Research

You can also find my articles on my Google Scholar profile.

Causal Carbon: Baselines and Additionality with Potential Outcomes

Preprint

Recent work has questioned the credibility of forest carbon offsets as a climate solution. This threatens both investor confidence and genuine climate mitigation efforts in the voluntary carbon market, which has contracted by over 75% since 2021. Despite updated methodologies and widespread advice to invest only in high-integrity or high-quality credits, it remains unclear which credits genuinely meet these criteria. Here, we draw on the fields of statistics and causal inference to develop a generalized analytical framework for evaluating the additionality of credits generated by carbon offset protocols, addressing persistent ambiguities and limitations in current approaches. This framework comes from a systematic evaluation of all existing forest carbon offset protocols. We translated each protocol’s baseline methodology into a statistical causal estimator and derived the assumptions necessary for it to produce accurate estimates of generated credits. By translating those assumptions back into the language of forest carbon credits we provide a set of conditions that buyers should believe in order to conclude that the credits they purchase are additional. We demonstrate that strategic enrollment, combined with even minor measurement errors in carbon stocks, can lead to significant market distortions, undermining both environmental integrity and financial reliability of offsets. Our analysis highlights an inherent tradeoff between ensuring accurate carbon measurement and expanding participation in offset programs. This framework clarifies the assumptions purchasers of carbon offsets must accept for credits under each protocol to reliably represent genuine climate impacts, providing a transparent basis for buyers, sellers, critics, and advocates to constructively engage and reconcile divergent views.

Recommended citation: Ayers, M., Sanford, L., Gardner, W., & Kuebbing, S. (2026, May 4). Causal Carbon: Baselines and Additionality with Potential Outcomes. https://doi.org/10.31219/osf.io/5pcuh_v3
Download Paper

Adversarial Debiasing for Unbiased Parameter Recovery

Proceedings of the 29th International Conference on Artificial Intelligence and Statistics, 2026

Advances in machine learning and the increasing availability of high-dimensional data have led to the proliferation of social science research that uses the predictions of machine learning models as proxies for measures of human activity or environmental outcomes. However, prediction errors from machine learning models can lead to bias in the estimates of regression coefficients. In this paper, we show how this bias can arise, propose a test for detecting bias, and demonstrate the use of an adversarial machine learning algorithm in order to de-bias predictions. These methods are applicable to any setting where machine-learned predictions are the dependent variable in a regression. We conduct simulations and empirical exercises using ground truth and satellite data on forest cover in Africa. Using the predictions from a naive machine learning model leads to biased parameter estimates, while the predictions from the adversarial model recover the true coefficients.

Recommended citation: Sanford, Luke, Megan Ayers, Matthew Gordon, and Eliana Stone. “Adversarial Debiasing for Parameter Recovery.” Proceedings of the 29th International Conference on Artificial Intelligence and Statistics, 2026.
Download Paper

Discovering Influential Text Using Convolutional Neural Networks

Findings of the Association for Computational Linguistics ACL, 2024

Experimental methods for estimating the impacts of text on human evaluation have been widely used in the social sciences. However, researchers in experimental settings are usually limited to testing a small number of pre-specified text treatments. While efforts to mine unstructured texts for features that causally affect outcomes have been ongoing in recent years, these models have primarily focused on the topics or specific words of text, which may not always be the mechanism of the effect. We connect these efforts with NLP interpretability techniques and present a method for flexibly discovering clusters of similar text phrases that are predictive of human reactions to texts using convolutional neural networks. When used in an experimental setting, this method can identify text treatments and their effects under certain assumptions. We apply the method to two data sets. The first enables direct validation of the model’s ability to detect phrases known to cause the outcome. The second demonstrates its ability to flexibly discover text treatments with varying textual structures. In both cases, the model learns a greater variety of text treatments compared to benchmark methods, and these text features quantitatively meet or exceed the ability of benchmark methods to predict the outcome.

Recommended citation: Megan Ayers, Luke Sanford, Margaret Roberts, and Eddie Yang. 2024. Discovering influential text using convolutional neural networks. In Findings of the Association for Computational Linguistics ACL 2024, pages 12002–12027, Bangkok, Thailand and virtual meeting. Association for Computational Linguistics.
Download Paper

Changes in Global Warming’s Six Americas: An Analysis of Repeat Respondents

Climatic Change, 2024

Building public consensus about the threat of climate change is critical for enacting meaningful action to address it. To understand how Americans are changing their beliefs about climate change, research typically relies on cross-sectional survey responses. Data that is collected from the same individuals over time– panel data– provides clearer evidence about whether people’s beliefs are shifting. In this article, we investigate changes in climate beliefs among the American public using panel data from 2,135 survey respondents, analyzing opinion changes through the “Global Warming’s Six Americas” framework– an audience segmentation tool that identifies the people who are the most worried about global warming (the Alarmed) to the least worried (the Dismissive). Our findings indicate that many Americans are changing their minds about climate change and becoming more worried over time, and that these shifts correlate with changes in support for climate policy and behavioral engagement. However, these trends vary within key segments of the population and indicate that while climate communication may be shifting the beliefs of many, strategies for reaching particular audiences may need to be adapted.

Recommended citation: Ayers, M., Marlon, J.R., Ballew, M.T. et al. Changes in Global Warming’s Six Americas: an analysis of repeat respondents. Climatic Change 177, 96 (2024). https://doi.org/10.1007/s10584-024-03754-x
Download Paper