Interaction Studies - Volume 15, Issue 3, 2014

Humans routinely transmit and interpret subtle information about their mental states through the language they use, even when only the language text is available. This suggests humans can utilize the linguistic signature of a mental state (its mindprint), comprised of features in the text. Once the relevant features are identified, mindprints can be used to automatically identify mental states communicated via language. We focus on the mindprints of eight mental states resulting from intentions, attitudes, and emotions, and present a mindprint-based machine learning technique to automatically identify these mental states in realistic language data. By using linguistic features that leverage available semantic, syntactic, and valence information, our approach achieves near-human performance on average and even exceeds human performance on occasion. Given this, we believe mindprints could be very valuable for intelligent systems interacting linguistically with humans. Keywords: mental state; linguistic features; mindprint; natural language processing; information extraction

We introduce the CUBISM system for the analysis and deep understanding of multi-participant dialogues. CUBISM brings together two typically separate forms of discourse analysis: semantic analysis and sociolinguistic analysis. In the paper proper, we describe and illustrate major components of the CUBISM system, and discuss the challenge posed by the system’s ultimate purpose, which is to automatically detect anomalous changes in participants’ expressed or implied beliefs about the world and each other, including shifts toward or away from cultural and community norms.

Mental model ascription – also called mindreading – is the process of inferring the mental states of others, which happens as a matter of course in social interactions. But although ubiquitous, mindreading is presumably a highly variable process: people mindread to different extents and with different results. We hypothesize that human mindreading ability relies on a large number of personal and contextual features: the inherent abilities of specific individuals, their current physical and mental states, their knowledge of the domain of discourse, their familiarity with the interlocutor, the risks associated with an incorrect assessment of intent, and so on. This paper presents a theory of mindreading that models diverse artificial intelligent agents using an inventory of parameters and value sets that represent traits of humans and features of discourse contexts. Examples are drawn from Maryland Virtual Patient, a prototype system that will permit medical trainees to diagnose and treat cognitively modeled virtual patients with the optional assistance of a virtual tutor. Since real patients vary greatly with respect to physiological and cognitive features, so must a society of virtual patients. Modeling such variation is one of the goals of the overall OntoAgent program of research and development.

This paper examines the task of understanding dialogues in terms of the mental states of the participating agents. We present a motivating example that clarifies the challenges this problem involves and then outline a theory of dialogue interpretation based on abductive inference of these unobserved beliefs and goals, incremental construction of explanations, and reliance on domain-independent knowledge. After this, we describe UMBRA, an implementation of the theory that embodies these assumptions. We report experiments with the system that demonstrate its ability to accurately infer the conversants’ mental states even when some speech acts are unavailable. We conclude by reviewing related research on dialogue and discussing avenues for future study.

Behavior oftentimes allows for many possible interpretations in terms of mental states, such as goals, beliefs, desires, and intentions. Reasoning about the relation between behavior and mental states is therefore considered to be an effortful process. We argue that people use simple strategies to deal with high cognitive demands of mental state inference. To test this hypothesis, we developed a computational cognitive model, which was able to simulate previous empirical findings: In two-player games, people apply simple strategies at first. They only start revising their strategies when these do not pay off. The model could simulate these findings by recursively attributing its own problem solving skills to the other player, thus increasing the complexity of its own inferences. The model was validated by means of a comparison with findings from a developmental study in which the children demonstrated similar strategic developments.