Recommended citation: Block Jr., Ray, Michael Burnham, Kayla Kahn, Rachel Peng, Jeremy Seeman, and Christopher Seto. 2022. Perceived risk, political polarization, and the willingness to follow COVID-19 mitigation guidelines. Social Science & Medicine. 305.
View paper
Recommended citation: Palmer, Glenn, Roseanne W McManus, Vito D’Orazio, Michael R Kenwick, Mikaela Karstens, Chase Bloch, Nick Dietrich, Kayla Kahn, Kellan Ritter, and Michael J Soules. 2022. Conflict Management and Peace Science. 39(4).
View paper
Details
Under review
Abstract: Social scientists quickly adopted large language models due to their ability to annotate documents without supervised training, an ability known as zero-shot learning. However, due to their compute demands, cost, and often proprietary nature, these models are often at odds with replication and open science standards. This paper introduces the Political DEBATE (DeBERTa Algorithm for Textual Entailment) language models for zero-shot and few-shot classification of political documents. These models are not only as good, or better than, state-of-the art large language models at zero and few-shot classification, but are orders of magnitude more efficient and completely open source. By training the models on a simple random sample of 10-25 documents, they can outperform supervised classifiers trained on hundreds or thousands of documents and state-of-the-art generative models with complex, engineered prompts. Additionally, we release the PolNLI dataset used to train these models -- a corpus of over 200,000 political documents with highly accurate labels across over 800 classification tasks.
Models available on Hugging Face
Recommended citation: Burnham, Michael, Kayla Kahn, Ryan Yang Wang, and Rachel X. Peng. 2024. "Political DEBATE: Efficient Zero-shot and Few-shot Classifiers for Political Text." https://arxiv.org/abs/2409.02078.
View paper
Details
Abstract: There is a consensus within existing literature in the terrorism field that cooperation between terrorist groups increases their survival. Such a consensus is lacking where lethality is concerned, in no small part due to lethality rarely being studied as a primary explanatory variable for survival. Furthermore, existing literature does not use statistical network methods to examine survival as a dependent variable. This article uses network analysis to examine the effect that both lethality and alliances have on terrorist group survival. I find that the consensus regarding cooperation holds; even when taking network dependencies into account, cooperation leads to longer survival. I also find support for lethality having a curvilinear effect on survival.
Other details: Uses RSiena and accelerated failure time models
Details
Abstract: The terrorism field has long been embroiled in a debate over whether terrorism is an effective coercive tactic. Some argue that terrorism is effective because groups choose the best method available to achieve their goals, while others argue that terrorists do not achieve their long-term goals. Missing from this debate is the distinction between key organizational attributes that may benefit or hinder success. I explore organizational lethality as one such attribute and examine the impact that it has on group success. I show that there is a nonlinear relationship between lethality, success, and failure. Moderate levels of lethality are where groups are most likely to see success, but this is also when they are most likely to be forcibly eliminated.
Other details: Uses cause-specific competing risks Cox proportional hazards models
Details
Abstract: Existing literature in the terrorism field emphasizes the connection between terrorist group alliances, rivalries, and lethality. Building off of the extant literature, this study uses original data on alliances and rivalries in order to assess lethality while accounting for dependence between terrorist groups. I find little evidence that the count of alliances drives lethality. Instead, it is embeddedness of a group within the broader alliance network that leads to increased lethality. I also find support for the outbidding hypothesis.
Other details: Uses temporal network autocorrelation models
You can also find my articles on my Google Scholar profile.