- Agent-based modeling of new product market diffusion: an overview of strengths and criticisms , ANNALS OF OPERATIONS RESEARCH (2021)
- Apart we ride together: The motivations behind users of mixed-reality sports , JOURNAL OF BUSINESS RESEARCH (2021)
- Inferring mechanisms of response prioritization on social media under information overload , SCIENTIFIC REPORTS (2021)
- Influence Cascades: Entropy-Based Characterization of Behavioral Influence Patterns in Social Media , ENTROPY (2021)
- Real-Time Brand Reputation Tracking Using Social Media , JOURNAL OF MARKETING (2021)
- Is the Grass Greener? On the Strategic Implications of Moving Along the Value Chain for IT Service Providers , INFORMATION SYSTEMS RESEARCH (2020)
- The effects of information overload on online conversation dynamics , COMPUTATIONAL AND MATHEMATICAL ORGANIZATION THEORY (2020)
- Characterising climate change discourse on social media during extreme weather events , Global Environmental Change (2019)
- Inferring models of opinion dynamics from aggregated jury data , PLOS ONE (2019)
- Letting the Computers Take Over: Using AI to Solve Marketing Problems , CALIFORNIA MANAGEMENT REVIEW (2019)
NC State will support Perceptronics on this effort in the areas of: Anomaly Detection using Causal State Models, Image Analysis, Natural Language Processing.
Social Media provides a popular platform for marketers to provide content to their consumers and stakeholders. However, many organizations also fear the ability that social media gives to malignant users to spread disinformation. The fear of the potential effects of malicious influencers (e.g., bots, trolls, extremists) spreading falsehoods or attempting to radicalize the general public on social media has been recognized since as early as 2005. In recent years, these concerns have been validated and pose a significant threat to public opinion and internet word of mouth. In this project, we propose a methodology based on functional data techniques to understand the users' daily behavior on these social media platforms. We will model the user data as a â€œcurveâ€ with 0/1 values or categorical values and understand a userâ€™s behavior on a social media platform by pooling across all users of the same kind. Our methods have the potential to identify the minimum number of days over which to follow users to determine their true nature (eg. genuine user or malicious). In this proposal, we will also investigate functional-data based methods that focus on the interpretability of the results, so that we can formally assess whether there are certain times during the day when the social media activity is different across multiple types of users. The methods will be deployed as packaged in R and will be applied to several social media data.
The objective of this project is to create an informative and timely newsletter for CHAI users that highlights various trending topics from the worldâ€™s most read literature. The initial focus of the newsletter will be on managing the SARS-Cov-2 virus and the disease it causes, Covid-19.
Dr. Rand will utilize his expertise in social media, information diffusion, agent-based modeling, and evolutionary computation to advise the team on several major aspects of the prime contract. In addition, he will help with the supervision of students and / or postdocs involved with the project. In particular, Prof. Rand will facilitate the following four major areas of the project: (1) Helping with the development of the Deep Agent Framework (DAF) Modeling Markup Language (DMML) (2) Assisting in the creation of three plausible submodels for the emotional, social, and cognitive components of DMML (3) Work on the architecture and development of the Modeling for Information Spread and Evolution in Online Environments (MISEOE) (4) Collaborate on the development and articulation of the information Evaluation module within the MISEOE"
Specifically the project will attempt to answer the following questions: 1. Can we use social media data (Twitter) to identify self-suspected cases of COVID-19 in LMIC countries in Sub-Saharan Africa, and then predict future caseload? 2. What is the best way to collect and organize social media data about the concerns of Health Care Workers from LMIC countries in Sub-Saharan Africa related to the COVID-19 pandemic? And what initial information can be found in this data? 3. The last question is more exploratory in nature than the previous two. Can we start to characterize patientsâ€™ journeys through COVID-19 based on their Twitter timeline?
Specifically the project will attempt to answer the following questions: What is the current status of clinical risk factors that correlate to disease severity? Type of CVD or other condition, age, gender, medication use, ethnicity, type of exposure. What are the most commonly used treatments for managing COVID disease and symptoms? What do we know about how much is being used per patient? (This will help triage the WHO list.) What treatments for COVID-induced pneumonia work best?
PAIT - Statement of Work NCSU NC State will support Perceptronics in this effort in the following roles: Anomaly Detection using Causal State Models - A key component of PAIT involves determining if a particular datastream tied to an individual is exhibiting anomalous behavior. We will take the Causal State Modeling approach that has been developed for batch data and construct an online version of it, by combining this with new methods of using CSM to detect anomalies we will build an online anomaly detector for streaming data. Image Analysis - We will work with Perceptronics to develop methods for detecting threats in images. The NCSU team has experience working with Convolutional Neural Networks and processing social media images. Building on this and applying new transfer learning methods, we will work with Pereceptronics to develop new approaches to detect threats in online images. Natural Language Processing - The NCSU team has considerable experience with analyzing social media data. Though NCSU will not be developing new natural language methods, they will play an advisory role in helping with the natural language processing of the social media data used for this contract.
LAS DO1 Rand -3.3 Computational Social Sciences
The North Carolina State University team agrees to help in accomplishing the following goals as part of this proposal: 1. Alternate Features for Causal State Modeling â€“ Explore and evaluate alternative features for causal state modeling such as different time resolutions, transducer models, alternate input encoding, etc., 2. Causal State Modeling for User Classification into Groups â€“ Extend previous work for using CSM to inform classification of users into known groups. 3. Causal State Modeling for Change Detection â€“ Explore using CSM techniques to detect changes / anomalies in trend data. 4. Individual and Content-Based Recommendations â€“ Explore combining individual-level and content-level recommendations together to improve the performance of each other. 5. Predicting Influential Users â€“ Assist and advise in the development of methods to predict which users will be influential in the future, based on dynamics observed in the past. The North Carolina State University team will provide updates to Perceptronics on an as-needed basis, and in accordance with DARPAâ€™s requirements for Perceptronicsâ€™ reporting.
NC State will support Perceptronics in this effort in the following roles: 1. Identifying Suspect Accounts - The NCSU team will expand on its previous work using statistical methods, specifically Causal State Modeling and Functional Data Analysis, to identify suspect accounts on social media. We have developed techniques to separate genuine users from trolls from bots, and we will apply and improve these methods while working with Perceptronics. 2. Analyzing and Characterizing Suspect Accounts - The NCSU team will further explore the identified suspect accounts to examine whether those accounts are acting in a coordinated way. We will also attempt to identify latent characteristics of these users such as country affiliation, group affiliation, and potential role in the social media landscape. 3. Visualization - The NCSU team has considerable experience with visualizing social media data. Though NCSU will not be developing new visualization methods, they will play an advisory role in helping with the visualization of the social media analysis used for this contract.