Libby Hemphill

Wenyue Hua, Lizhou Fan, Lingyao Li et al.

Can we avoid wars at the crossroads of history? This question has been pursued by individuals, scholars, policymakers, and organizations throughout human history. In this research, we attempt to answer the question based on the recent advances of Artificial Intelligence (AI) and Large Language Models (LLMs). We propose \textbf{WarAgent}, an LLM-powered multi-agent AI system, to simulate the participating countries, their decisions, and the consequences, in historical international conflicts, including the World War I (WWI), the World War II (WWII), and the Warring States Period (WSP) in Ancient China. By evaluating the simulation effectiveness, we examine the advancements and limitations of cutting-edge AI systems' abilities in studying complex collective human behaviors such as international conflicts under diverse settings. In these simulations, the emergent interactions among agents also offer a novel perspective for examining the triggers and conditions that lead to war. Our findings offer data-driven and AI-augmented insights that can redefine how we approach conflict resolution and peacekeeping strategies. The implications stretch beyond historical analysis, offering a blueprint for using AI to understand human history and possibly prevent future international conflicts. Code and data are available at \url{https://github.com/agiresearch/WarAgent}.

Lizhou Fan, Lingyao Li, Zihui Ma et al.

Large language models (LLMs) are a class of language models that have demonstrated outstanding performance across a range of natural language processing (NLP) tasks and have become a highly sought-after research area, because of their ability to generate human-like language and their potential to revolutionize science and technology. In this study, we conduct bibliometric and discourse analyses of scholarly literature on LLMs. Synthesizing over 5,000 publications, this paper serves as a roadmap for researchers, practitioners, and policymakers to navigate the current landscape of LLMs research. We present the research trends from 2017 to early 2023, identifying patterns in research paradigms and collaborations. We start with analyzing the core algorithm developments and NLP tasks that are fundamental in LLMs research. We then investigate the applications of LLMs in various fields and domains including medicine, engineering, social science, and humanities. Our review also reveals the dynamic, fast-paced evolution of LLMs research. Overall, this paper offers valuable insights into the current state, impact, and potential of LLMs research and its applications.

Lingyao Li, Lizhou Fan, Shubham Atreja et al.

Harmful content is pervasive on social media, poisoning online communities and negatively impacting participation. A common approach to address this issue is to develop detection models that rely on human annotations. However, the tasks required to build such models expose annotators to harmful and offensive content and may require significant time and cost to complete. Generative AI models have the potential to understand and detect harmful content. To investigate this potential, we used ChatGPT and compared its performance with MTurker annotations for three frequently discussed concepts related to harmful content: Hateful, Offensive, and Toxic (HOT). We designed five prompts to interact with ChatGPT and conducted four experiments eliciting HOT classifications. Our results show that ChatGPT can achieve an accuracy of approximately 80% when compared to MTurker annotations. Specifically, the model displays a more consistent classification for non-HOT comments than HOT comments compared to human annotations. Our findings also suggest that ChatGPT classifications align with provided HOT definitions, but ChatGPT classifies "hateful" and "offensive" as subsets of "toxic." Moreover, the choice of prompts used to interact with ChatGPT impacts its performance. Based on these in-sights, our study provides several meaningful implications for employing ChatGPT to detect HOT content, particularly regarding the reliability and consistency of its performance, its understand-ing and reasoning of the HOT concept, and the impact of prompts on its performance. Overall, our study provides guidance about the potential of using generative AI models to moderate large volumes of user-generated content on social media.

Lizhou Fan, Sara Lafia, David Bleckley et al.

Data citations provide a foundation for studying research data impact. Collecting and managing data citations is a new frontier in archival science and scholarly communication. However, the discovery and curation of research data citations is labor intensive. Data citations that reference unique identifiers (i.e. DOIs) are readily findable; however, informal mentions made to research data are more challenging to infer. We propose a natural language processing (NLP) paradigm to support the human task of identifying informal mentions made to research datasets. The work of discovering informal data mentions is currently performed by librarians and their staff in the Inter-university Consortium for Political and Social Research (ICPSR), a large social science data archive that maintains a large bibliography of data-related literature. The NLP model is bootstrapped from data citations actively collected by librarians at ICPSR. The model combines pattern matching with multiple iterations of human annotations to learn additional rules for detecting informal data mentions. These examples are then used to train an NLP pipeline. The librarian-in-the-loop paradigm is centered in the data work performed by ICPSR librarians, supporting broader efforts to build a more comprehensive bibliography of data-related literature that reflects the scholarly communities of research data users.

Angela Schöpke-Gonzalez, Siqi Wu, Sagar Kumar et al.

Computational social science research has made advances in machine learning and natural language processing that support content moderators in detecting harmful content. These advances often rely on training datasets annotated by crowdworkers for harmful content. In designing instructions for annotation tasks to generate training data for these algorithms, researchers often treat the harm concepts that we train algorithms to detect - 'hateful', 'offensive', 'toxic', 'racist', 'sexist', etc. - as interchangeable. In this work, we studied whether the way that researchers define 'harm' affects annotation outcomes. Using Venn diagrams, information gain comparisons, and content analyses, we reveal that annotators do not use the concepts 'hateful', 'offensive', and 'toxic' interchangeably. We identify that features of harm definitions and annotators' individual characteristics explain much of how annotators use these terms differently. Our results offer empirical evidence discouraging the common practice of using harm concepts interchangeably in content moderation research. Instead, researchers should make specific choices about which harm concepts to analyze based on their research goals. Recognizing that researchers are often resource constrained, we also encourage researchers to provide information to bound their findings when their concepts of interest differ from concepts that off-the-shelf harmful content detection algorithms identify. Finally, we encourage algorithm providers to ensure their instruments can adapt to contextually-specific content detection goals (e.g., soliciting instrument users' feedback).

Sara Lafia, Lizhou Fan, Libby Hemphill

Discovering authoritative links between publications and the datasets that they use can be a labor-intensive process. We introduce a natural language processing pipeline that retrieves and reviews publications for informal references to research datasets, which complements the work of data librarians. We first describe the components of the pipeline and then apply it to expand an authoritative bibliography linking thousands of social science studies to the data-related publications in which they are used. The pipeline increases recall for literature to review for inclusion in data-related collections of publications and makes it possible to detect informal data references at scale. We contribute (1) a novel Named Entity Recognition (NER) model that reliably detects informal data references and (2) a dataset connecting items from social science literature with datasets they reference. Together, these contributions enable future work on data reference, data citation networks, and data reuse.

Lizhou Fan, Lingyao Li, Libby Hemphill

Background: Online toxicity, encompassing behaviors such as harassment, bullying, hate speech, and the dissemination of misinformation, has become a pressing social concern in the digital age. The 2022 Mpox outbreak, initially termed "Monkeypox" but subsequently renamed to mitigate associated stigmas and societal concerns, serves as a poignant backdrop to this issue. Objective: In this research, we undertake a comprehensive analysis of the toxic online discourse surrounding the 2022 Mpox outbreak. Our objective is to dissect its origins, characterize its nature and content, trace its dissemination patterns, and assess its broader societal implications, with the goal of providing insights that can inform strategies to mitigate such toxicity in future crises. Methods: We collected more than 1.6 million unique tweets and analyzed them from five dimensions, including context, extent, content, speaker, and intent. Utilizing BERT-based topic modeling and social network community clustering, we delineated the toxic dynamics on Twitter. Results: We identified five high-level topic categories in the toxic online discourse on Twitter, including disease (46.6%), health policy and healthcare (19.3%), homophobia (23.9%), politics (6.0%), and racism (4.1%). Through the toxicity diffusion networks of mentions, retweets, and the top users, we found that retweets of toxic content were widespread, while influential users rarely engaged with or countered this toxicity through retweets. Conclusions: By tracking topical dynamics, we can track the changing popularity of toxic content online, providing a better understanding of societal challenges. Network dynamics spotlight key social media influencers and their intents, indicating that addressing these central figures in toxic discourse can enhance crisis communication and inform policy-making.

Zihui Ma, Lingyao Li, Libby Hemphill et al.

Effective disaster response is critical for affected communities. Responders and decision-makers would benefit from reliable, timely measures of the issues impacting their communities during a disaster, and social media offers a potentially rich data source. Social media can reflect public concerns and demands during a disaster, offering valuable insights for decision-makers to understand evolving situations and optimize resource allocation. We used Bidirectional Encoder Representations from Transformers (BERT) topic modeling to cluster topics from Twitter data. Then, we conducted a temporal-spatial analysis to examine the distribution of these topics across different regions during the 2020 western U.S. wildfire season. Our results show that Twitter users mainly focused on three topics:"health impact," "damage," and "evacuation." We used the Susceptible-Infected-Recovered (SIR) theory to explore the magnitude and velocity of topic diffusion on Twitter. The results displayed a clear relationship between topic trends and wildfire propagation patterns. The estimated parameters obtained from the SIR model in selected cities revealed that residents exhibited a high level of several concerns during the wildfire. Our study details how the SIR model and topic modeling using social media data can provide decision-makers with a quantitative approach to measure disaster response and support their decision-making processes.

Megan A. Brown, Shubham Atreja, Libby Hemphill et al.

Researchers have proposed the use of generative large language models (LLMs) to label data for research and applied settings. This literature emphasizes the improved performance of these models relative to other natural language models, noting that generative LLMs typically outperform other models and even humans across several metrics. Previous literature has examined bias across many applications and contexts, but less work has focused specifically on bias in generative LLMs' responses to subjective annotation tasks. This bias could result in labels applied by LLMs that disproportionately align with majority groups over a more diverse set of viewpoints. In this paper, we evaluate how LLMs represent diverse viewpoints on these contentious tasks. Across four annotation tasks on four datasets, we show that LLMs do not show systematic substantial disagreement with annotators on the basis of demographics. Rather, we find that multiple LLMs tend to be biased in the same directions on the same demographic categories within the same datasets. Moreover, the disagreement between human annotators on the labeling task -- a measure of item difficulty -- is far more predictive of LLM agreement with human annotators. We conclude with a discussion of the implications for researchers and practitioners using LLMs for automated data annotation tasks. Specifically, we emphasize that fairness evaluations must be contextual, model choice alone will not solve potential issues of bias, and item difficulty must be integrated into bias assessments.

Shubham Atreja, Joshua Ashkinaze, Lingyao Li et al.

Manually annotating data for computational social science tasks can be costly, time-consuming, and emotionally draining. While recent work suggests that LLMs can perform such annotation tasks in zero-shot settings, little is known about how prompt design impacts LLMs' compliance and accuracy. We conduct a large-scale multi-prompt experiment to test how model selection (ChatGPT, PaLM2, and Falcon7b) and prompt design features (definition inclusion, output type, explanation, and prompt length) impact the compliance and accuracy of LLM-generated annotations on four CSS tasks (toxicity, sentiment, rumor stance, and news frames). Our results show that LLM compliance and accuracy are highly prompt-dependent. For instance, prompting for numerical scores instead of labels reduces all LLMs' compliance and accuracy. The overall best prompting setup is task-dependent, and minor prompt changes can cause large changes in the distribution of generated labels. By showing that prompt design significantly impacts the quality and distribution of LLM-generated annotations, this work serves as both a warning and practical guide for researchers and practitioners.

Andrea K. Thomer, Dharma Akmon, Jeremy York et al.

Data curation is the process of making a dataset fit-for-use and archiveable. It is critical to data-intensive science because it makes complex data pipelines possible, makes studies reproducible, and makes data (re)usable. Yet the complexities of the hands-on, technical and intellectual work of data curation is frequently overlooked or downplayed. Obscuring the work of data curation not only renders the labor and contributions of the data curators invisible; it also makes it harder to tease out the impact curators' work has on the later usability, reliability, and reproducibility of data. To better understand the specific work of data curation -- and thereby, explore ways of showing curators' impact -- we conducted a close examination of data curation at a large social science data repository, the Inter-university Consortium of Political and Social Research (ICPSR). We asked, What does curatorial work entail at ICPSR, and what work is more or less visible to different stakeholders and in different contexts? And, how is that curatorial work coordinated across the organization? We triangulate accounts of data curation from interviews and records of curation in Jira tickets to develop a rich and detailed account of curatorial work. We find that curators describe a number of craft practices needed to perform their work, which defies the rote sequence of events implied by many lifecycle or workflow models. Further, we show how best practices and craft practices are deeply intertwined.

Shubham Atreja, Libby Hemphill, Paul Resnick

To reduce the spread of misinformation, social media platforms may take enforcement actions against offending content, such as adding informational warning labels, reducing distribution, or removing content entirely. However, both their actions and their inactions have been controversial and plagued by allegations of partisan bias. When it comes to specific content items, surprisingly little is known about what ordinary people want the platforms to do. We provide empirical evidence about a politically balanced panel of lay raters' preferences for three potential platform actions on 368 news articles. Our results confirm that on many articles there is a lack of consensus on which actions to take. We find a clear hierarchy of perceived severity of actions with a majority of raters wanting informational labels on the most articles and removal on the fewest. There was no partisan difference in terms of how many articles deserve platform actions but conservatives did prefer somewhat more action on content from liberal sources, and vice versa. We also find that judgments about two holistic properties, misleadingness and harm, could serve as an effective proxy to determine what actions would be approved by a majority of raters.

Sara Lafia, Andrea Thomer, David Bleckley et al.

This paper describes a machine learning approach for annotating and analyzing data curation work logs at ICPSR, a large social sciences data archive. The systems we studied track curation work and coordinate team decision-making at ICPSR. Repository staff use these systems to organize, prioritize, and document curation work done on datasets, making them promising resources for studying curation work and its impact on data reuse, especially in combination with data usage analytics. A key challenge, however, is classifying similar activities so that they can be measured and associated with impact metrics. This paper contributes: 1) a schema of data curation activities; 2) a computational model for identifying curation actions in work log descriptions; and 3) an analysis of frequent data curation activities at ICPSR over time. We first propose a schema of data curation actions to help us analyze the impact of curation work. We then use this schema to annotate a set of data curation logs, which contain records of data transformations and project management decisions completed by repository staff. Finally, we train a text classifier to detect the frequency of curation actions in a large set of work logs. Our approach supports the analysis of curation work documented in work log systems as an important step toward studying the relationship between research data curation and data reuse.

Libby Hemphill, Angela M. Schöpke-Gonzalez

Understanding how political attention is divided and over what subjects is crucial for research on areas such as agenda setting, framing, and political rhetoric. Existing methods for measuring attention, such as manual labeling according to established codebooks, are expensive and can be restrictive. We describe two computational models that automatically distinguish topics in politicians' social media content. Our models---one supervised classifier and one unsupervised topic model---provide different benefits. The supervised classifier reduces the labor required to classify content according to pre-determined topic list. However, tweets do more than communicate policy positions. Our unsupervised model uncovers both political topics and other Twitter uses (e.g., constituent service). These models are effective, inexpensive computational tools for political communication and social media research. We demonstrate their utility and discuss the different analyses they afford by applying both models to the tweets posted by members of the 115th U.S. Congress.

David Jurgens, Eshwar Chandrasekharan, Libby Hemphill

Online abusive behavior affects millions and the NLP community has attempted to mitigate this problem by developing technologies to detect abuse. However, current methods have largely focused on a narrow definition of abuse to detriment of victims who seek both validation and solutions. In this position paper, we argue that the community needs to make three substantive changes: (1) expanding our scope of problems to tackle both more subtle and more serious forms of abuse, (2) developing proactive technologies that counter or inhibit abuse before it harms, and (3) reframing our effort within a framework of justice to promote healthy communities.

Libby Hemphill, A. J. Million, Ingrid Erickson

We present findings from interviews with 23 individuals affiliated with non-profit organizations (NPOs) to understand how they deploy information and communication technologies (ICTs) in civic engagement efforts. Existing research about NPO ICT use is largely critical, but we did not find evidence that NPOs fail to use tools effectively. Rather, we detail how various ICT use on the part of NPOs intersects with unique affordance perceptions and adoption causes. Overall, we find that existing theories about technology choice (e.g., task-technology fit, uses and gratifications) do not explain the assemblages NPOs describe. We argue that NPOs fashion infrastructures in accordance with their moral economy frameworks rather than selecting tools based on utility. Together, the rhetorics of infrastructure and moral economies capture the motivations and constraints our participants expressed and challenge how prevailing theories of ICT usage describe the non-profit landscape.

Ping Liu, Joshua Guberman, Libby Hemphill et al.

Online antisocial behavior, such as cyberbullying, harassment, and trolling, is a widespread problem that threatens free discussion and has negative physical and mental health consequences for victims and communities. While prior work has proposed automated methods to identify hostile comments in online discussions, these methods work retrospectively on comments that have already been posted, making it difficult to intervene before an interaction escalates. In this paper we instead consider the problem of forecasting future hostilities in online discussions, which we decompose into two tasks: (1) given an initial sequence of non-hostile comments in a discussion, predict whether some future comment will contain hostility; and (2) given the first hostile comment in a discussion, predict whether this will lead to an escalation of hostility in subsequent comments. Thus, we aim to forecast both the presence and intensity of hostile comments based on linguistic and social features from earlier comments. To evaluate our approach, we introduce a corpus of over 30K annotated Instagram comments from over 1,100 posts. Our approach is able to predict the appearance of a hostile comment on an Instagram post ten or more hours in the future with an AUC of .82 (task 1), and can furthermore distinguish between high and low levels of future hostility with an AUC of .91 (task 2).

Libby Hemphill

To address "bad actors" online, I argue for more specific definitions of acceptable and unacceptable behaviors and explicit attention to the social structures in which behaviors occur.