We present an approach that uses learning from demonstration in a computer role playing game to create a controller for a companion team member. We describe a behavior engine that uses case-based reasoning. The behavior engine accepts observation traces of human playing decisions and produces a sequence of actions which can then be carried out by an artificial agent within the gaming environment. Our work focuses on team-based role playing games, where the agents produced by the behavior engine act as team members within a mixed human-agent team. We present the results of a study we conducted, where we assess both the quantitative and qualitative performance difference between human-only teams compared with hybrid human-agent teams. The results of our study show that human-agent teams were more successful at task completion and, for some qualitative dimensions, hybrid teams were perceived more favorably than human-only teams.

This paper focuses on case acquisition strategies in the context of Case-based Learning from Observation (CBLfO). In Learning from Observation (LfO), a system learns behaviors by observing an expert rather than being explicitly programmed. Specifically, we focus on the problem of learning behaviors from experts that reason using internal state information, that is, information that can not be directly observed. The unobservability of this state information means that the behaviors can not be represented by a simple perception-to-action mapping. We propose a new case acquisition strategy called "Similarity-based Chunking", and compare it with existing strategies to address this problem. Additionally, since standard classification accuracy in predicting the expert's actions is known to be a poor measure for evaluating LfO systems, we propose a new evaluation procedure based on two complementary metrics: behavior performance and similarity with the expert.

Instance-based regression methods generate solutions from prior solutions within a neighborhood of the input query. Their performance depends on both neighborhood selection criteria and on the method for generating new solutions from the values of prior instances. This paper proposes a new approach to addressing both problems, in which solutions are generated by an ensemble of solutions of local linear regression models built for a collection of "stretched" neighborhoods of the query. Each neighborhood is generated by relaxing a different dimension of the problem space. The rationale is to enable major change trends along that dimension to have increased influence on the corresponding model. The approach is evaluated for two candidate relaxation approaches, gradient-based and based on fixed profiles, and compared to baselines of k-NN and using a radius-based spherical neighborhood in n-dimensional space. Results in four test domains show up to 15 percent improvement over baselines, and suggest that the approach could be particularly useful in domains for which the space of prior instances is sparse.

Traces of user behaviour can be a valuable source of knowledge that can be used during case-based reasoning. This paper presents an approach for analyzing and cleaning user traces. The analysis looks to identify three properties in traces: reasoning with an internal state, non-deterministic behaviour and error. The existence of any of these properties may influence how a system should reason or store knowledge in cases. Initially, each trace is examined to see areas that might contain one of the three properties. Multiple versions of the trace are then generated in order to determine which specific property is present. The analysis is applied to traces generated by observing both a computer and human controller for an obstacle avoidance robot. The results demonstrate that the analysis is able to successfully identify which properties are present and clean many of the errors that exist in the traces.

This paper addresses Trace-Based Reasoning (TBR) by using Case-Based Reasoning (CBR) as a descriptive framework. TBR is a reasoning paradigm in which inferences are made on specific objects called traces. Traces are sequential records of events observed and stored during an interactive process. We report two contributions. First, we propose a review of the current researches related to TBR. Then, we compare CBR and TBR. From this comparison, we show that the exploitation of traces instead of cases as knowledge sources raises very specific challenges. More precisely, new methods for defining similarity measures and for performing adaptation of traces are required. These new methods have to take into account the sequential properties of traces. We emphasis the benefits of using traces as a knowledge container in a reasoning process and we pinpoint promising applications of TBR.

Introduction to the papers presented at the 2013 FLAIRS Conference Special Track on Case-Based Reasoning

One of the major challenges in bioinformatics is selecting the appropriate genes for a given problem, and moreover, choosing the best gene selection technique for this task. Many such techniques have been developed, each with its own characteristics and complexities. Recently, some works have addressed this by introducing ensemble gene selection, which is the process of performing multiple runs of gene selection and aggregating the results into a single final list. The question is, will ensemble gene selection improve the results over those obtained when using single gene selection techniques (e.g., filter-based gene selection techniques on their own without any ensemble approach)? We compare how five filter-based feature (gene) selection techniques work with and without a data diversity ensemble approach (using a single feature selection technique on multiple sampled datasets created from an original one) when used for building models to label cancerous cells (or predict cancer treatment response) based on gene expression levels. Eleven bioinformatics (gene microarray) datasets are employed, along with four feature subset sizes and five learners. Our results show that the techniques Fold Change Ratio and Information Gain will produce better classification results when an ensemble approach is applied, while Probability Ratio and Signal-to-Noise will, in general, perform better without the ensemble approach. For the Area Under the ROC (Receiver Operating Characteristics) Curve ranker, the classification results are similar with or without the ensemble approach. This is, to our knowledge, the first paper to comprehensively examine the difference between the ensemble and single approaches for gene selection in the biomedical and bioinformatics domains.

National Cancer Institute's (NCI) Cancer Information Service (CIS) offers online instant messaging based information service called LiveHelp to patients, family members, friends, and other cancer information consumers. A cancer information specialist (IS) 'chats' with a consumer and provides information on a variety of topics including clinical trials. After a LiveHelp chat session is finished, the IS codes about 20 different elements of metadata about the session in electronic contact record forms (ECRF), which are to be later used for quality control and reporting. Besides straightforward elements like age and gender, more specific elements to be coded include the purpose of contact, the subjects of interaction, and the different responses provided to the consumer, the latter two often taking on multiple values. As such, ECRF coding is a time consuming task and automating this process could help ISs to focus more on their primary goal of helping consumers with valuable cancer related information. As a first attempt in this task, we explored multi-label and multi-class text classification approaches to code the purpose, subjects of interaction, and the responses provided based on the chat transcripts. With a sample dataset of about 673 transcripts, we achieved example-based F-scores of 0.67 (for subjects) and 0.58 (responses). We also achieved label-based micro F-scores of 0.65 (for subjects), 0.62 (for responses), and 0.61 (for purpose). To our knowledge this is the first attempt in automatic coding of LiveHelp transcripts and our initial results on the smaller corpus indicate promising future directions in this task.

In this article, we present ongoing research, EMO, an affective embodied conversational agent platform, aimed at depicting multi-ethnic, multi-modal communication patterns in a credible manner. We employ the methodology of integrating counseling concepts early in the design to effectively target a specific domain. The system is geared to augment solution focused therapy. We present a prototype of the architecture as proof of concept and evaluate the platform for affect portrayal. 

Introduction to the papers presented at the 2013 FLAIRS Conference Special Track on Artificial Intelligence in 
Healthcare Informatics.