Marco Gillies

Building Expression into Virtual Characters

Virtual characters are an important part of many 3D graphical simulations. In entertainment or training applications, virtual characters might be one of the main mechanisms for creating and developing content and scenarios. In such applications the user may need to interact with a number of different characters that need to invoke specific responses in the user, so that the user interprets the scenario in the way that the designer intended. Whilst representations of virtual characters have come a long way in recent years, interactive virtual characters tend to be a bit “wooden” with respect to their perceived behaviour. In this STAR we give an overview of work on expressive virtual characters. In particular, we assume that a virtual character representation is already available, and we describe a variety of models and methods that are used to give the characters more “depth” so that they are less wooden and more plausible. We cover models of individual characters’ emotion and personality, models of interpersonal behaviour and methods for generating expression.

Comparing and evaluating real time character engines for virtual environments

As animated characters increasingly become vital parts of virtual environments, then the engines that drive these characters increasingly become vital parts of virtual environment software. This paper gives an overview of the state of the art in character engines, and proposes a taxonomy of the features that are commonly found in them. This taxonomy can be used as a tool for comparison and evaluation of different engines. In order to demonstrate this we use it to compare three engines. The first is Cal3D, the most commonly used open source engine. We also introduce two engines created by the authors, Piavca and HALCA. The paper ends with a brief discussion of some other popular engines.

Integrating Autonomous Behavior and User Control for Believable Agents

Autonomous agents can help users by taking on a substantial workload, and performing tasks that are too complex for a human. However, in some systems complete autonomy is undesirable as it removes control from the user. It is therefore important to include some user control while maintaining the reduced workload associated with an autonomous system. This is particularly true of user avatars in virtual worlds, appropriate non-verbal communication is too complex to be directed explicitly and should therefore be controlled by an agent. However, the non-verbal communication should express the feelings of the user, so a degree of human input is needed. This paper presents Demeanour, an autonomous system for generating body language in avatars, which integrates user input based on a three level methodology of customization and control.

Socially anxious and confident men interact with a forward virtual woman: an experimental study.

Background Male volunteers entered an immersive virtual reality that depicted a party, where they were approached by a lone virtual woman who initiated a conversation. The goal was to study how socially anxious and socially confident men would react to this event. Interest focused on whether the socially anxious participants would exhibit sustained anxiety during the conversation or whether this would diminish over time, and differ from the responses of the more socially confident men. Methodology The scenario was a party with five virtual characters, four sitting at a distance from the participant and talking amongst themselves and one lone woman standing closer. The woman approached the participant, introduced herself and initiated a conversation that was first about mundane matters and then became more personal and intimate. Participants were men who were either relatively socially confident (18) or socially anxious in their relationships with women (18). A second experimental factor was whether or not the other four characters occasionally looked towards the participant. There was a post-trial questionnaire about social anxiety in relation to the experience, and skin conductance and ECG physiological measures were recorded. Our expectation was that the socially anxious participants would show greater anxiety throughout. Conclusions Compared to baseline readings both socially confident and socially anxious groups on average showed signs of significantly increased stress at the initial approach of the virtual woman. The stress then diminished once the conversation entered into the mundane phase and then did not significantly change. Comparing pre- and post-questionnaire anxiety scores there was no change for the more confident participants but a significant decrease in average score amongst the anxious group. The methodology of placing socially anxious participants in a virtual reality where they can gain experience of how to act in a stressful situation promises a novel way forward for treating social anxiety.

Male Bodily Responses during an Interaction with a Virtual Woman

This work presents the analysis of the body movement of male participants, while talking with a life-size virtual woman in a virtual social encounter within a CAVE-like system. We consider independent and explanatory variables including whether the participant is the centre of attention in the scenario, whether the participant is shy or confident, and his relationship status. We also examine whether this interaction between the participant and the virtual character changes as the conversation progresses. The results show that the participants tend to have different hand movements, head movements, and posture depending on these conditions. This research therefore provides strong evidence for using body movement as a systematic method to assess the responses of people within a virtual environment, especially when the participant interacts with a virtual character. These results also point the way towards the application of this technology to the treatment of social phobic males.

A Model of Interpersonal Attitude and Posture Generation

We present a model of interpersonal attitude used for generating expressive postures for computer animated characters. Our model consists of two principle dimensions, affiliation and status. It takes into account the relationships between the attitudes of two characters and allows for a large degree of variation between characters, both in how they react to other characters’ behaviour and in the ways in which they express attitude.

Human-Centred Machine Learning

Machine learning is one of the most important and successful techniques in contemporary computer science. It involves the statistical inference of models (such as classifiers) from data. It is often conceived in a very impersonal way, with algorithms working autonomously on passively collected data. However, this viewpoint hides considerable human work of tuning the algorithms, gathering the data, and even deciding what should be modeled in the first place. Examining machine learning from a human-centered perspective includes explicitly recognising this human work, as well as reframing machine learning workflows based on situated human working practices, and exploring the co-adaptation of humans and systems. A human-centered understanding of machine learning in human context can lead not only to more usable machine learning tools, but to new ways of framing learning computationally. This workshop will bring together researchers to discuss these issues and suggest future research questions aimed at creating a human-centered approach to machine learning.

Learning Finite-State Machine Controllers From Motion Capture Data

With characters in computer games and interactive media increasingly being based on real actors, the individuality of an actors performance should not only be reflected in the appearance and animation of the character but also in the AI that governs the characters behavior and interactions with the environment. Machine learning methods applied to motion capture data provide a way of doing this. This paper presents a method for learning the parameters of a finite-state machine (FSM) controller. The method learns both the transition probabilities of the FSM and also how to select animations based on the current state.

Non-representational Interaction Design

This paper presents how non-representational views of cognition can inform interaction design as it moves from traditional graphical user interfaces to more bodily forms of interaction such as gesture or movement tracking. We argue that the true value of these “bodily” interfaces is that they can tap our prior skills for interacting in the world. However, these skills are highly non-representational and so traditional representational approaches to interaction design will fail to capture them effectively. We propose interactive machine learning as an alternative approach to interaction design that is able to capture non-representational sensori-motor couplings by allowing people to design by performing actions rather than by representing them. We present an example of this approach applied to designing interactions with video game characters.

Piavca: A Framework for Heterogeneous Interactions with Virtual Characters

This paper presents a virtual character animation system for real time multi-modal interaction, which can handle many different styles of interaction and animation simultaneously. Our system is based on a model of animation that represents different types of animations as instantiations of an abstract function representation. This makes it easy to combine different types of animation and encourages the creation of behavior out of basic building blocks.