publications

2026

1. R pkg

   factorverse: Tools for Bayesian Factor Analysis with Rcpp

   Peter Dunson

   2026

   Abs Bib Code

   An R package for fitting a variety of linear Gaussian Bayesian factor models under different shrinkage and sparsity priors on the loading matrix, using Gibbs samplers implemented in C++ via RcppArmadillo. All methods share a single interface, so switching priors is a one-word change. The package also provides simulation and diagnostic functions for evaluating fits.

   @software{dunson2026factorverse,
     title = {factorverse: Tools for Bayesian Factor Analysis with Rcpp},
     author = {Dunson, Peter},
     year = {2026},
     version = {0.1.0},
     url = {https://github.com/peterdunson/factorverse},
   }

2. arXiv

   Tutorial for Bayesian Factor Models

   Peter Dunson and Ciprian M. Crainiceanu

   arXiv preprint arXiv:2607.11819, 2026

   Abs arXiv Bib HTML Code

   Bayesian Factor Models (BFM) are well-established models that decompose the observed variability in a set of mean-zero, independent, and uncorrelated factors (random effects). While Factor Analysis (FA) was introduced in 1904 by Spearman, there has been renewed interest in inferential and computational methods that can adapt to large and complex modern data sets that are now routinely collected in a variety of applications. We provide reproducible, harmonized, and fast software for a variety of recent BFMs that allows the direct comparison of methods and provides a one-stop tutorial for the BFMs and their implementation. We neither endorse nor recommend any of the methods for a particular application; we simply provide a previously unavailable harmonized and reproducible common platform for BFMs.

   @article{dunson2026tutorial,
     title = {Tutorial for {B}ayesian Factor Models},
     author = {Dunson, Peter and Crainiceanu, Ciprian M.},
     year = {2026},
     journal = {arXiv preprint arXiv:2607.11819},
   }

3. CEI: A Benchmark for Evaluating Pragmatic Reasoning in Language Models

   Jon Chun, Hannah Sussman, Adrian Mangine, and 13 more authors

   arXiv preprint arXiv:2603.09993, 2026

   Abs arXiv Bib HTML

   Pragmatic reasoning, inferring intended meaning beyond literal semantics, underpins everyday communication yet remains difficult for large language models. We present the Contextual Emotional Inference (CEI) Benchmark: 300 human-validated scenarios for evaluating how well LLMs disambiguate pragmatically complex utterances. Each scenario pairs a situational context and speaker-listener roles (with explicit power relations) against an ambiguous utterance. The dataset covers five pragmatic subtypes (sarcasm/irony, mixed signals, strategic politeness, passive aggression, deflection/misdirection) drawn from workplace, family, social, and service settings, with three power configurations (peer, higher-to-lower, lower-to-higher). Three trained annotators independently labeled every scenario. Inter-annotator agreement (Fleiss’ kappa = 0.06-0.25 by subtype) is low but expected: pragmatic inference admits multiple valid readings, and the disagreement itself is informative. We describe our annotation methodology, including a 4-level quality control pipeline that combines automated statistical checks with expert adjudication. CEI is released under CC-BY-4.0.

   @article{chun2026cei,
     title = {{CEI}: A Benchmark for Evaluating Pragmatic Reasoning in Language Models},
     author = {Chun, Jon and Sussman, Hannah and Mangine, Adrian and Kocaman, Murathan and Sidorko, Kirill and Koirala, Abhigya and McCloud, Andre and Eisenbeis, Gwen and Akanwe, Wisdom and Gassama, Moustapha and {Gonzalez Chirinos}, Eliezer and Enright, Anne-Duncan and Dunson, Peter and Ng, Tiffanie and {von Rosenstiel}, Anna and Idowu, Godwin},
     year = {2026},
     journal = {arXiv preprint arXiv:2603.09993},
   }

2025

1. Disagreement Among AI Models as a Metric of Economic Uncertainty: Testing AI Disagreement Against Traditional Uncertainty Metrics

   Peter Dunson, Andre McCloud, and Gwen Eisenbeis

   2025

   Abs Bib HTML

   This study proposes an AI-based economic uncertainty metric built from disagreement among three frontier large language models (Claude, Google Gemini, and ChatGPT) that score daily market sentiment across five dimensions: equities, inflation, labor, consumer confidence, and forward guidance. We quantify both across-model and within-model disagreement using cosine distance between 5-dimensional sentiment vectors and assess whether these disagreement measures track or relate to established uncertainty benchmarks (VIX, EPU, and the Citi Economic Surprise Index). Results suggested that across-model disagreement moved more closely with news and policy-based uncertainty than with option-implied volatility, as it was strongly positively correlated with the Economic Policy Uncertainty Index and moderately positively correlated with the Citi Economic Surprise Index, but strongly negatively correlated with the VIX. In contrast, within-model disagreement varied substantially by model and date, indicating model-specific prompt sensitivity, rather than a uniform relationship with traditional uncertainty measures.

   @misc{dunson2025disagreement,
     title = {Disagreement Among {AI} Models as a Metric of Economic Uncertainty: Testing {AI} Disagreement Against Traditional Uncertainty Metrics},
     author = {Dunson, Peter and McCloud, Andre and Eisenbeis, Gwen},
     year = {2025},
     howpublished = {Kenyon College Digital Commons},
     url = {https://digital.kenyon.edu/dh_iphs_391/10/},
   }