Concerns about potential fetal harm caused by alcohol and drug use have motivated policymakers and public health agencies to dedicate increasing attention and resources to limiting substance use in the context of pregnancy over the past 50 years. Most states now have at least one pregnancy-specific alcohol and drug policy in place. Public health agencies have issued warnings and recommended abstinence before, during, and after pregnancy. Clinical counseling interventions for patients who use alcohol and drugs typically focus exclusively on contraception and the prevention of pregnancy without attention to nuanced pregnancy preferences and reproductive autonomy. This dissertation explores causal inference and multi-level modeling approaches to assess the intended and unintended effects of policies focused on alcohol and drug use in the context of pregnancy, taking into consideration both fetal health and reproductive autonomy. Additionally, it applies a reproductive autonomy framework to questions related to substance use and pregnancy, pushing back against traditional public health frameworks, which are focused on preventing “unintended” pregnancy and neglect to account for individual pregnancy preferences.Chapter 1 assesses whether the oldest and most common type of pregnancy-specific alcohol and drug use policy is associated with changes in birth outcomes. Specifically, we investigated the impacts of Arizona’s alcohol-only and alcohol-plus-drug child abuse and neglect policies on birth outcomes. We used state- level policy information and vital statistics data from 1.5 million singleton births conceived between January 2000 and December 2012 in Arizona and New Mexico and conducted quasi-experimental difference-in-difference approaches to compare pre-to-post policy changes in gestational length and birthweight between the states. Stratified subgroup analyses examined policy associations by race and ethnicity. Contrary to hypotheses, we found apparent population-level improvements in birthweight, and non-significant changes for gestational age, associated with Arizona’s alcohol and drug child abuse and neglect policies relative to New Mexico, where neither policy was implemented. Improvements associated with the policy were specifically among non-Hispanic white and Hispanic/Latinx births. Due to the close timing of the tobacco policies and the CACN policies of interest (particularly when considering lags in policy implementation), it was not possible to distinguish between changes in birth outcomes associated with the tobacco policies and the policies of interest. Given rigorous evidence showing associations between tobacco smoking and adverse birth outcomes and between tobacco control policies and reductions in smoking behavior (including among pregnant people), unexpected improvements in birthweight may be due to co-commencing tobacco policies instead of CACN policies. Chapter 2 evaluates associations between state-level legal recreational cannabis and three outcomes: cannabis use during pregnancy, beliefs about safety of cannabis use during pregnancy, and perceptions of community stigma around cannabis use during pregnancy. We used cross-sectional multi-level survey data collected from 3,571 pregnant or recently pregnant people in 37 states in 2022 and fit generalized linear models and mixed effects ordinal logistic regression models with random effects for state. Findings do not support concerns that legal recreational cannabis is associated with cannabis use during pregnancy or beliefs about safety. Yet, results indicate that legal recreational cannabis may be associated with lower community stigma around cannabis use during pregnancy, which could have implications for pregnant people’s disclosure of use and care-seeking behavior. Chapter 3 investigates relationships between alcohol and drug use, the desire to avoid pregnancy, and incident pregnancy using longitudinal survey data collected from 2,015 individuals with the capacity to become pregnant who were recruited from 23 clinics in 5 Southwestern states from 2019-2023. We used multi-level linear mixed effects regression to explore associations between alcohol, cannabis, or illicit drug use and desire to avoid pregnancy and we used cox proportional hazard models to explore whether there was significant interaction between desire to avoid pregnancy and alcohol, cannabis, or illicit drug use in predicting incident pregnancy. Results indicate that more frequent cannabis use and any illicit drug use, but not heavy alcohol use, were associated with higher desire to avoid pregnancy and that the attainment of pregnancy preferences (or the association between desire to avoid pregnancy and incident pregnancy) differed significantly by heavy alcohol use, but not by cannabis or illicit drug use. Participants who reported heavy alcohol use were less likely to attain their pregnancy preferences over one year relative to those who did not report heavy alcohol use, with those most wanting to avoid pregnancy at significantly increased risk of pregnancy if they reported heavy alcohol use. Among those more open to pregnancy, there was no significant difference in the rate of incident pregnancy by heavy alcohol use.

This study introduces the concept of “asymptotic labor” through three case studies examining the political economy of contemporary information-processing systems. It aims to contribute additional historical and theoretical perspective to the strong foundation of existing critical research that has revealed the “hidden” human workforce that props up the vast digital infrastructure of artificial intelligence. The first chapter chronicles the design and development of CAPTCHA and reCAPTCHA, suggesting a critical periodization of these cybersecurity systems based on their different methods of validating human and nonhuman users. The evolution of CAPTCHA is deeply intertwined with the rapid ascendance of machine learning as the dominant form of artificial intelligence in the mid-2000s, and presages the emergent methods of value capture that undergird these data-intensive systems. The next chapter builds on this latter premise, utilizing semiotics as a method for dissecting the mechanisms of meaning-making and value production at the core of these complex information-processing systems, as well as the ways that they have become implicated in a broader set of political-economic conditions. Finally, an ethnographic account of a specialized class of workers at the United States Postal Service ties together the more theoretically-laden arguments of the preceding chapters by demonstrating the social, political, and material implications that the development and implementation of these information-processing systems have on the dwindling number of humans that remain embroiled in their continuing operations.