Social, Cultural, and Behavioral Modeling

Frontmatter

Analyzing the Dabiq Magazine: The Language and the Propaganda Structure of ISIS

Abstract

The Islamic State of Iraq and Sham (ISIS) still poses a significant concern worldwide due to its brutal attacks and unconventional recruitment strategy despite its recent defeat and loss of territory. ISIS distinguished itself from other notorious terrorist organizations regarding Techniques, Tactics, and Procedures (TTP). It has been observed that ISIS is highly capable of attracting foreign fighters through its improved “netwar” skills. Whereas its propaganda videos and images have been extensively analyzed, a systematic analysis of textual content is still lacking. Therefore, we examine the Dabig magazine to discover propagandist elements by performing natural language processing (NLP) and text mining methods. Namely, we first automatically detect three types of entities (person, location, organization) in each article for fifteen Dabiq issues. Then we build entity networks based on co-occurrence of entities to observe the entity relationships over time. We further employ topic modeling on all articles and calculate statistics for entities. We observe entities revolve around the term “jihad,” and the ISIS consistently seems to exploit the sources of Islam in their propaganda. The analysis also revealed that ISIS primarily targets Shiites by using derogatory language about their belief system and try to justify their attacks against them.

Halil Bisgin, Hasan Arslan, Yusuf Korkmaz

Characterizing Organizational Micro-climates in Structural Groups

Abstract

We use text to characterize micro-climates in a specific organizational context. We use Louvain clustering to create trees, and identify communities and smaller groups, which we call leaves. In comparing these structural groups, we see that most structural groups within this organization share locations and organizational functions. Our analyses show that location, gender, and minority status dominance tend to increase the strength of the textual distinction of the found structural groups.

Geoffrey P. Morgan, Kathleen M. Carley

Pro/Con: Neural Detection of Stance in Argumentative Opinions

Abstract

Accurate information from both sides of the contemporary issues is known to be an ‘antidote in confirmation bias’. While these types of information help the educators to improve their vital skills including critical thinking and open-mindedness, they are relatively rare and hard to find online. With the well-researched argumentative opinions (arguments) on controversial issues shared by Procon.org in a non-partisan format, detecting the stance of arguments is a crucial step to automate organizing such resources. We use a universal pretrained language model with weight-dropped LSTM neural network to leverage the context of an argument for stance detection on the proposed dataset. Experimental results show that the dataset is challenging, however, utilizing the pretrained language model fine-tuned on context information yields a general model that beats the competitive baselines. We also provide analysis to find the informative segments of an argument to our stance detection model and investigate the relationship between the sentiment of an argument with its stance.

Marjan Hosseinia, Eduard Dragut, Arjun Mukherjee

Modeling Gender Inequity in Household Decision-Making

Abstract

The Food and Agriculture Organization (FAO) estimates that if female farmers in developing countries had access to the same resources as men, the number of undernourished people would decrease by 12%–17% [9]. Clearly, gender equity is a vital part of increasing agricultural production to feed the world’s projected 9.7 billion people by 2050. However, programs designed to empower women in agricultural systems are expensive, and no quantitative model exists to define and explore the efficacy of policies in cultural contexts. We introduce a formal model of household decisions embedded in an agent-based model of community gender dynamics and show how the explicit definition of gender inequity can help inform decision-making about programs intended to empower women.

Allegra A. Beal Cohen, Paul R. Cohen, Gregory Kiker

Bot Detection: Will Focusing on Recall Cause Overall Performance Deterioration?

Abstract

Social bots are an effective tool in the arsenal of malicious actors who manipulate discussions on social media. Bots help spread misinformation, promote political propaganda, and inflate the popularity of users and content. Hence, it is necessary to differentiate bot accounts and human users. There are several bot detection methods that approach this problem. Conventional methods either focus on precision regardless of the overall performance or optimize overall performance, say \(F_1\), without monitoring its effect on precision or recall. Focusing on precision means that those users marked as bots are more likely than not bots but a large portion of the bots could remain undetected. From a user’s perspective, however, it is more desirable to have less interaction with bots, even if it would incur a loss in precision. This can be achieved by a detection method with higher recall. A trivial, but useless, solution for high recall is to classify every account (human or bot) as bot, hence, resulting in poor overall performance.

In this work, we investigate if it is feasible for a method to focus on recall without considerable loss in overall performance. Extensive experiments with recall and precision trade-off suggest that high recall can be achieved without much overall performance deterioration. This research leads to a recall-focused approach to bot detection, REFOCUS, with some lessons learned and future directions.

Tahora H. Nazer, Matthew Davis, Mansooreh Karami, Leman Akoglu, David Koelle, Huan Liu

A Quantitative Portrait of Legislative Change in Ukraine

Abstract

Over the past decade, Ukraine has undergone tremendous socio-political changes, which continue to this day. While such changes may be analyzed and interpreted from a variety of sources, we utilize recent advancements in the quantitative analysis of culture to identify how these changes are encoded within Ukraine’s legislation. Our goal is to provide a new picture of Ukrainian governance that may be used by subject matter experts as a complement to existing forms of political data. To do so, we apply probabilistic topic modeling to compress over a decade of Ukrainian legislation into patterns of word usage. We then apply a recently developed calculation of novelty to measure how different each draft law is from the draft laws which precede it. We find an interesting pattern of legislative changes and identify some of the drivers of these changes. Finally, we discuss the relationship between our results and the broader context of Ukrainian political changes and suggest steps to explore this relationship further.

Zachary K. Stine, Nitin Agarwal

Synthesizing Machine-Learning Datasets from Parameterizable Agents Using Constrained Combinatorial Search

Abstract

The tedious, often hand-modeled, activity of designing and implementing simulation scenarios can benefit from modern-day data-driven methods, i.e., machine-learning (ML). We envision a toolchain that exploits information obtained during live operations, such as the observed maneuvers, techniques, and procedures of all interacting players in live operational settings, that serves as input into an ML-based scenario authoring process. We present a mechanism, called the Parameter Diversifier (PD), that takes a base scenario structure and synthesizes the comprehensive datasets needed for the supervised machine-learning of a scenario authoring model. The design of the PD explores and exploits low-level agent state search space as it relates to it high-level implications at the scenario level. This work demonstrates an explicit sampling of the scenario parameter search space to build an implicit model for use in simulation scenario generation.

Victor Hung, Joshua Haley, Robert Bridgman, Norb Timpko, Robert Wray

Exploiting Emojis for Sarcasm Detection

Abstract

Modern social media platforms largely rely on text. However, the written text lacks the emotional cues of spoken and face-to-face dialogue, ambiguities are common, which is exacerbated in the short, informal nature of many social media posts. Sarcasm represents the nuanced form of language that individuals state the opposite of what is implied. Sarcasm detection on social media is important for users to understand the underlying messages. The majority of existing sarcasm detection algorithms focus on text information; while emotion information expressed such as emojis are ignored. In real scenarios, emojis are widely used as emotion signals, which have great potentials to advance sarcasm detection. Therefore, in this paper, we study the novel problem of exploiting emojis for sarcasm detection on social media. We propose a new framework ESD, which simultaneously captures various signals from text and emojis for sarcasm detection. Experimental results on real-world datasets demonstrate the effectiveness of the proposed framework.

Jayashree Subramanian, Varun Sridharan, Kai Shu, Huan Liu

Condorcet Optimal Clustering with Delaunay Triangulation: Climate Zones and World Happiness Insights

Abstract

Condorcet clustering methods have the attractive features of producing clusterings which place similar points in the same cluster and dissimilar points in different clusters as well as not requiring a priori specification of the number of clusters. They have the disadvantages of being combinatorially hard and the method produces only convex clusters. We propose a novel modification to this method, which improves it significantly on both accounts and works particularly well when applied to social network type data sets. Specifically, we reduce the domain of the clustering to be over a Delaunay triangulation, whose size scales as \(O(n^{\lfloor m/2 \rfloor })\) where n is the number of records and m is the number of attributes used for the clustering. The triangulation also limits focus to local structure, which allows for non-convex clusterings. We demonstrate its use in comparison to other well-known heuristic methods using several constructed datasets, then use it to cluster real-world datasets.

Max Bassett, Blake Newton, Joseph Schlessinger, Jacob Schmidt, Scott Lynch, Patrick Kuiper, Ryan Miller, Steven Morse, James Pleuss, Travis Russell, William Pulleyblank

Using Common Enemy Graphs to Identify Communities of Coordinated Social Media Activity

Abstract

Increased use of and reliance on social media has led to a responsive rise in the creation of automated accounts on such platforms. Recent approaches to identification of individual automated accounts has relied on machine learning methods utilizing features drawn predominantly from text content and profile metadata. In this work we explore a novel use of graph theoretic measures, specifically common enemy graphs, to identify and characterize groups of accounts exhibiting shared behavior in online social media, particularly those exhibiting characteristics of automation and/or potential coordination. In addition, we develop edge weight variants of fuzzy competition graphs to further characterize common group behavior of automated accounts within subnetworks of social media ecosystems.

Lucas A. Overbey, Bryan Ek, Kevin Pinzhoffer, Bryan Williams

Chronological Semantics Modeling: A Topic Evolution Approach in Online User-Generated Medical Data

Abstract

Online medical discussion forums/question answering sites have become one of the major resources for people to look for healthcare information. These sites typically contain tremendous user-generated content (UGC) that possesses complex domain-specific information in layman’s terms, which is the opposite of formal medical records kept in hospitals (i.e. Electronic Health Record). The goal of this project is to dissect semantics and extract valuable information systematically from UGC composed in unstructured and unconstrained format. We propose an automatic medical content analyzer that takes into account language semantics as well as progression (evolution) of medical events. The preliminary evaluation on the WebMD dataset shows that evolution-based recommendation uncovers broader domain semantic which might be ignored when using word-level or concept-based features.

Cheng-Yu Chung, I-Han Hsiao

Massive-Scale Models of Urban Infrastructure and Populations

Abstract

As the world becomes more dense, connected, and complex, it is increasingly difficult to answer “what-if” questions about our cities and populations. Most modeling and simulation tools struggle with scale and connectivity. We present a new method for creating digital twin simulations of city infrastructure and populations from open source and commercial data. We transform cellular location data into activity patterns for synthetic agents and use geospatial data to create the infrastructure and world in which these agents interact. We then leverage technologies and techniques intended for massive online gaming to create 1:1 scale simulations to answer these “what-if” questions about the future.

Daniel Baeder, Eric Christensen, Anhvinh Doanvo, Andrew Han, Ben F. M. Intoy, Steven Hardy, Zachary Humayun, Melissa Kain, Kevin Liberman, Adrian Myers, Meera Patel, William J. Porter III, Lenny Ramos, Michelle Shen, Lance Sparks, Allan Toriel, Benjamin Wu

Dynamic Resource Allocation During Natural Disasters Using Multi-agent Environment

Abstract

Natural disasters are devastating for a country and effective allocation of critical resources can mitigate the impact. While traditional approaches usually have difficulties in making optimal critical resource allocation, in this paper we introduce a novel hierarchical multi-agent reinforcement learning framework to model optimal resource allocation for natural disasters in real-time. On the lower level a set of agents navigate with the continuous time environment using deep reinforcement algorithms. On the higher level, a lead agent takes care of the global decision-making. Our framework achieves more efficient resource allocation in response to dynamic events and is applicable to problems where disaster evolves alongside the response efforts, where delays in response can lead to increased disaster severity and thus a greater need for resources.

Alina Vereshchaka, Wen Dong

Parallelizing Convergent Cross Mapping Using Apache Spark

Abstract

Identifying the causal relationships between subjects or variables remains an important problem across various scientific fields. This is particularly important but challenging in complex systems, such as those involving human behavior, sociotechnical contexts, and natural ecosystems. By exploiting state space reconstruction via lagged embedding of time series, convergent cross mapping (CCM) serves as an important method for addressing this problem. While powerful, CCM is computationally costly; moreover, CCM results are highly sensitive to several parameter values. While best practice entails exploring a range of parameter settings when assessing casual relationships, the resulting computational burden can raise barriers to practical use, especially for long time series exhibiting weak causal linkages. We demonstrate here several means of accelerating CCM by harnessing the distributed Apache Spark platform. We characterize and report on results of several experiments with parallelized solutions that demonstrate high scalability and a capacity for over an order of magnitude performance improvement for the baseline configuration. Such economies in computation time can speed learning and robust identification of causal drivers in complex systems.

Bo Pu, Lujie Duan, Nathaniel D. Osgood

Continuous-Time Simulation of Epidemic Processes on Dynamic Interaction Networks

Abstract

Contagious processes on networks, such as spread of disease through physical proximity or information diffusion over social media, are continuous-time processes that depend upon the pattern of interactions between the individuals in the network. Continuous-time stochastic epidemic models are a natural fit for modeling the dynamics of such processes. However, prior work on such continuous-time models doesn’t consider the dynamics of the underlying interaction network which involves addition and removal of edges over time. Instead, researchers have typically simulated these processes using discrete-time approximations, in which one has to trade off between high simulation accuracy and short computation time. In this paper, we incorporate continuous-time network dynamics (addition and removal of edges) into continuous-time epidemic simulations. We propose a rejection-sampling based approach coupled with the well-known Gillespie algorithm that enables exact simulation of the continuous-time epidemic process. Our proposed approach gives exact results, and the computation time required for simulation is reduced as compared to discrete-time approximations of comparable accuracy.

Rehan Ahmad, Kevin S. Xu

Characterizing Bot Networks on Twitter: An Empirical Analysis of Contentious Issues in the Asia-Pacific

Abstract

This paper empirically analyzes bot activity in contentious Twitter conversations using case studies from the Asia-Pacific. Bot activity is measured and characterized using a series of interoperable tools leveraging dynamic network analysis and machine learning. We apply this novel and flexible methodological framework to derive insights about information operations in three contexts: the senatorial elections in the Philippines, the presidential elections in Indonesia, and the relocation of a military base in Okinawa. Varying levels of bot prevalence and influence are identified across case studies. The presented findings demonstrate principles of social cyber-security in concrete settings and highlight conceptual and methodological issues to inform further development of analytic pipelines in studying online information operations.

Joshua Uyheng, Kathleen M. Carley

A Hybrid Cellular Model for Predicting Organizational Recruitment in a k-Dimensional Space

Abstract

Ecological models are useful in modeling organizations and their competition over resources. However, the traditional approaches, particularly Blau space models, are restrictive in their dependence on a continuous space. In addition, these models are susceptible to indicating competition in sparsely populated areas of an ecology. To deal with these problems we propose a reconceptualization of Blau space that utilizes a cellular structure to model a wider number of variable types, and simple probabilistic urn models to evaluate competition between organizations. We briefly review the basic concepts of Blau Space, demonstrate the issues with traditional Blau space modeling, and present a new model referred to as the Hybrid model.

Nicolas L. Harder, Matthew E. Brashears

A Challenging Dataset for Bias Detection: The Case of the Crisis in the Ukraine

Abstract

The use of disinformation and purposefully biased reportage to sway public opinion has become a serious concern. We present a new dataset related to the Ukrainian Crisis of 2014–2015 which can be used by other researchers to train, test, and compare bias detection algorithms. The dataset comprises 4,538 articles in English related to the crisis from 227 news sources in 43 countries (including the Ukraine) comprising 1.7M words. We manually classified the bias of each article as either pro-Russian, pro-Western, or Neutral, and also aligned each article with a master timeline of 17 major events. When trained on the whole dataset a simple baseline SVM classifier using doc2vec embeddings as features achieves an \(F_{1}\) score of 0.86. This performance is deceptively high, however, because (1) the model is almost completely unable to correctly classify articles published in the Ukraine (0.07 \(F_{1}\)), and (2) the model performs nearly as well when trained on unrelated geopolitics articles written by the same publishers and tested on the dataset. As has been pointed out by other researchers, these results suggest that models of this type are learning journalistic styles rather than actually modeling bias. This implies that more sophisticated approaches will be necessary for true bias detection and classification, and this dataset can serve as an incisive test of new approaches.

Andres Cremisini, Daniela Aguilar, Mark A. Finlayson

Does Causal Coherence Predict Online Spread of Social Media?

Abstract

Online misinformation is primarily spread by humans deciding to do so. We therefore seek to understand the factors making this content compelling and, ultimately, driving online sharing. Fuzzy-Trace Theory, a leading account of decision making, posits that humans encode stimuli, such as online content, at multiple levels of representation; namely, gist, or bottom-line meaning, and verbatim, or surface-level details. Both of these levels of representation are expected to contribute independently to online information spread, with the effects of gist dominating. Important aspects of gist in the context of online content include the presence of a clear causal structure, and semantic coherence – both of which aid in meaning extraction. In this paper, we test the hypothesis that causal and semantic coherence are associated with online sharing of misinformative social media content using Coh-Metrix – a widely-used set of psycholinguistic measures. Results support Fuzzy-Trace Theory’s predictions regarding the role of causally- and semantically-coherent content in promoting online sharing and motivate better measures of these key constructs.

Pedram Hosseini, Mona Diab, David A. Broniatowski

Detecting Disruption: Identifying Structural Changes in the Verkhovna Rada

Abstract

We identify time periods of disruption in the voting patterns of the Ukrainian parliament for the last three convocations. We compare two methods: ideal point estimation (PolSci) and faction detection (CS). Both methods identify the revolution in Ukraine in 2014. The faction detection method also detects structural changes prior to the revolution (election of the president whose tenure was ended early by the revolution), while the ideal points method performs stronger after 2014, identifying a disruption around voting on constitutional changes to implement Minsk II agreements between separatists and Ukraine. The ideal point method is better at detecting position changes of the members of parliament, while the faction method is better at detecting changes in relationships between different MPs. The results suggest that after 2014, the Ukrainian parliament has become more consolidated, but the distribution of its political positions continues to evolve in response to changes in geo-political conditions.

Thomas Magelinski, Jialin Hou, Tymofiy Mylovanov, Kathleen M. Carley

Lost in Online Stores? Agent-Based Modeling of Cognitive Limitations of Elderly Online Consumers

Abstract

We have developed an agent-based model of e-commerce platform users’ behavior with emphasis on reflecting decision-making characteristics of elderly adults. The model has been used to verify how cognitive deficits of older customers influence the effectiveness of collaborative filtering and content-based recommender systems. The results from our simulation suggest that the effectiveness of recommender systems in improving quality of elderly consumers choices is low for population of agents with strong cognitive deficits.

Justyna Pawlowska, Radoslaw Nielek, Adam Wierzbicki

Identifying Toxicity Within YouTube Video Comment

Abstract

Online Social Networks (OSNs), once regarded as safe havens for sharing information and providing mutual support among groups of people, have become breeding grounds for spreading toxic behaviors, political propaganda, and radicalizing content. Toxic individuals often hide under the auspices of anonymity to create fruitless arguments and divert the attention of other users from the core objectives of a community. In this study, we examined five recurring forms of toxicity among the comments posted on pro- and anti-NATO channels on YouTube. We leveraged the YouTube Data API to collect video and comment data from eight channels. We then utilized Google’s Perspective API to assign toxic scores to each comment. Our analysis suggests that, on average, commenters on the anti-NATO channels are more likely to be more toxic than those on the pro-NATO channels. We further discovered that commenters on pro-NATO channels tend to use a mixture of toxic and innocuous comments. We generated word clouds to get an idea of word use frequency, as well as applied the Latent Dirichlet Allocation topic model to classify the comments into their overall topics. The topics extracted from the pro-NATO channels’ comments were primarily positive, such as “Alliance” and “United”; whereas, the topics extracted from anti-NATO channels’ comments were more geared towards geographical locations, such as “Russia”, and negative components such as “Profanity” and “Fake News”. By identifying and examining the toxic behaviors of commenters on YouTube, our analysis lends aid to the pressing need for understanding this toxicity.

Adewale Obadimu, Esther Mead, Muhammad Nihal Hussain, Nitin Agarwal

Examining Intensive Groups in YouTube Commenter Networks

Abstract

Focal structures are the sets of individuals in social networks that are not influential on their own but are influential collectively. These individuals, when coordinating, can be responsible for massive information diffusion, influence operations, or could coordinate (cyber)-attacks. These communities have high tension than other communities in the social network and can mobilize crowds. In this research, we propose a two-level decomposition optimization method for identifying these intensive groups in the complex social networks by constructing a two-level optimization problem for maximizing the local individual’s degree centrality values and the global modularity measures. We also demonstrate the assembled centrality modularity method by applying to a network of YouTube users commenting on conspiracy theory videos to identify coordinating commenters. The dataset consisted of 9,661 users commenting on 4,145 conspiracy theory videos and the derived commenter network contained more than 4.4 million edges. Focal structure analysis was applied to this network to identify sets of users that are coordinating to promote disinformation dissemination. Our proposed model identifies smallest atomic units having high influence, interactions, higher reachability for information propagation. A multi-criteria optimization problem is also employed to rank the identified sets for further investigations.

Mustafa Alassad, Nitin Agarwal, Muhammad Nihal Hussain

User Behavior Modelling for Fake Information Mitigation on Social Web

Abstract

The propagation of fake information on social networks is now a societal problem. Design of mitigation and intervention strategies for fake information has received less attention in social media research, mainly due to the challenge of designing relevant user behavior models. In this paper we lay the groundwork towards such models and present a novel, data-driven approach for user behavior analysis and characterization. We leverage unsupervised learning to define user behavioral categories over key behavior dimensions. We then relate these categories to content-based, user-based, and network-based features that can be extracted in near-real time and identify the most discriminative features. Finally, we build predictive models via supervised learning that leverage these features to determine a user’s behavior category. Rigorous evaluation indicates that the constructed models can be valuable in predicting user behavior from recent activity. These models can be employed to rapidly identify users for intervention in mitigation strategies, crisis communication, and brand management.

Zahra Rajabi, Amarda Shehu, Hemant Purohit

Effect of E-Cigarette Use and Social Network on Smoking Behavior Change: An Agent-Based Model of E-Cigarette and Cigarette Interaction

Abstract

Despite a general reduction in smoking in many areas of the developed world, it remains one of the biggest public health threats. As an alternative to tobacco, the use of electronic cigarettes (ECig) has been increased dramatically over the last decade. ECig use is hypothesized to impact smoking behavior through several pathways, not only as a means of quitting cigarettes and lowering risk of relapse, but also as both an alternative nicotine delivery device to cigarettes, as a visible use of nicotine that can lead to imitative behavior in the form of smoking, and as a gateway nicotine delivery technology that can build high levels of nicotine tolerance and pave the way for initiation of smoking. Evidence regarding the effect of ECig use on smoking behavior change remains inconclusive. To address these challenges, we built an agent-based model (ABM) of smoking and ECig use to examine the effects of ECig use on smoking behavior change. The impact of social network (SN) on the initiation of smoking and ECig use were also explored. Findings from the simulation suggest that the use of ECig generates substantially lower prevalence of current smoker (PCS), which demonstrates the potential for reducing smoking and lowering the risk of relapse. The effects of proximity-based influences within SN increases the prevalence of current ECig user (PCEU). The model also suggests the importance of improved understanding of drivers in cessation and relapse in ECig use, in light of findings that such aspects of behavior change may notably influence smoking behavior change and burden.

Yang Qin, Rojiemiahd Edjoc, Nathaniel D. Osgood

Multi-scale Simulation Modeling for Prevention and Public Health Management of Diabetes in Pregnancy and Sequelae

Abstract

Diabetes in pregnancy (DIP) is an increasing public health priority in the Australian Capital Territory, particularly due to its impact on risk for developing Type 2 diabetes. While earlier diagnostic screening results in greater capacity for early detection and treatment, such benefits must be balanced with the greater demands this imposes on public health services. To address such planning challenges, a multi-scale hybrid simulation model of DIP was built to explore the interaction of risk factors and capture the dynamics underlying the development of DIP. The impact of interventions on health outcomes at the physiological, health service and population level is measured. Of particular central significance in the model is a compartmental model representing the underlying physiological regulation of glycemic status based on beta-cell dynamics and insulin resistance. The model also simulated the dynamics of continuous BMI evolution, glycemic status change during pregnancy and diabetes classification driven by the individual-level physiological model. We further modeled public health service pathways providing diagnosis and care for DIP to explore the optimization of resource use during service delivery. The model was extensively calibrated against empirical data.

Yang Qin, Louise Freebairn, Jo-An Atkinson, Weicheng Qian, Anahita Safarishahrbijari, Nathaniel D. Osgood

Cough Detection Using Hidden Markov Models

Abstract

Respiratory infections and chronic respiratory diseases impose a heavy health burden worldwide. Coughing is one of the most common symptoms of many such infections, and can be indicative of flare-ups of chronic respiratory diseases. Whether at a clinical or public health level, the capacity to identify bouts of coughing can aid understanding of population and individual health status. Developing health monitoring models in the context of respiratory diseases and also seasonal diseases with symptoms such as cough has the potential to improve quality of life, help clinicians and public health authorities with their decisions and decrease the cost of health services. In this paper, we investigated the ability to which a simple machine learning approach in the form of Hidden Markov Models (HMMs) could be used to classify different states of coughing using univariate (with a single energy band as the input feature) and multivariate (with a multiple energy band as the input features) binned time series using both of cough data. We further used the model to distinguish cough events from other events and environmental noise. Our Hidden Markov algorithm achieved 92% AUR (Area Under Receiver Operating Characteristic Curve) in classifying coughing events in noisy environments. Moreover, comparison of univariate with multivariate HMMs suggest a high accuracy of multivariate HMMs for cough event classifications.

Aydin Teyhouee, Nathaniel D. Osgood

Modeling Belief Divergence and Opinion Polarization with Bayesian Networks and Agent-Based Simulation

A Study on Traditional Healing Use in South Africa

Abstract

This study uses agent-based simulation with human settlement patterns to model belief revision and information exchange about health care options. We adopt two recent microeconomic theories based on Bayesian Network formulations for individual belief update then examine the macro-level effects of the belief revision process. This model tries to explain traditional healing usage at the village and regional level while providing a causal mechanism with a single conceptual factor, mobility, at the individual level. The resulting simulation estimates the dependency on traditional healing in villages in Limpopo, South Africa, and the estimates are validated with empirical data.

Kamwoo Lee, Jeanine Braithwaite

Backmatter