Learning Data-Driven Models of Non-Verbal Behaviors for Building Rapport Using an Intelligent Virtual Agent

There is a growing societal need to address the increasing prevalence of behavioral health issues, such as obesity, alcohol or drug use, and general lack of treatment adherence for a variety of health problems. The statistics, worldwide and in the USA, are daunting. Excessive alcohol use is the third leading preventable cause of death in the United States (with 79,000 deaths annually), and is responsible for a wide range of health and social problems. On the positive side though, these behavioral health issues (and associated possible diseases) can often be prevented with relatively simple lifestyle changes, such as losing weight with a diet and/or physical exercise, or learning how to reduce alcohol consumption. Medicine has therefore started to move toward finding ways of preventively promoting wellness, rather than solely treating already established illness. Evidence-based patient-centered Brief Motivational Interviewing (BMI) interven- tions have been found particularly effective in helping people find intrinsic motivation to change problem behaviors after short counseling sessions, and to maintain healthy lifestyles over the long-term. Lack of locally available personnel well-trained in BMI, however, often limits access to successful interventions for people in need. To fill this accessibility gap, Computer-Based Interventions (CBIs) have started to emerge. Success of the CBIs, however, critically relies on insuring engagement and retention of CBI users so that they remain motivated to use these systems and come back to use them over the long term as necessary. Because of their text-only interfaces, current CBIs can therefore only express limited empathy and rapport, which are the most important factors of health interventions. Fortunately, in the last decade, computer science research has progressed in the design of simulated human characters with anthropomorphic communicative abilities. Virtual characters interact using humans’ innate communication modalities, such as facial expressions, body language, speech, and natural language understanding. By advancing research in Artificial Intelligence (AI), we can improve the ability of artificial agents to help us solve CBI problems. To facilitate successful communication and social interaction between artificial agents and human partners, it is essential that aspects of human social behavior, especially empathy and rapport, be considered when designing human-computer interfaces. Hence, the goal of the present dissertation is to provide a computational model of rapport to enhance an artificial agent’s social behavior, and to provide an experimental tool for the psychological theories shaping the model. Parts of this thesis were already published in [LYL+12, AYL12, AL13, ALYR13, LAYR13, YALR13, ALY14].

Using structural equation modeling to assess two theoretical models that explain dropout

This study evaluated the relative fit of both Finn's (1989) Participation-Identification and Wehlage, Rutter, Smith, Lesko and Fernandez's (1989) School Membership models of high school completion to a sample of 4,597 eighth graders taken from the National Educational Longitudinal Study of 1988, (NELS:88), utilizing structural equation modeling techniques. This study found support for the importance of educational engagement as a factor in understanding academic achievement. The Participation-Identification model was particularly well fitting when applied to the sample of high school completers, dropouts (both overall and White dropouts) and African-American students. This study also confirmed the contribution of school environmental factors (i.e., size, diversity of economic and ethnic status among students) and family resources (i.e., availability of learning resources in the home and parent educational level) to students' educational engagement. Based on these findings, school social workers will need to be more attentive to utilizing macro-level interventions (i.e., community organization, interagency coordination) to achieve the organizational restructuring needed to address future challenges. The support found for the Participation-Identification model supports a shift in school social workers' attention from reactive attempts to improve the affective-interpersonal lives of students to proactive attention to their academic lives. The model concentrates school social work practices on the central mission of schools, which is educational engagement. School social workers guided by this model would be encouraged to seek changes in school policies and organization that would facilitate educational engagement. ^

Learning data-driven models of non-verbal behaviors for building rapport using an intelligent virtual agent

There is a growing societal need to address the increasing prevalence of behavioral health issues, such as obesity, alcohol or drug use, and general lack of treatment adherence for a variety of health problems. The statistics, worldwide and in the USA, are daunting. Excessive alcohol use is the third leading preventable cause of death in the United States (with 79,000 deaths annually), and is responsible for a wide range of health and social problems. On the positive side though, these behavioral health issues (and associated possible diseases) can often be prevented with relatively simple lifestyle changes, such as losing weight with a diet and/or physical exercise, or learning how to reduce alcohol consumption. Medicine has therefore started to move toward finding ways of preventively promoting wellness, rather than solely treating already established illness.^ Evidence-based patient-centered Brief Motivational Interviewing (BMI) interventions have been found particularly effective in helping people find intrinsic motivation to change problem behaviors after short counseling sessions, and to maintain healthy lifestyles over the long-term. Lack of locally available personnel well-trained in BMI, however, often limits access to successful interventions for people in need. To fill this accessibility gap, Computer-Based Interventions (CBIs) have started to emerge. Success of the CBIs, however, critically relies on insuring engagement and retention of CBI users so that they remain motivated to use these systems and come back to use them over the long term as necessary.^ Because of their text-only interfaces, current CBIs can therefore only express limited empathy and rapport, which are the most important factors of health interventions. Fortunately, in the last decade, computer science research has progressed in the design of simulated human characters with anthropomorphic communicative abilities. Virtual characters interact using humans’ innate communication modalities, such as facial expressions, body language, speech, and natural language understanding. By advancing research in Artificial Intelligence (AI), we can improve the ability of artificial agents to help us solve CBI problems.^ To facilitate successful communication and social interaction between artificial agents and human partners, it is essential that aspects of human social behavior, especially empathy and rapport, be considered when designing human-computer interfaces. Hence, the goal of the present dissertation is to provide a computational model of rapport to enhance an artificial agent’s social behavior, and to provide an experimental tool for the psychological theories shaping the model. Parts of this thesis were already published in [LYL+12, AYL12, AL13, ALYR13, LAYR13, YALR13, ALY14].^

Improving resource management in virtualized data centers using application performance models

The rapid growth of virtualized data centers and cloud hosting services is making the management of physical resources such as CPU, memory, and I/O bandwidth in data center servers increasingly important. Server management now involves dealing with multiple dissimilar applications with varying Service-Level-Agreements (SLAs) and multiple resource dimensions. The multiplicity and diversity of resources and applications are rendering administrative tasks more complex and challenging. This thesis aimed to develop a framework and techniques that would help substantially reduce data center management complexity.^ We specifically addressed two crucial data center operations. First, we precisely estimated capacity requirements of client virtual machines (VMs) while renting server space in cloud environment. Second, we proposed a systematic process to efficiently allocate physical resources to hosted VMs in a data center. To realize these dual objectives, accurately capturing the effects of resource allocations on application performance is vital. The benefits of accurate application performance modeling are multifold. Cloud users can size their VMs appropriately and pay only for the resources that they need; service providers can also offer a new charging model based on the VMs performance instead of their configured sizes. As a result, clients will pay exactly for the performance they are actually experiencing; on the other hand, administrators will be able to maximize their total revenue by utilizing application performance models and SLAs. ^ This thesis made the following contributions. First, we identified resource control parameters crucial for distributing physical resources and characterizing contention for virtualized applications in a shared hosting environment. Second, we explored several modeling techniques and confirmed the suitability of two machine learning tools, Artificial Neural Network and Support Vector Machine, to accurately model the performance of virtualized applications. Moreover, we suggested and evaluated modeling optimizations necessary to improve prediction accuracy when using these modeling tools. Third, we presented an approach to optimal VM sizing by employing the performance models we created. Finally, we proposed a revenue-driven resource allocation algorithm which maximizes the SLA-generated revenue for a data center.^

Improving Resource Management in Virtualized Data Centers using Application Performance Models

The rapid growth of virtualized data centers and cloud hosting services is making the management of physical resources such as CPU, memory, and I/O bandwidth in data center servers increasingly important. Server management now involves dealing with multiple dissimilar applications with varying Service-Level-Agreements (SLAs) and multiple resource dimensions. The multiplicity and diversity of resources and applications are rendering administrative tasks more complex and challenging. This thesis aimed to develop a framework and techniques that would help substantially reduce data center management complexity. We specifically addressed two crucial data center operations. First, we precisely estimated capacity requirements of client virtual machines (VMs) while renting server space in cloud environment. Second, we proposed a systematic process to efficiently allocate physical resources to hosted VMs in a data center. To realize these dual objectives, accurately capturing the effects of resource allocations on application performance is vital. The benefits of accurate application performance modeling are multifold. Cloud users can size their VMs appropriately and pay only for the resources that they need; service providers can also offer a new charging model based on the VMs performance instead of their configured sizes. As a result, clients will pay exactly for the performance they are actually experiencing; on the other hand, administrators will be able to maximize their total revenue by utilizing application performance models and SLAs. This thesis made the following contributions. First, we identified resource control parameters crucial for distributing physical resources and characterizing contention for virtualized applications in a shared hosting environment. Second, we explored several modeling techniques and confirmed the suitability of two machine learning tools, Artificial Neural Network and Support Vector Machine, to accurately model the performance of virtualized applications. Moreover, we suggested and evaluated modeling optimizations necessary to improve prediction accuracy when using these modeling tools. Third, we presented an approach to optimal VM sizing by employing the performance models we created. Finally, we proposed a revenue-driven resource allocation algorithm which maximizes the SLA-generated revenue for a data center.

Development of methods for spatial modeling of sediments quality in Lake Okeechobee, Florida

The major objectives of this dissertation were to develop optimal spatial techniques to model the spatial-temporal changes of the lake sediments and their nutrients from 1988 to 2006, and evaluate the impacts of the hurricanes occurred during 1998–2006. Mud zone reduced about 10.5% from 1988 to 1998, and increased about 6.2% from 1998 to 2006. Mud areas, volumes and weight were calculated using validated Kriging models. From 1988 to 1998, mud thicknesses increased up to 26 cm in the central lake area. The mud area and volume decreased about 13.78% and 10.26%, respectively. From 1998 to 2006, mud depths declined by up to 41 cm in the central lake area, mud volume reduced about 27%. Mud weight increased up to 29.32% from 1988 to 1998, but reduced over 20% from 1998 to 2006. The reduction of mud sediments is likely due to re-suspension and redistribution by waves and currents produced by large storm events, particularly Hurricanes Frances and Jeanne in 2004 and Wilma in 2005. Regression, kriging, geographically weighted regression (GWR) and regression-kriging models have been calibrated and validated for the spatial analysis of the sediments TP and TN of the lake. GWR models provide the most accurate predictions for TP and TN based on model performance and error analysis. TP values declined from an average of 651 to 593 mg/kg from 1998 to 2006, especially in the lake’s western and southern regions. From 1988 to 1998, TP declined in the northern and southern areas, and increased in the central-western part of the lake. The TP weights increased about 37.99%–43.68% from 1988 to 1998 and decreased about 29.72%–34.42% from 1998 to 2006. From 1988 to 1998, TN decreased in most areas, especially in the northern and southern lake regions; western littoral zone had the biggest increase, up to 40,000 mg/kg. From 1998 to 2006, TN declined from an average of 9,363 to 8,926 mg/kg, especially in the central and southern regions. The biggest increases occurred in the northern lake and southern edge areas. TN weights increased about 15%–16.2% from 1988 to 1998, and decreased about 7%–11% from 1998 to 2006.

Using extreme value theory to value stock market returns

Extreme stock price movements are of great concern to both investors and the entire economy. For investors, a single negative return, or a combination of several smaller returns, can possible wipe out so much capital that the firm or portfolio becomes illiquid or insolvent. If enough investors experience this loss, it could shock the entire economy. An example of such a case is the stock market crash of 1987. Furthermore, there has been a lot of recent interest regarding the increasing volatility of stock prices. ^ This study presents an analysis of extreme stock price movements. The data utilized was the daily returns for the Standard and Poor's 500 index from January 3, 1978 to May 31, 2001. Research questions were analyzed using the statistical models provided by extreme value theory. One of the difficulties in examining stock price data is that there is no consensus regarding the correct shape of the distribution function generating the data. An advantage with extreme value theory is that no detailed knowledge of this distribution function is required to apply the asymptotic theory. We focus on the tail of the distribution. ^ Extreme value theory allows us to estimate a tail index, which we use to derive bounds on the returns for very low probabilities on an excess. Such information is useful in evaluating the volatility of stock prices. There are three possible limit laws for the maximum: Gumbel (thick-tailed), Fréchet (thin-tailed) or Weibull (no tail). Results indicated that extreme returns during the time period studied follow a Fréchet distribution. Thus, this study finds that extreme value analysis is a valuable tool for examining stock price movements and can be more efficient than the usual variance in measuring risk. ^

A methodology for formally modeling and analyzing software architecture of mobile agent systems

A methodology for formally modeling and analyzing software architecture of mobile agent systems provides a solid basis to develop high quality mobile agent systems, and the methodology is helpful to study other distributed and concurrent systems as well. However, it is a challenge to provide the methodology because of the agent mobility in mobile agent systems.^ The methodology was defined from two essential parts of software architecture: a formalism to define the architectural models and an analysis method to formally verify system properties. The formalism is two-layer Predicate/Transition (PrT) nets extended with dynamic channels, and the analysis method is a hierarchical approach to verify models on different levels. The two-layer modeling formalism smoothly transforms physical models of mobile agent systems into their architectural models. Dynamic channels facilitate the synchronous communication between nets, and they naturally capture the dynamic architecture configuration and agent mobility of mobile agent systems. Component properties are verified based on transformed individual components, system properties are checked in a simplified system model, and interaction properties are analyzed on models composing from involved nets. Based on the formalism and the analysis method, this researcher formally modeled and analyzed a software architecture of mobile agent systems, and designed an architectural model of a medical information processing system based on mobile agents. The model checking tool SPIN was used to verify system properties such as reachability, concurrency and safety of the medical information processing system. ^ From successful modeling and analyzing the software architecture of mobile agent systems, the conclusion is that PrT nets extended with channels are a powerful tool to model mobile agent systems, and the hierarchical analysis method provides a rigorous foundation for the modeling tool. The hierarchical analysis method not only reduces the complexity of the analysis, but also expands the application scope of model checking techniques. The results of formally modeling and analyzing the software architecture of the medical information processing system show that model checking is an effective and an efficient way to verify software architecture. Moreover, this system shows a high level of flexibility, efficiency and low cost of mobile agent technologies. ^

Anxiety sensitivity's facets in relation to anxious and depressive symptoms in youth

Anxiety sensitivity is a multifaceted cognitive risk factor currently being examined in relation to anxiety and depression. The paucity of research on the relative contribution of the facets of anxiety sensitivity to anxiety and depression, coupled with variations in existing findings, indicate that the relations remain inadequately understood. In the present study, the relations between the facets of anxiety sensitivity, anxiety, and depression were examined in 730 Hispanic-Latino and European-American youth referred to an anxiety specialty clinic. Youth completed the Childhood Anxiety Sensitivity Index, the Revised Children’s Manifest Anxiety Scale, and the Children’s Depression Inventory. The factor structure of the Childhood Anxiety Sensitivity Index was examined using ordered-categorical confirmatory factor analytic techniques. Goodness-of-fit criteria indicated that a two-factor model fit the data best. The identified facets of anxiety sensitivity included Physical/Mental Concerns and Social Concerns. Support was also found for cross-ethnic equivalence of the two-factor model across Hispanic-Latino and European-American youth. Structural equation modeling was used to examine models involving anxiety sensitivity, anxiety, and depression. Results indicated that an overall measure of anxiety sensitivity was positively associated with both anxiety and depression, while the facets of anxiety sensitivity showed differential relations to anxiety and depression symptoms. Both facets of anxiety sensitivity were related to overall anxiety and its symptom dimensions, with the exception being that Social Concerns was not related to physiological anxiety symptoms. Physical/Mental Concerns were strongly associated with overall depression and with all depression symptom dimensions. Social Concerns was not significantly associated with depression or its symptom dimensions. These findings highlight that anxiety sensitivity’s relations to youth psychiatric symptoms are complex. Results suggest that focusing on anxiety sensitivity’s facets is important to fully understand its role in psychopathology. Clinicians may want to target all facets of anxiety sensitivity when treating anxious youth. However, in the context of depression, it might be sufficient for clinicians to target Physical/Mental Incapacitation Concerns.

Three empirical essays on startups' survival using the Kauffman Firm Survey

This dissertation studies newly founded U.S. firms' survival using three different releases of the Kauffman Firm Survey. I study firms' survival from a different perspective in each chapter. ^ The first essay studies firms' survival through an analysis of their initial state at startup and the current state of the firms as they gain maturity. The probability of survival is determined using three probit models, using both firm-specific variables and an industry scale variable to control for the environment of operation. The firm's specific variables include size, experience and leverage as a debt-to-value ratio. The results indicate that size and relevant experience are both positive predictors for the initial and current states. Debt appears to be a predictor of exit if not justified wisely by acquiring assets. As suggested previously in the literature, entering a smaller-scale industry is a positive predictor of survival from birth. Finally, a smaller-scale industry diminishes the negative effects of debt. ^ The second essay makes use of a hazard model to confirm that new service-providing (SP) firms are more likely to survive than new product providers (PPs). I investigate the possible explanations for the higher survival rate of SPs using a Cox proportional hazard model. I examine six hypotheses (variations in capital per worker, expenses per worker, owners' experience, industry wages, assets and size), none of which appear to explain why SPs are more likely than PPs to survive. Two other possibilities are discussed: tax evasion and human/social relations, but these could not be tested due to lack of data. ^ The third essay investigates women-owned firms' higher failure rates using a Cox proportional hazard on two models. I make use of a never-before used variable that proxies for owners' confidence. This variable represents the owners' self-evaluated competitive advantage. ^ The first empirical model allows me to compare women's and men's hazard rates for each variable. In the second model I successively add the variables that could potentially explain why women have a higher failure rate. Unfortunately, I am not able to fully explain the gender effect on the firms' survival. Nonetheless, the second empirical approach allows me to confirm that social and psychological differences among genders are important in explaining the higher likelihood to fail in women-owned firms.^

Transforming Participation: A Case Study of an Introductory Physics Student in a Modeling Instruction Class

We present a case study on how participation of one student changed during her first semester of introductory physics class using Modeling Instruction. Using video recordings, we explore how her behavior is consistent with a change from thinking of group learning as a parallel activity to one that is collaborative.

Development of effective approaches to the large-scale aerodynamic testing of low-rise buildings

Low-rise buildings are often subjected to high wind loads during hurricanes that lead to severe damage and cause water intrusion. It is therefore important to estimate accurate wind pressures for design purposes to reduce losses. Wind loads on low-rise buildings can differ significantly depending upon the laboratory in which they were measured. The differences are due in large part to inadequate simulations of the low-frequency content of atmospheric velocity fluctuations in the laboratory and to the small scale of the models used for the measurements. A new partial turbulence simulation methodology was developed for simulating the effect of low-frequency flow fluctuations on low-rise buildings more effectively from the point of view of testing accuracy and repeatability than is currently the case. The methodology was validated by comparing aerodynamic pressure data for building models obtained in the open-jet 12-Fan Wall of Wind (WOW) facility against their counterparts in a boundary-layer wind tunnel. Field measurements of pressures on Texas Tech University building and Silsoe building were also used for validation purposes. The tests in partial simulation are freed of integral length scale constraints, meaning that model length scales in such testing are only limited by blockage considerations. Thus the partial simulation methodology can be used to produce aerodynamic data for low-rise buildings by using large-scale models in wind tunnels and WOW-like facilities. This is a major advantage, because large-scale models allow for accurate modeling of architectural details, testing at higher Reynolds number, using greater spatial resolution of the pressure taps in high pressure zones, and assessing the performance of aerodynamic devices to reduce wind effects. The technique eliminates a major cause of discrepancies among measurements conducted in different laboratories and can help to standardize flow simulations for testing residential homes as well as significantly improving testing accuracy and repeatability. Partial turbulence simulation was used in the WOW to determine the performance of discontinuous perforated parapets in mitigating roof pressures. The comparisons of pressures with and without parapets showed significant reductions in pressure coefficients in the zones with high suctions. This demonstrated the potential of such aerodynamic add-on devices to reduce uplift forces.

A middleware to support services delivery in a domain-specific virtual machine

The increasing use of model-driven software development has renewed emphasis on using domain-specific models during application development. More specifically, there has been emphasis on using domain-specific modeling languages (DSMLs) to capture user-specified requirements when creating applications. The current approach to realizing these applications is to translate DSML models into source code using several model-to-model and model-to-code transformations. This approach is still dependent on the underlying source code representation and only raises the level of abstraction during development. Experience has shown that developers will many times be required to manually modify the generated source code, which can be error-prone and time consuming. ^ An alternative to the aforementioned approach involves using an interpreted domain-specific modeling language (i-DSML) whose models can be directly executed using a Domain Specific Virtual Machine (DSVM). Direct execution of i-DSML models require a semantically rich platform that reduces the gap between the application models and the underlying services required to realize the application. One layer in this platform is the domain-specific middleware that is responsible for the management and delivery of services in the specific domain. ^ In this dissertation, we investigated the problem of designing the domain-specific middleware of the DSVM to facilitate the bifurcation of the semantics of the domain and the model of execution (MoE) while supporting runtime adaptation and validation. We approached our investigation by seeking solutions to the following sub-problems: (1) How can the domain-specific knowledge (DSK) semantics be separated from the MoE for a given domain? (2) How do we define a generic model of execution (GMoE) of the middleware so that it is adaptable and realizes DSK operations to support delivery of services? (3) How do we validate the realization of DSK operations at runtime? ^ Our research into the domain-specific middleware was done using an i-DSML for the user-centric communication domain, Communication Modeling Language (CML), and for microgrid energy management domain, Microgrid Modeling Language (MGridML). We have successfully developed a methodology to separate the DSK and GMoE of the middleware of a DSVM that supports specialization for a given domain, and is able to perform adaptation and validation at runtime. ^