Andrew J. Cowell

Manipulation of non-verbal interaction style and demographic embodiment to increase anthropomorphic computer character credibility

For years, people have sought more natural means of communicating with their computers. Many have suggested that interaction with a computer should be as easy as interacting with other people, taking advantage of the multimodal nature of human communication. While users should, in theory, gravitate to such anthropomorphic embodiments, quite the contrary has been experienced; users generally have been dissatisfied and abandoned their use. This suggests a disconnect between factors that make human-human communication engaging and those used by designers to support human-agent interaction. This paper discusses a set of empirical studies that attempted to replicate human-human non-verbal behavior. The focus revolved around behaviors that portray a credible facade, thereby helping embodied conversational agents (ECAs) to form a successful cooperative dyad with users. Based on a review of the non-verbal literature, a framework was created that identified trustworthy and credible non-verbal behaviors across five areas and formed design guidelines for character interaction. The design suggestions for those areas emanating from the facial region were experimentally supported but there was no concordant increase in perceived trust when bodily regions (posture, gesture) were added. In addition, in examining the importance of demographic elements in embodiment, it was found that users prefer to interact with characters that match their ethnicity and are young looking. There was no significant preference for gender. The implications of these results, as well as other interesting consequences are discussed.

Embodiment and Interaction Guidelines for Designing Credible, Trustworthy Embodied Conversational Agents

This paper discusses our recent studies on Embodied Conversational Agent (ECA) design strategies to encourage credible and trustworthy dialogue. We approach the problem from two specific directions: the embodiment that the character ‘wears’ during its interchange with the user, and the method of interaction used by the ECA to engage the user. Our results indicate that while users generally prefer to interact with a youthful character matching their ethnicity, no significant preferences were indicated for character gender. For interaction, our results indicated that a character that portrayed trusting nonverbal behaviors was rated as being significantly more credible than a character portraying no nonverbal behavior, or one that portrayed non-trusting behaviors. Other interesting results from this work are also discussed.

Evaluating Agent Architectures: Cougaar, Aglets and AAA

Research and development organizations are constantly evaluating new technologies in order to implement the next generation of advanced applications. At Pacific Northwest National Laboratory, agent technologies are perceived as an approach that can provide a competitive advantage in the construction of highly sophisticated software systems in a range of application areas. To determine the sophistication, utility, performance, and other critical aspects of such systems, a project was instigated to evaluate three candidate agent toolkits. This paper reports on the outcomes of this evaluation, the knowledge accumulated from carrying out this project, and provides insights into the capabilities of the agent technologies evaluated.

Construction and validation of a neurophysio-technological framework for imagery analysis

Intelligence analysts are bombarded with enormous volumes of imagery, which they must visually filter through to identify relevant areas of interest. Interpretation of such data is subject to error due to (1) large data volumes, implying the need for faster and more effective processing, and (2) misinterpretation, implying the need for enhanced analyst/system effectiveness. This paper outlines the Revolutionary Accelerated Processing Image Detection (RAPID) System, designed to significantly improve data throughput and interpretation by incorporating advancing neurophysiological technology to monitor processes associated with detection and identification of relevant target stimuli in a non-invasive and temporally precise manner. Specifically, this work includes the development of innovative electroencephalographic (EEG) and eye tracking technologies to detect and flag areas of interest, potentially without an analysts conscious intervention or motor responses, while detecting and mitigating problems with tacit knowledge, such as anchoring bias in real-time to reduce the possibility of human error.

Understanding the dynamics of collaborative multi-party discourse

In this paper, we discuss the efforts underway at the Pacific Northwest National Laboratory in understanding the dynamics of multi-party discourse across a number of communication modalities, such as email, instant messaging traffic and meeting data. Two prototype systems are discussed. The Conversation Analysis Tool (ChAT) is an experimental test-bed for the development of computational linguistic components and enables users to easily identify topics or persons of interest within multi-party conversations, including who talked to whom, when, the entities that were discussed, etc. The Retrospective Analysis of Communication Events (RACE) prototype, leveraging many of the ChAT components, is an application built specifically for knowledge workers and focuses on merging different types of communication data so that the underlying message can be discovered in an efficient, timely fashion.

Social media and social reality

Social Media provide an exciting and novel view into social phenomena. The vast amounts of data that can be gathered from the Internet coupled with massively parallel supercomputers such as the Cray XMT open new vistas for research. Conclusions drawn from such analysis must recognize that social media are distinct from the underlying social reality. Rigorous validation is essential. This paper briefly presents results obtained from computational analysis of social media - utilizing both blog and twitter data. Validation of these results is discussed in the context of a framework of established methodologies from the social sciences. Finally, an outline for a set of supporting studies is proposed.

Supporting Mutual Understanding in a Visual Dialogue between Analyst and Computer

The Knowledge Associates for Novel Intelligence (KANI) project is developing a system of automated “associates” to actively support and participate in the information analysis task. The primary goal of KANI is to use automatically extracted information in a reasoning system that draws on the strengths of both a human analyst and automated reasoning. The interface between the two agents is a key element in achieving this goal. The KANI interface seeks to support a visual dialogue with mixed-initiative manipulation of information and reasoning components. The interface must achieve mutual understanding between the analyst and KANI of the others actions. Toward this mutual understanding, KANI allows the analyst to work at multiple levels of abstraction over the reasoning process, links the information presented across these levels to make use of interaction context, and provides querying facilities to allow exploration and explanation.

Knowledge encapsulation framework for technosocial predictive modeling

Analysts who use predictive analytics methods need actionable evidence to support their models and simulations. Commonly, this evidence is distilled from large data sets with significant amount of culling and searching through a variety of sources including traditional and social media. The time/cost effectiveness and quality of the evidence marshaling process can be greatly enhanced by combining component technologies that support directed content harvesting, automated semantic annotation, and content analysis within a collaborative environment, with a functional interface to models and simulations. Existing evidence extraction tools provide some, but not all, the critical components that would empower such an integrated knowledge management environment. This paper describes a novel evidence marshaling solution that significantly advances the state of the art. Its embodiment, the Knowledge Encapsulation Framework (KEF), offers a suite of semi-automated and configurable content harvesting, vetting, annotation and analysis capabilities within a wiki-enabled and user-friendly visual interface that supports collaborative work across distributed teams of analysts. After a summarization of related work, our motivation, and the technical implementation of KEF, we will explore the model for using KEF and results of our research.

Using Physiological Measures to Discriminate Signal Detection Outcome during Imagery Analysis

An experiment was conducted to explore the feasibility of using physiological indicators (i.e. eye-tracking and electroencephalography [EEG]) to drive identification of relevant areas of interest during imagery analysis. Results indicate that ocular fixations are longer when a target is believed to be present. Furthermore, the accuracy of correct identification of targets could be identified based on fixation duration, given that fixations were significantly longer when a target was actually present. In addition, by synching eye-tracking fixation points to EEG, fixation-locked event-related potentials (FLERPs) show potential for detecting distinctive patterns and scalp distributions for various types of fixations, which may be used to classify fixation points based on level of interest. This paper reports findings from a study and summarizes challenges and implications for constructing a system where eye tracking is used to drive EEG ERP evaluation of interest during a defined search task within complex static images.

The Human-Information Workspace (HI-Space): Ambient Table Top Entertainment

This paper introduces the Human Information Workspace (HI-Space) as a test-bed for evaluating new information exploration mechanisms. In moving from dated interaction devices and small computer monitors, we aim to utilize more natural surfaces such as tables and walls as our interaction space. In testing our theories, we have produced a number of gaming applications as test cases. Here, we report on our most popular application, Virtual Hockey.