For affective computing to have an impact outside the laboratory, facial expressions must be studied in rich naturalistic situations. We argue negotiations are one such situation as they are ubiquitous in daily life, often evoke strong emotions, and perceived emotion shapes decisions and outcomes. Negotiations are a growing focus in AI research and applications, including agents that negotiate directly with people and attempt to use affective information. We introduce the DyNego-WOZ Corpus, which includes dyadic negotiation between participants and wizard-controlled virtual humans. We demonstrate the value of this corpus to the affective computing community by examining participants’ facial expressions in response to a virtual human negotiation partner. We show that people's facial expressions typically co-occur with the end of their partner's speech (suggesting they reflect a reaction to the content of this speech), that these reactions do not correspond to prototypical emotional expressions, and that these reactions can help predict the expresser's subsequent action. We highlight challenges in working with such naturalistic data, including difficulties of expression recognition during speech, and the extreme variability of expressions, both across participants and within a negotiation. Our findings reinforce arguments that facial expressions convey more than emotional state but serve important communicative functions.

Women earn less than men in technical fields. Competing theories have been offered to explain this disparity. Some argue that women underperform in negotiating their salary, in-part due to language in job descriptions, called gender triggers, which leave women feeling disadvantaged in salary negotiations. Others point to structural and institutional bias: i.e., recruiters make better offers to men even when women exhibit equal negotiation skills. As a final salary is co-constructed though an interaction between employees and recruiters, it is difficult to disentangle these views. Here, we discuss how intelligent virtual agents serve as powerful methodological tools that lend new insight into this psychological debate. We use virtual negotiators to examine the impact of gender triggers on computer science (CS) undergraduates that engaged in a simulated salary negotiation with an automated recruiter. We find that, regardless of gender, CS students are reluctant to negotiate, and this hesitancy likely lowers their starting salary. Even when they negotiate, students show little skill in discovering tradeoffs that could enhance their salary, highlighting the need for negotiation training in technical fields. Most importantly, we find little evidence that gender triggers impact women's negotiated outcomes, at least within the field of CS. We argue that findings that emphasize women's individual deficits may reflect a lack of experimental control, which intelligent agents can help correct, and that structural and institutional explanations of inequity deserve greater attention.

This demonstration showcases a virtual agent, Mr. Clue, capable of acting in the role of clue-giver in a word-guessing game. The agent has the ability to automatically generate clues and update its dialogue policy dynamically based on user input.

We apply Reinforcement Learning (RL) to the problem of incremental dialogue policy learning in the context of a fast-paced dialogue game. We compare the policy learned by RL with a high-performance baseline policy which has been shown to perform very efficiently (nearly as well as humans) in this dialogue game. The RL policy outperforms the baseline policy in offline simulations (based on real user data). We provide a detailed comparison of the RL policy and the baseline policy, including information about how much effort and time it took to develop each one of them. We also highlight the cases where the RL policy performs better, and show that understanding the RL policy can provide valuable insights which can inform the creation of an even better rule-based policy.

Although negotiation is an integral part of daily life, most people are unskilled negotiators. To improve one's skill set, a range of costly options including self-study guides, courses, and training programs are offered by various companies and educational institutions. For those who can't afford costly training options, virtual role playing agents offer a low-cost alternative. To be effective, these systems must allow students to engage in experiential learning exercises and provide personalized feedback on the learner's performance. In this paper, we show how a number of negotiation principles can be formalized and quantified. We then establish the pedagogical relevance of several automatic metrics, and show that these metrics are significantly correlated with negotiation outcomes in a human-agent negotiation. This illustrates the realism and helps to validate these principles. It also shows the potential of technology being used to quantify feedback that is traditionally provided through more qualitative approaches. The metrics we describe can provide students with personalized feedback on the errors they make in a negotiation exercise and thereby support guided experiential learning.

Spoken dialogue researchers have recently demonstrated highly interactive systems in several domains. This paper considers how to build on these advances to make systems more robust, easier to develop, and more scientifically significant. We identify key challenges whose solution would lead to improvements in dialogue systems and beyond.

Real-world scenes typically have complex structure, and utterances about them consequently do as well.We devise and evaluate a model that processes descriptions of complex configurations of geometric shapes and can identify the described scenes among a set of candidates, including similar distractors.The model works with raw images of scenes, and by design can work word-by-word incrementally.Hence, it can be used in highly-responsive interactive and situated settings.Using a corpus of descriptions from game-play between human subjects (who found this to be a challenging task), we show that reconstruction of description structure in our system contributes to task success and supports the performance of the word-based model of grounded semantics that we use.

In this paper, we present and evaluate an approach to incremental dialogue act (DA) segmentation and classification.Our approach utilizes prosodic, lexico-syntactic and contextual features, and achieves an encouraging level of performance in offline corpus-based evaluation as well as in simulated human-agent dialogues.Our approach uses a pipeline of sequential processing steps, and we investigate the contribution of different processing steps to DA segmentation errors.We present our results using both existing and new metrics for DA segmentation.The incremental DA segmentation capability described here may help future systems to allow more natural speech from users and enable more natural patterns of interaction.

PentoRef is a corpus of task-oriented dialogues collected in systematically manipulated settings. The corpus is multilingual, with English and German sections, and overall comprises more than 20000 utterances. The dialogues are fully transcribed and annotated with referring expressions mapped to objects in corresponding visual scenes, which makes the corpus a rich resource for research on spoken referring expressions in generation and resolution. The corpus includes several sub-corpora that correspond to different dialogue situations where parameters related to interactivity, visual access, and verbal channel have been manipulated in systematic ways. The corpus thus lends itself to very targeted studies of reference in spontaneous dialogue.

This article examines the potential for teaching negotiation with virtual humans. Many people find negotiations to be aversive. We conjecture that students may be more comfortable practicing negotiation skills with an agent than with another person. We test this using the Conflict Resolution Agent, a semi-automated virtual human that negotiates with people via natural language. In a between-participants design, we independently manipulated two pedagogically-relevant factors while participants engaged in repeated negotiations with the agent: perceived agency (participants either believed they were negotiating with a computer program or another person) and pedagogical feedback (participants received instructional advice or no advice between negotiations). Findings indicate that novice negotiators were more comfortable negotiating with a computer program (they self-reported more comfort and punished their opponent less often) and expended more effort on the exercise following instructional feedback (both in time spent and in self-reported effort). These findings lend support to the notion of using virtual humans to teach interpersonal skills.

This paper presents and analyzes an approach to crowd-sourced spoken dialogue data collection. Our approach enables low cost collection of browser-based spoken dialogue interactions between two remote human participants (human-human condition) as well as one remote human participant and an automated dialogue system (human-agent condition). We present a case study in which 200 remote participants were recruited to participate in a fast-paced image matching game, and which included both human-human and human-agent conditions. We discuss several technical challenges encountered in achieving this crowd-sourced data collection, and analyze the costs in time and money of carrying out the study. Our results suggest the potential of crowdsourced spoken dialogue data to lower costs and facilitate a range of research in dialogue modeling, dialogue system design, and system evaluation.

This paper introduces Eve, a highperformance agent that plays a fast-paced image matching game in a spoken dialogue with a human partner. The agent can be optimized and operated in three different modes of incremental speech processing that optionally include incremental speech recognition, language understanding, and dialogue policies. We present our framework for training and evaluating the agent’s dialogue policies. In a user study involving 125 human participants, we evaluate three incremental architectures against each other and also compare their performance to human-human gameplay. Our study reveals that the most fully incremental agent achieves game scores that are comparable to those achieved in human-human gameplay, are higher than those achieved by partially and nonincremental versions, and are accompanied by improved user perceptions of efficiency, understanding of speech, and naturalness of interaction.

We present the SimSensei system, a fully automatic virtual agent that conducts interviews to assess indicators of psychological distress. With this demo, we focus our attention on the perception part of the system, a multimodal framework which captures and analyzes user state behavior for both behavioral understanding and interactional purposes. We will demonstrate real-time user state sensing as a part of the SimSensei architecture and discuss how this technology enabled automatic analysis of behaviors related to psychological distress.

In this paper we assess our progress toward creating a virtual human negotiation agent with fluid turn-taking skills. To facilitate the design of this agent, we have collected a corpus of human-human negotiation roleplays as well as a corpus of Wizard-controlled human-agent negotiations in the same roleplay scenario. We compare the natural turn-taking behavior in our human-human corpus with that achieved in our Wizard-of-Oz corpus, and quantify our virtual human's turn-taking skills using a combination of subjective and objective metrics. We also discuss our design for a Wizard user interface to support real-time control of the virtual human's turn-taking and dialogue behavior, and analyze our wizard's usage of this interface.

Systems capable of highly-interactive dialog have recently been developed in several domains. This paper considers how to build on these successes to make systems more robust, easier to develop, more adaptable, and more scientifically significant.

We argue for the importance of negotiation as a challenge problem for virtual human research, and introduce a virtual conversational agent that allows people to practice a wide range of negotiation skills. We describe the multi-issue bargaining task, which has become a de facto standard for teaching and research on negotiation in both the social and computer sciences. This task is popular as it allows scientists or instructors to create a variety of distinct situations that arise in real-life negotiations, simply by manipulating a small number of mathematical parameters. We describe the development of a virtual human that will allow students to practice the interpersonal skills they need to recognize and navigate these situations. An evaluation of an early wizard-controlled version of the system demonstrates the promise of this technology for teaching negotiation and supporting scientific research on social intelligence.

This paper presents a multimodal corpus of spoken human-human dialogues collected as participants played a series of Rapid Dialogue Games (RDGs). The corpus consists of a collection of about 11 hours of spoken audio, video, and Microsoft Kinect data taken from 384 game interactions (dialogues). The games used for collecting the corpus required participants to give verbal descriptions of linguistic expressions or visual images and were specifically designed to engage players in a fast-paced conversation under time pressure. As a result, the corpus contains many examples of participants attempting to communicate quickly in specific game situations, and it also includes a variety of spontaneous conversational phenomena such as hesitations, filled pauses, overlapping speech, and low-latency responses. The corpus has been created to facilitate research in incremental speech processing for spoken dialogue systems. Potentially, the corpus could be used in several areas of speech and language research, including speech recognition, natural language understanding, natural language generation, and dialogue management.

We present a computational model of incremental grounding, including state updates and action selection. The model is inspired by corpus-based examples of overlapping utterances of several sorts, including backchannels and completions. The model has also been partially implemented within a virtual human system that includes incremental understanding, and can be used to track grounding and provide overlapping verbal and non-verbal behaviors from a listener, before a speaker has completed her utterance.