JWJake Williams

Jake Williams

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Jake Williams has a total of 14 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Understanding how young adults assess the source credibility of misinformation and AI-generated content labels

    by Erica Shusas Racine, Andrea Forte, Jake Williams

    The advent of the internet brought with it unprecedented access to information and potential for many people. However, the ease and speed with which information can be spread has also facilitated the proliferation of false and misleading information that has had a widespread and injurious impact on many aspects of people's lives, including their health, access to fair elections, and personal safety. Social media platforms have been implicated as facilitators of the spread of misinformation, and the American public has been divided in the role that social media platforms should play in the moderation of problematic content. This problem is particularly concerning for young adults, who receive a proportionally high amount of their news from social media platforms. Content labels are the preeminent means to communicate to social media users whether a social media post contains misinformation, and more recently, AI-generated content. However, research has shown that content labels are less effective on young adults. Prior work has also shown that young adults value the source of the information identification of a post (usually fact checkers, AI tools, or other social media users) in their decision to value the content label's assessment. This dissertation examines how young adults (ages 18-25) assess the source credibility of content labels for misinformation and AI-generated content. I first present findings from a survey of 1,462 adults that was implemented to establish a baseline understanding of which content label sources are more effective for young adults for different types of content (Chapter 4) as well as results from analyses of that survey data that tested for differences among different demographic groups within the young adult population (Chapter 5). I then share findings from a diary study with 22 participants and follow-up group interviews with 9 of those participants on why they found some sources more credible than others for text information and AI-generated images (Chapter 6). I found that for young adults, AI Tools are most effective for labels on false text and AI-generated images, professional fact checkers are most effective for true text labels, and labels from other social media users are most effective for labels that identify an image is not AI-generated. My findings also include key distinctions between demographic groups in their source credibility assessment of content labels as well as identify demographic groups for which the labels were least impactful. Additionally, I contribute a framework of strategies commonly used by young adults to assess the source credibility of content labels for text misinformation and AI-generated images. Finally, I discuss the implications of these findings as well as directions for future research (Chapter 7).


  • Modeling Emerging News Stories Across Digital Publications and Social Media

    by Munif Ishad Mujib, Jake Williams

    The complex process of news production, through intense competition for audience attention among outlets, generates content that presents significant interaction challenges to the end user. In addition to the information overload introduced by the transposition of the 24-hour news cycle onto the internet, interaction and intersection of digital news production networks and social network platforms have introduced additionally complex phenomena in the form of novel digital artifacts. One such phenomenon is the interactivity-enabled inclusion, or "embedding'' of user-generated content from social media within digital news articles. The preponderance of such complexity within news stories place great cognitive burdens on users, i.e. news consumers, attempting to process this multifaceted information stream. News aggregators offer an avenue for users to organize and direct their consumption, however, major news aggregators in use today remain rudimentary by simply replicating hyperlinked representations of conventional newspaper-style story organization. This landscape of news retrieval and consumption technologies presents the opportunity to conceptualize and design more intelligent systems and tools that can help reduce the challenging complexity of news consumption through distributed cognition. In this dissertation, I present a multi-year ongoing data collection effort, NewsTweet, that acquires digital news articles, embedded social media content from Twitter, and further potentially newsworthy tweets. I analyze the dataset acquired by NewsTweet to investigate the mechanisms and associated journalistic practices of tweet embedding through the perspective of temporality as well as outlet and embedded user dynamics. Next, I describe a network construction approach from embedded tweets and connected articles, investigate network structures, identify the topical significance of connected subgraphs, and evaluate methods for segmentation of larger connected subgraphs into further topically distinct groups of articles. I introduce, implement, and evaluate state-of-the-art semantic featurization and unsupervised clustering methods for generalized news aggregation. Further, I describe and implement an algorithmic approach for combining embedded subgraph segmentation with semantic clustering so that network structure from granular subgraphs may be preserved within a larger aggregation pipeline. Through a series of experiments, I demonstrate the potential for modular and scalable design of a novel model of news aggregation that constructs hierarchies of stories in the news, which may be extended and deployed within user-facing applications that can help news consumers navigate contexts and hierarchies associated with news stories interactively, engendering a novel class of news aggregator design. The aggregation model presented in this work incorporates state-of-the-art semantic technologies to implement a topological organization of related news stories, allowing for guided exploration of interconnected events and associated temporal dynamics. A novel class of news aggregator design enabled by the technology developed in this work can help improve user experience of news consumption and create a positive impact in terms of digital social health by rethinking news discovery mechanics and optimizing for context navigation and exploration rather than engagement.

  • Open Access

    Motivational Interviewing Conversational Agent for Parents as Proxies for Their Children in Healthy Eating: Development and User Testing

    by Diva Smriti, Tsui-Sui Annie Kao, Rahil Rathod, Ji Youn Shin, Wei Peng, Jake Williams, Munif Ishad Mujib, Meghan Colosimo, Jina Huh-Yoo

    Background: Increased adoption of off-the-shelf conversational agents (CAs) brings opportunities to integrate therapeutic interventions. Motivational Interviewing (MI) can then be integrated with CAs for cost-effective access to it. MI can be especially beneficial for parents who often have low motivation because of limited time and resources to eat healthy together with their children. Objective: We developed a Motivational Interviewing Conversational Agent (MICA) to improve healthy eating in parents who serve as a proxy for health behavior change in their children. Proxy relationships involve a person serving as a catalyst for behavior change in another person. Parents, serving as proxies, can bring about behavior change in their children. Methods: We conducted user test sessions of the MICA prototype to understand the perceived acceptability and usefulness of the MICA prototype by parents. A total of 24 parents of young children participated in 2 user test sessions with MICA, approximately 2 weeks apart. After parents’ interaction with the MICA prototype in each user test session, we used qualitative interviews to understand parents’ perceptions and suggestions for improvements in MICA. Results: Findings showed participants’ perceived usefulness of MICAs for helping them self-reflect and motivating them to adopt healthier eating habits together with their children. Participants further suggested various ways in which MICA can help them safely manage their children’s eating behaviors and provide customized support for their proxy needs and goals. Conclusions: We have discussed how the user experience of CAs can be improved to uniquely offer support to parents who serve as proxies in changing the behavior of their children. We have concluded with implications for a larger context of designing MI-based CAs for supporting proxy relationships for health behavior change.


  • A framework for measuring and mitigating bias in electronic health record data

    by Elizabeth Campbell, Jake Williams, Aaron J. Masino

    Electronic Health Records (EHRs) are valuable sources of healthcare data with tremendous potential to improve patient care and reduce financial costs. Advanced data mining and analytical methods are necessary to obtain, transform, and analyze EHR data for secondary uses such as clinical and health policy research, and machine learning methods are key to addressing challenges in secondary EHR uses. However, developing machine learning models equitably is a significant challenge. As machine learning tools become more ubiquitous in healthcare research and clinical practice, it is critical that predictive models are developed that work to alleviate, rather than exacerbate, existing disparities in the healthcare system. Bias may be introduced into a machine learning model through the data used to train the model, algorithm selection, or in how a model is implemented in an applied setting. In this work, we present a framework for assessing and alleviating bias in dataset development for a classification model derived from EHR data. Classification, which refers to inputting observations from data into categories, is a common machine learning task and a method that is increasingly used in medical research. Classification may be used to predict future health outcomes, identify patients who are at risk for developing a condition, and to make clinical diagnoses. In this work, we present a case study in childhood obesity incidence prediction where a classification model is developed that aims to predict future obesity among pediatric patients based on demographic data and the presence of clinical diagnostic patterns in their medical history. We use easily accessible EHR data that may be obtained to predict other outcomes of interest, and present a method to evaluate data to develop classifiers that may be used for prediction of other chronic diseases (such as heart disease, diabetes, and asthma). First, temporal pattern mining is applied to EHR data to identify common diagnostic patterns that surround pediatric obesity incidence. Next, a dataset is developed for the study cohort, that indicates whether or not patients had the temporal condition patterns that were previously identified and demographic information (such as race/ethnicity, gender, etc.) obtained from their EHR. Then, classification models are trained and tested using this dataset. We develop a framework for assessing bias in EHR-derived datasets, and present its application in our pediatric obesity incidence case study. We conclude by discussing implications of our work in healthcare research and outlining directions for future research, including assessing and alleviating bias in algorithm selection and model development.

  • Low-power, phase-preserving 2R amplitude regenerator

    by Taras I. Lakoba, Jake R. Williams, Michael Vasilyev