset consisting of 39 healthy adults interacting with the humanoid iCub. Our analysis shows that it is possible to predict to a good level (64 %) the Extraversion of a human from a thin slice of interaction relying only on non-verbal movement features. Our results are comparable to the state-of-the-art obtained in HHI [23].

Contextual Categorization: A Mechanism Linking Perception and Knowledge in Modeling and Simulating Perceived Events as Actions

The specific objective of this paper is to introduce the computer model ACACIA (Action by Contextually Automated Categorizing Interactive Agents) capable of simulating the way in which context is taken into account for the interpretation of perceived actions elaborated by a number of autonomous moving agents in a bidimensional space. With this in mind, we will examine some different modeling approaches in Artificial Intelligence and Artificial Life and emphasize the strong and weak points of each approach in relation to the set of issues addressed by our theory based on Contextual Categorization. Second, we provide a theoretical explanation of how contextual categorization accounts for temporal and environmental context to interpret ongoing situations in terms of perceived action. Finally, we describe the computer implementation of ACACIA, and we propose a preliminary simulation of a simple situation using StarLogo software.

The Role of Context in Interpreting Perceived Events as Actions

In this paper we discuss the role of context in interpreting and understanding perceived events as actions carried out by other people. The context is defined both as temporal (relations between events) and as being comprised of the state of the situation as cognitively processed (object properties) at any given time. We begin by presenting this overlooked field through previous work and perceived action models in psychology and in artificial intelligence. We will argue that the principal mechanism involved in perceiving action is categorisation of the objects in the environment (environmental context) and of the temporal relations between events (temporal context). The mechanisms involved in this process are modelled using the Dynamic Allocation of Meaning Model (C.A.D.S.). This model proposes an explanation of the manner in which temporal information and the perceived properties of a situation interact.

Understanding actions: contextual dimensions and heuristics

The aim of this paper is to analyze how people build an internal representation of actions done by others by matching observable data (objects, events, spatial and temporal context) with knowledge and by making inferences about not observable (intention, goal, anticipation, and planning) data. For that purpose, we list the set of dimensions that define “what an action is” accompanied with their contextual aspects, and the heuristics used by the human cognitive system. Finally, we propose that processing objects properties in terms of function-to-patient relation between objects, and in terms of patient-to-function within objects, is a possible way to explain the context effects with respect to the fine-grained vs. the large-grained semantic organization of the spatial and temporal context.

Reasoning based on categorisation for interpreting and acting: a first approach

Taking a detour to reach a goal is intelligent behavior based on making inferences. The main purpose of the present research is to show how such apparently complex behavior can emerge from basic mechanisms such as contextual categorisation and goal attribution when perceiving people. We presentacacia (Action by Contextually Automated Categorising Interactive Agents), a computer model implemented using StarLogo software, grounded in the principles of Artificial Life (Al), capable of simulating the behavior of a group of agents with a goal (for instance, to find a “treasure” in a “treasure hunt”) in an environment where obstacles mask the goal site. The results of the simulations show that agents reach the goal the fastest when they follow each other and take detours. We argue that these results indicate that intelligent adaptive behavior is based on the contextual categorisation of environmental constrainst (that is, obstacles and other agents).

How Much Is “About”? Fuzzy Interpretation of Approximate Numerical Expressions

Approximate Numerical Expressions (ANEs) are linguistic expressions involving numbers and referring to imprecise ranges of values, such as “about 100”. This paper proposes to interpret ANEs as fuzzy numbers. A model, taking into account the cognitive salience of numbers and based on critical points from Pareto frontiers, is proposed to characterise the support, the kernel and the 0.5-cut of the corresponding membership functions. An experimental study, based on real data, is performed to assess the quality of these estimated parameters.

Towards Behavioral Objects: A Twofold Approach for a System of Notation to Design and Implement Behaviors in Non-anthropomorphic Robotic Artifacts

Among robots, non-anthropomorphic robotic artifacts are in an interesting position: the fact that they do not resemble living beings, yet impart a sense of agency through the way they move, motivates to consider motion as a source of expressivity in itself, independently of any morphological cues. This problematic is considered in parallel to the question of movement notation and the different levels of abstraction that one may consider when reflecting on movement and its relation to a spatial, temporal and social context. This is through a twofold perspective, drawing on both dance notation and cognitive psychology, that we consider the question of movement notation, and its relation to expressive gestures and psychological attributes. To progress in the direction of a system of notation that could integrate the qualitative, relational, and behavioral aspects of movement, we propose different typologies and a model of constraints to analyze, conceive and implement behaviors in robotic artifacts.

Mis)behavioral Objects

While the movement toward granting ever more power to users is very real, can we understand it not from the viewpoint of humans, but that of the objects? How can we design the empowerment of both users and objects, but starting from the objects? Could the object then change its status and become a subject, or at least an agent? With that as our starting position, we pose the hypothesis, theoretical and practical, that in order to engage with such prospects, the objects in question must be endowed with behaviors. And rather than resorting to types of expression related to the morphology or dressing of the objects, we want to focus on the expressive capacity of movement, on the objects’ power to act as embodied fundamentally by movement, by actions that give shape to what would then be possible to describe as “behaviors”. Our approach is first and foremost a matter of art and design, while entertaining a dialogue between two other disciplines: robotics, to put into practice our reflection, and cognitive science, to better understand and prepare this new form of interaction “object-human” that we are seeking. More specifically we are looking to develop objects that misbehave, as a way for these objects to go against the function they have been designed for, and thus develop a certain subjectivity that could enable an affective relationship that is valid for itself. From a state of knowledge in art and design, and from an analysis of behaviour from the point of view of action perception, we propose a design space based on the behaviours produced and the interpretations they elicit in terms of mental states. This conceptual apparatus is put into practice through workshops during which we propose a form of collective experimentation, with the help of our modular robotics toolkit MisB KIT, open-source and accessible both to non-programmer practitioners and to developers. Following a first series of workshops, in particular the one held at TEI 2014 conference, along with the Tangible Media group from the MIT Medialab, we draw some conclusions and discuss some new perspectives regarding the development of (mis)behavioral objects.

Behavioral objects: the rise of the evocative machines

A new race of artifacts comes equipped with behavioral properties. Those properties transmute the very nature of the object, granting it a life of its own and a special status that stems from the psychological attributions humans naturally produce when confronted by autonomous movements. This article examines what makes behavioral objects special in terms of the psychological properties they evoke in an observer. We look into the notion of behavior and evaluate to what extent the concept of anthropomorphism is a valid construct when considering the behavior of artificial objects. Based on recent research in cognitive psychology, we propose a framework to conceptualize the way people infer psychological attributes from movement, and the way it applies to behavioral objects.

Interacting with Non-anthropomorphic Robotic Artworks and Interpreting Their Behaviour

Art installations involving robotic artifacts provide an opportunity to examine human relationships with robots designed solely for the purpose of sustaining evocative behaviours. In an attempt to determine the behavioural characteristics and personality traits attributed by a human to a robotic artifact, we investigated an audience’s experience of an installation that presented three robotic artifacts moving autonomously in an exhibition space. In order to describe the audience’s experience, we present two studies that revealed the psychological attributions spontaneously produced from observing the robots, and visitors’ physical exploration patterns inside the exhibition. We propose a psychological profile for the artwork, and a tentative organization for the attribution process. Using a cluster analysis performed on visitors’ trajectories inside the installation, we highlight four different exploration and interaction heuristics characterized by patterns of approach or withdrawal, passive observation and exploration.