This paper argues the position that an essential approach to the advancement of the state of the art in cognitive systems is to focus on systems that deeply integrate knowledge representations, cognitive capabilities, and knowledge content. Integration is the path to aggregating constraints in ways that improve the science of cognitive systems. However, evaluating the role of knowledge among these constraints has largely been ignored, in part because it is difficult to build and evaluate systems that incorporate large amounts of knowledge. We provide suggestions for evaluating such systems and argue that such evaluations will become easier as we come closer to applying usefully new, integrated learning mechanisms that are capable of acquiring large and effective knowledge bases.

In this paper, we consider the issue of guiding ethical behavior in human-robot teams from a systemic viewpoint. Considering a team as a sociotechnical complex, we look at how responsibility for actions can arise through the interaction between the different actors in the team while playing specific roles. We define the notions of role, discuss how they establish a social network, and then use logical notions of multi-agent trust to formalize responsibility as accountability against capabilities that are invoked during collaboration.

Social cognition is a key feature of human-level intelligence. However, social reasoning faculties are rarely included in cognitive systems. To encourage research in this direction, we introduce a practical, computational framework that enables socially aware inference. We demonstrate the framework's ability to model a common, complex, and under-investigated aspect of human social behavior: deception. Moreover, we show how a system implementing this framework could dynamically respond once it has detected a lie. We then discuss some of the challenges associated with deception, ending with an outline of future research directions.

Beyond belief change and meme adoption, both genetics and infection have been spoken of in terms of information transfer. What we examine here, concentrating on the specific case of transfer between sub-networks, are the differences in network dynamics in these cases: the different network dynamics of germs, genes, and memes. Germs and memes, it turns out, exhibit a very different dynamics across networks. For infection, measured in terms of time to total infection, it is network type rather than degree of linkage between sub-networks that is of primary importance. For belief transfer, measured in terms of time to consensus, it is degree of linkage rather than network type that is crucial. Genes model each of these other dynamics in part, but match neither in full. For genetics, like belief transfer and unlike infection, network type makes little difference. Like infection and unlike belief, on the other hand, the dynamics of genetic information transfer within single and between linked networks are much the same. In ways both surprising and intriguing, transfer of genetic information seems to be robust across network differences crucial for the other two.

In this paper we address the design of robots that can be successful partners to humans in joint activity. The paper outlines an approach to achieving adjustable autonomy during execution- and hence to achieve resilient multi-actor joint action - based on both temporal and epistemic situation projection. The approach is based on non-deterministic planning techniques based on the situations calculus.

One challenge for cognitive architectures is to effectively use different forms of knowledge and learning. We present a case study of Soar agents that play a multiplayer dice game, in which probabilistic reasoning and heuristic symbolic knowledge appear to play a central role. We develop and evaluate a collection of agents that use different combinations of probabilistic decision making, heuristic symbolic reasoning, opponent modeling, and learning. We demonstrate agents that use Soar’s rule learning mechanism (chunking) to convert deliberate reasoning with probabilities into implicit reasoning, and then use reinforcement learning to further tune performance.

The better the team mental model, the better the teamwork. An important aspect of what determines a good team model is the extent to which the model is shared by the team members. This paper presents suggestions for measuring the extent to which teams have a shared mental model and describes how these measures are related to team performance. The most promising measures of sharedness proposed so far rely on using a compositional approach for team modeling and on a situation-sensitive relevance relation that indicates to what extent components contribute to team performance. A case study illustrates the approach and initial results on measuring performance when teams use different levels of sharedness.

In this paper, we address the problem of computing the policy of an autonomous agent, taking human recommendations into account which could be appropriate for mixed initiative, or adjustable autonomy. For this purpose, we present Goal Biased Markov Decision Process (GBMDP) which assume two kinds of recommendation. The human recommends to the agent to avoid some situations (represented by undesirable states), or he recommends favorable situations represented by desirable states. The agent takes those recommendations into account by updating its policy (only updating the states concerned by the recommendations, not the whole policy). We show that GBMDP is efficient and it improves the human's intervention by reducing its time of attention paid to the agent. Moreover, GBMDP optimizes robot's computation time by updating only the necessary states. We also show how GBMDP can consider more than one recommendation. Finally, our experiments show how we update policies which are intractable by standard approaches.

This paper is about generating questions in human-robot interaction. We survey existing work on the forms and meanings of questions in English and discuss the pragmatic effects resulting from an interplay between the choice of syntactic form and intonation. We propose an approach to formalization based on a notion of common ground and commitment, set in a model of situated dialogue as part of collaborative activity where we explicitly model the beliefs and intentions of both the robot and the human. Questions come about by abductively inferring an intentional structure grounded in the belief model and indicating commitments. Content planning and surface realization turn this into a question of the appropriate form.

This paper discusses the challenge of designing instructable agents that can learn through interaction with a human expert. Learning through instruction is a powerful paradigm for acquiring knowledge because it limits the complexity of the learning task in a variety of ways. To support learning through instruction, the agent must be able to effectively communicate its lack of knowledge to the human, comprehend instructions, and apply them to the ongoing task. Weidentify some problems of concern when designing instructable agents. We propose an agent design that addresses some of these problems. We instantiate this design in the Soar cognitive architecture and analyze its capabilities on a learning task.