Browsing by Subject "Risk assessment"

A moral dilemma confronts criminal justice in unjust states. If the state punishes marginalized citizens whose crimes are connected to conditions of systemic injustice the state has failed to alleviate, it perpetuates a further injustice to those citizens. If the state does not punish, it perpetuates an injustice to victims of crime whose protection is the duty of the criminal justice system. Thus, no reaction to crime by the unjust state appears to avoid perpetuating further injustice. Tommie Shelby proposes a new solution to this old dilemma, suggesting that certain theoretical and practical qualifications can save the unjust state from perpetuating injustice. He argues that punishment can be just even as society remains unjust if it is: (a) administered through a fair criminal justice apparatus; (b) only directed at mala in se crimes; and (c) not expressive of moral judgment. In the first part of this thesis, I explore Shelby’s solution to show that certain aspects of his framework are superior to alternative ones, but that it nonetheless fails to resolve the dilemma. In Part 2, I use a novel technological reform that promises to make criminal justice fairer, the AI risk assessment, as a case study to show why even punishment that meets Shelby’s criteria will continue to perpetuate injustice as long as it operates under systemic social injustice. Punishment can only be just if society is.

Background

Methods

Results

Conclusions

Trial registration

Objectives

Methods

Results

Conclusion

Key points

This dissertation encompasses three essays on policing and criminal justice, algorithms and inequality. The first two essays examine the efficacy and equity implications of data-driven algorithms that are increasingly used in important life-altering decision-making contexts. The third essay investigates when crime responds to punishment.

The first essay studies the impacts of neighborhood targeting of police presence brought about by predictive policing algorithms on crime and arrests. While predictive policing is widely used, the impacts of neighborhood targeting brought about by predictive policing on crime, and whether there are disproportionate racial impacts are open questions. Using a novel dataset, I isolate quasi-experimental variation in police presence induced by predictive-policing algorithms to estimate the causal impacts of algorithm-induced police presence. I find that algorithm-induced police presence decreases serious violent and property crime, and evidence that algorithm-induced neighborhood targeting of police presence has disproportionate racial impacts on traffic incident arrests and serious violent crime incident arrests.

The second essay investigates how data-driven algorithms can maximize overall predictive power at the cost of racial and economic justice. Examining a tool that is already widely used in pretrial decision-making, I build a framework to evaluate how input variables trades off overall predictive power, and racial and economic disparities in the scores that defendants receive. I find that using information on neighborhoods where defendants live only marginally contributes to overall predictive power. However, the use of defendant neighborhood data substantially increases racial and economic disparities, suggesting that machine learning objectives tuned to maximize overall predictive power risk being in conflict with racial and economic justice.

Finally, in the third essay, joint with Sarah Komisarow and Robert Gonzalez, we examine when crime responds to punishment severity increases. While economic theory suggests that crime should respond to punishment severity, empirical evidence on this link is ambiguous. We propose an explanation for this empirical evidence -- the effect of punishment severity increases depends on the probability of detection; punishments deter crime when the probability of detection is moderate. We test and validate this explanation using increases in punishment severity in drug-free school zones along with changes in the probability of detection resulting from a community crime-monitoring program.

Nanomaterials are being increasingly produced and used across a myriad of applications while their novel properties are still in the midst of being designed and explored. Thus the full implications of introducing these materials into the environment cannot be understood, yet the need to assess potential risks is already upon us. This work discusses a comprehensive view of environmental impact with respect to material flows from across the value chain into all compartments of the environment, whereby interactions and potential hazardous effects become possible. A subset of this broad system is then chosen for evaluation; a model is derived to describe the fate of nanomaterials released to wastewater.

This analysis considers the wastewater treatment plant (WWTP) as a complete mixed reactor aerobic secondary clarifier, and predicts whether nanomaterials will associate with effluent or sludge to project potential concentrations in each. The concentration of nanomaterials reaching a WWTP is estimated based on a linear weighting of total production, and the fate of nanomaterials within the WWTP is based on a characteristic inherent to the material, partition coefficient, and on design parameters of the WWTP, such as retention times and suspended solids concentration.

Due to the uncertainty inherent to this problem, a probabilistic approach is employed. Monte Carlo simulation is used, sampling from probability distributions assigned to each of the input parameters to calculate a distribution for the predicted concentrations in sludge and effluent. Input parameter distributions are estimated from values reported in the literature where possible. Where data do not yet exist, studies are carried out to enable parameter estimation. In particular, nanomaterial production is investigated to provide a basis to estimate the magnitude of potential exposure. Nanomaterial partitioning behavior is also studied in this work, through laboratory experiments for several types of nano-silver.

The results presented here illustrate the use of nanomaterial inventory data in predicting environmentally relevant concentrations. Estimates of effluent and sludge concentrations for nano-silver with four different types coatings suggest that these surface treatments affect the removal efficiency; the same nanomaterial with different coatings may have different environmental fates. Effluent concentration estimates for C60 and nano-TiO2 suggest that these nanomaterials could already be present at problematic concentrations at current levels of annual production.

Estimates of environmentally relevant concentrations may aid in interpretation of nanotoxicology studies. These relative estimates are also useful in that they may help inform future decisions regarding where to dedicate resources for future research. Beyond attempting to estimate environmental concentrations of nanomaterials, this type of streamlined model allows the consideration of scenarios, focusing on what happens as various input parameters change. Production quantity and the fraction of this quantity that is released to wastewater are found to greatly influence the model estimates for wastewater effluent concentrations; in the case of wastewater sludge concentrations, the model is sensitive to those parameters in addition to solids retention time.

Introduction: Chronic diseases have been increasing globally for decades, while the leading chronic diseases worldwide are cancer, cardiovascular disease (CVD), chronic respiratory disease, and diabetes.1 Behavioral risk factors of chronic diseases that can be modified include physical activity, diet, alcohol consumption and tobacco use.3 4 Several guidelines for screening and prevention recommend that family health history (FHH) is collected by primary care providers for disease risk stratification and management.6 7 MeTree, developed in 2014, is a computerized, patient-facing program that collects information about family health history and generates decision support for providers and patients.6 15 There are several potential barriers to implementation of an online FHH software tool including health literacy, computer skills, and behavioral components. This study collects FHH information through MeTree in a rural population in North Carolina through a unique implementation process using research assistants to manually and verbally assist participants. The aims of this study are to characterize the quality of pedigrees collected and to estimate familial disease aggregation among the families of participants.

Methods: This study enrolled 44 participants from an ongoing study conducted by collaborators from Duke University Health System, Duke Clinical Research Institute,

University of North Carolina Pembroke, and Southeastern Regional Medical Center. To collect FHH information, participants constructed family pedigree in MeTree, one family member at a time with the help of one study research assistant. Once participants created a full family pedigree, an individual risk assessment was generated by MeTree.

Results: More than half of the participants were female (n= 30, 68.2%). The ethnic group that composed the largest part of our study population were Lumbee Indians (n=23, 52.3%) followed by White/Caucasians (n=13, 29.5%) and African Americans (n=7, 15.9%). For quality, the average score across all pedigrees was higher than 65% for all seven components of the criteria. The total number of diseases present among all participants and relatives in the study was 930 (Table 3). Cancer was present in 81.8% of pedigrees and made up 12.2% of all reported diseases. Twenty-five percent of all pedigrees had at least one family member that was diagnosed with lung cancer. Diabetes was also frequently reported and was observed in 75% of all pedigrees. Kidney Disease was reported in at least one or more relatives in 52.3% of pedigrees.

Conclusions: Using a patient-facing online Health Information Technology tool such as MeTree could potentially lead to better health outcomes due to risk assessment and individually-targeted prevention strategies. MeTree may be an important tool to use to address the large burden of chronic diseases in this region.

Natural steroidal estrogen hormones in swine wastes generated from Concentrated Animal Feeding Operations (CAFOs) have become a potential pollutant to many aquatic environments due to their adverse impacts on the reproductive biology of aquatic organisms. In North Carolina, the swine CAFO industry is a major agricultural economic enterprise that is responsible for the generation of large volumes of waste. However, there is limited scientific understanding regarding the concentration, fate, and transport of the estrogenic compounds from these swine facilities into terrestrial and aquatic environments. To address this issue, my research involved the development and application of integrated Bayesian networks (BNs) models that can be used to better characterize and assess the generation, fate, and transport of site-specific swine CAFO-derived estrogen compounds. The developed model can be used as decision support tool towards estrogen risk assessment. Modularized and melded BN approaches were used to capture the predictive and casual relationships of the estrogen budget and its movement within and between the three major systems of a swine farrowing CAFO. These systems include the animal barns, the anaerobic waste lagoon, and the spray fields. For the animal barn system, a facility-wide estrogen budget was developed to assess the operation-specific estrogen excretion, using an object-oriented BN (OOBN) approach. The developed OOBN model provides a means to estimate and predict estrogen fluxes from the whole swine facility in the context of both estrogen type and animal operating unit. It also accounts for the uncertainties in the data and in our understanding of the system. Next, mass balance melding BN models were developed to predict the natural estrogen fates and budgets in two lagoon compartments, the slurry and the sludge storage. This involved utilizing mass balance equations to account for the mechanisms of flushing, sorption, transformation, settling, and burial reactions of estrogen compounds in the slurry and sludge storages. As an alternative approach, a regression based BN melding approach was developed to both characterize estrogen fate and budgets as a result of the sequential transformation processes between natural estrogen compounds and to assess the seasonal effects on the estrogen budgets in the two different lagoon compartments. Finally, a dynamic BN model was developed to characterize rainfall-driven estrogen runoff processes from the spray fields. The dynamic BN models were used to assess the potential risk of estrogen runoff to adjacent waterways. In addition, the dynamic model was used to quantify the effects of manure application rates, rainfall frequency, the time of rainfall and irrigation, crop types, on-farm best management practices, seasonal variability, and successive rainfall and manure application events on estrogen runoff.

The model results indicate that the farrowing barn is the biggest contributor of total estrogen as compared to the breeding and gestation operating barns. Once the estrogen reaches the anaerobic lagoon, settling and burial reactions were shown to be the most significant factors influencing estrogen levels in the slurry and sludge, respectively. The estrogen budgets in the lagoon were also found to vary by season, with higher slurry and sludge estrogen levels in the spring as compared to the summer. The risk of estrogen runoff was predicted to be lower in the summer as compared to the spring, primarily due to the spray field crop management plans adopted. The results also indicated that Bermuda grass performed more favorably when compared to soybean, when it came to retaining surface water runoff in the field. Model predictions indicated that there is a low risk of estrogen runoff losses from the spray fields under multiple irrigation and rainfall events, unless the time interval between irrigation was less than 10 days and/or in the event of a prolonged high magnitude rainstorm event. Overall, the estrone was the most persistent form of natural estrogens in the three major systems of the swine farrowing CAFO.

Chemicals in consumer products can influence our risk for developing adverse health conditions. This research addresses knowledge gaps in our ability to evaluate chemical safety, particularly for emerging substances on the market. Acknowledging the need for more high-throughput exposure and hazard models to support risk assessment, computational frameworks leveraging machine learning strategies and "big data" from public databases and mass social data sources were tested.

First, to understand consumer exposure, we require a better understanding of ingredient concentrations in products. A computational framework was developed to estimate chemical weight fractions for consumer products containing emerging substances. Nanomaterial-enabled products were used as a case study to represent such substances with limited physicochemical property data. Feature variables included chemical properties, functional use categories (e.g., antimicrobial), the type of product and its matrix. Weight fractions were classified as low, medium or high using a random forest or nonlinear support vector classifier. Performance of machine learning models was qualitatively compared with that of models from a second framework trained on data-rich, bulk-scale organic chemical product data. Models could roughly stratify material-product observations into weight fraction bins with moderate success. The best model achieved an average balanced accuracy of 73% on nanomaterials product data. Chemical functional use features served as particularly insightful predictors, suggesting that functional use data may be useful in evaluating the safety and sustainability of emerging chemicals. Investment in chemical and product data collection could see continued improvement of such machine learning models.

Shifting focus to the impact of chemicals on consumers, data on personal care products, ingredients, and customer reviews from online retailers and databases was collected to see if certain chemicals might increase risk of adverse reactions to products. The study scope was narrowed to shampoo products for hypothesis testing. Processing steps in the data pipeline included informatics and machine learning methods, namely, natural language processing for interpreting product reviews, text extraction from images of product labels, and feature reduction using chemical structure and ingredient source data. Fifty-one ingredient clusters were identified as having a significant correlation with higher adverse reaction rates in consumers when present in shampoos. Among these, there were a few common plant-based ingredients and synthetic preservatives known for causing skin sensitivity or irritation. In comparison with other constituents, however, the positively correlated ingredient groups had a general lack of published structural, physicochemical property and toxicity data. Results suggest an urgent need for targeted, higher-throughput chemical evaluations to safeguard consumers.

Together, these proof-of-concept studies progress our ability to quantify exposure and hazard of emerging and data-poor substances in consumer products. The outcomes of the computational frameworks can help prioritize potentially problematic substances for additional study to characterize risk.

Purpose of review

Recent findings

This work explores stochastic responses of various earth system models to different random sources, using polynomial chaos (PC) approaches. The following earth systems are considered, namely the HYbrid Coordinate Ocean Model (HYCOM, an ocean general circulation model (OGCM)) for the study of ocean circulation in the Gulf of Mexico (GoM); the Unified Wave INterface - Coupled Model (UWIN-CM, a dynamically coupled atmosphere-wave-ocean system) for Hurricane Earl (2010) modeling; and the earthquake seismology model for Bayesian inference of fault plane configurations.

In the OGCM study, we aim at analyzing the combined impact of uncertainties in initial conditions and wind forcing fields on ocean circulation using PC expansions. Empirical Orthogonal Functions (EOF) are used to represent both spatial perturbations of initial condition and space-time wind forcing fields, namely in the form of a superposition of modal components with uniformly distributed random amplitudes. The forward deterministic HYCOM simulations are used to propagate input uncertainties in ocean circulation in the GoM during the 2010 Deepwater Horizon (DWH) oil spill, and to generate a realization ensemble based on which PC surrogate models are constructed for both localized and field quantities of interest (QoIs), focusing specifically on Sea Surface Height (SSH) and Mixed Layer Depth (MLD). These PC surrogate models are constructed using Basis Pursuit DeNoising (BPDN) methodology, and their performance is assessed through various statistical measures. A global sensitivity analysis is then performed to quantify the impact of individual random sources as well as their interactions on ocean circulation. At the basin scale, SSH in the deep GoM is mostly sensitive to initial condition perturbations, while over the shelf it is sensitive to wind forcing perturbations. On the other hand, the basin MLD is almost exclusively sensitive to wind perturbations. For both quantities, the two random sources (initial condition and wind forcing) of uncertainties have limited interactions. Finally, computations indicate that whereas local quantities can exhibit complex behavior that necessitates a large number of realizations to build PC surrogate models, the modal analysis of field sensitivities can be suitably achieved with a moderate size ensemble.

It is noted that HYCOM simulations in the aforementioned OGCM study only focus on the ocean circulation, and ignore the oceanic feedback (e.g. momentum, energy, humidity etc) to the atmosphere. A more elaborated analysis is consequently performed to understand the atmosphere dynamics in a fully-coupled atmosphere-wave-ocean system. In particular, we explore the stochastic evolution of Hurricane Earl (2010) in response to uncertainties stemming from random perturbations in the storm's initial size, strength and rotational stretch. To this end, the UWIN-CM framework is employed as the forecasting system, which is used to propagate input uncertainties and generate a realization ensemble. PC surrogate models for time evolutions of both maximum wind speed and minimum sea level pressure (SLP) are constructed. These PC surrogates provide statistical insights on probability distributions of model responses throughout the simulation time span. Statistical analysis of rapid intensification (RI) process suggests that storms with enhanced initial intensity and counter-clockwise rotation perturbations are more likely to undergo a RI process. In addition, the RI process seems mostly sensitive to the mean wind strength and rotational stretch, rather than storm size and asymmetric wind amplitude perturbations. This is consistent with global sensitivity analysis of PC surrogate models. Finally we combine parametric storm perturbations with global stochastic kinetic energy backscatter (SKEBS) forcing in UWIN-CM simulations, and conclude that whereas the storm track is substantially influenced by global perturbations, it is weakly influenced by the properties of the initial storm.

The PC framework not only provides easy access to traditional statistical insights and global sensitivity indices, but also reduces the computational burden of sampling the system response, as performed for instance in Bayesian inference. These two advantages of PC approaches are well demonstrated in the study of earthquake seismology model response to random fault plane configurations. The PC statistical analysis suggests that the hypocenter location plays a dominant role in earthquake ground motion responses (in terms of peak ground velocities, PGVs), while elliptical patch properties only show secondary influence. In addition, Our PC based Bayesian analysis successfully identified the most `likely' fault plane configuration with respect to the chosen ground motion prediction equation (GMPE) curve, i.e. the hypocenter is more likely to be in the bottom right quadrant of the fault plane and the elliptical patch centers at the bottom left quadrant. To incorporate additional physical restrictions on fault plane configurations, a novel restricted sampling methodology is introduced. The results indicate that the restricted inference is more physically sensible, while retaining plausible consistency with findings from unrestricted inference.