Reports of the Workshops of the Thirty-First AAAI Conference on Artificial Intelligence

The AAAI-17 workshop program included 17 workshops covering a wide range of topics in AI. Workshops were held Sunday and Monday, February 4-5, 2017 at the Hilton San Francisco Union Square in San Francisco, California, USA. This report contains summaries of 12 of the workshops, and brief abstracts of the remaining 5

This paper presents the STAR system, a system for automated narrative comprehension, developed on top of an argumentation-theoretic formulation of defeasible reasoning, and strongly following guidelines from the psychology of comprehension. We discuss the system's use in psychological experiments on story comprehension, and our plans for its broader use in empirical studies concerning wider issues of commonsense reasoning.

Psychological evidence supporting the profound effortlessness (and often substantial carelessness) with which human cognition copes with typical daily life situations abounds. In line with this evidence, we propose a formal semantics for computational cognition that places emphasis on the existence of naturalistic and unpretentious algorithms for representing, acquiring, and manipulating knowledge. At the heart of the semantics lies the realization that the partial nature of perception is what ultimately necessitates — and hinders — cognition. Inexorably, this realization leads to the adoption of a unified treatment for all considered cognitive processes, and to the representation of knowledge via prioritized implication rules. Through discussion and the implementation of an early prototype cognitive system, we argue that such fast and loose semantics may offer a good basis for the development of machines with cognitive abilities.

Eliciting user preferences constitutes a major step towards developing recommender systems and decision support tools. Assuming that preferences are ceteris paribus allows for their concise representation as Conditional Preference Networks (CP-nets). This work presents the first empirical investigation of an algorithm for reliably and efficiently learning CP-nets in a manner that is minimally intrusive . At the same time, it introduces a novel process for efficiently reasoning with (the learned) preferences.

Among the issues faced by an intelligent agent, central is that of reconciling the, often contradictory, pieces of knowledge — be those given, learned, or sensed — at its disposal. This problem, known as knowledge qualification, requires that pieces of knowledge deemed reliable in some context be given preference over the others. These preferences are typically viewed as encodings of reasoning patterns; so, the frame axiom can be encoded as a preference of persistence over spontaneous change. Qualification, then, results by the principled application of these preferences. We illustrate how this can be naturally done through argumentation, by uniformly treating object-level knowledge and reasoning patterns alike as arguments that can be defeated by other stronger ones. We formulate an argumentation framework for Reasoning about Actions and Change that gives a semantics for Action Theories that include a State Default Theory. Due to their explicit encoding as preferences, reasoning patterns can be adapted, when and if needed, by a domain designer to suit a specific application domain. Furthermore, the reasoning patterns can be defeated in lieu of stronger external evidence, allowing, for instance, the frame axiom to be overridden when unexpected sensory information suggests that spontaneous change may have broken persistence in a particular situation.

Among the many aspects of human intelligence that currently elude the simulation by machines is that of story understanding. Although many theories of narrative have been proposed, several processes pertaining to narrative remain inadequately formalized and, hence, beyond full mechanization. This work proposes a general formal framework that attempts to make precise such processes and related notions, with first and foremost that of what constitutes a narrative. Emphasis is placed on identifying certain premises that narratives are expected to adhere to, and deriving the formal implications that these have in terms of the computability of the various relevant notions. Among others, it is established that checking whether a discourse is a narrative is decidable, and that narratives can be computably enumerated and, hence, unambiguously indexed.