A5101466158- Advanced Causal Inference Techniques
- Immune Cell Function and Interaction
- T-cell and B-cell Immunology
- Health Systems, Economic Evaluations, Quality of Life
- Immunotherapy and Immune Responses
- Immune Response and Inflammation
- Advanced Bandit Algorithms Research
- Ethics and Social Impacts of AI
- Crime Patterns and Interventions
- Statistical Methods in Clinical Trials
- Criminal Justice and Corrections Analysis
- Explainable Artificial Intelligence (XAI)
- Statistical Methods and Inference
- Healthcare Policy and Management
- Economic theories and models
- Semantic Web and Ontologies
- Machine Learning in Healthcare
- Distributed Sensor Networks and Detection Algorithms
- Age of Information Optimization
- Law, Economics, and Judicial Systems
- Mobile Crowdsensing and Crowdsourcing
- Insurance, Mortality, Demography, Risk Management
- Auction Theory and Applications
- Income, Poverty, and Inequality
- Seismic Imaging and Inversion Techniques
Sandia National Laboratories
2023
University of Southern California
2023
University Health Network
2022
Toronto General Hospital
2022
University of Toronto
2014-2021
Cornell University
2018-2021
Canada Research Chairs
2018-2021
Sandia National Laboratories California
2020-2021
University of California, Berkeley
2021
Centre for Social Innovation
2016
CN102453712siRNA PI4KB Inhibition of SARS coronavirus infection using siRNA molecules
The increasing impact of algorithmic decisions on people's lives compels us to scrutinize their fairness and, in particular, the disparate impacts that ostensibly-color-blind algorithms can have different groups. Examples include credit decisioning, hiring, advertising, criminal justice, personalized medicine, and targeted policymaking, where some cases legislative or regulatory frameworks for exist define specific protected classes. In this paper we study a fundamental challenge assessing...
The increasing impact of algorithmic decisions on people’s lives compels us to scrutinize their fairness and, in particular, the disparate impacts that ostensibly color-blind algorithms can have different groups. Examples include credit decisioning, hiring, advertising, criminal justice, personalized medicine, and targeted policy making, where some cases legislative or regulatory frameworks for exist define specific protected classes. In this paper we study a fundamental challenge assessing...
The induction of long-lived heterotypic T-cell protection against influenza virus remains elusive, despite the conservation epitopes. is critically dependent on lung-resident memory T cells (Trm). Here we show that intranasal administration 4-1BBL along with nucleoprotein in a replication-defective adenovirus vector to pre-immune mice induces remarkably stable circulating effector CD8 population characterized by higher IL-7Rα expression than control-boosted cells, as well substantial lung...
We study the problem of learning personalized decision policies from observational data while accounting for possible unobserved confounding. Previous approaches, which assume unconfoundedness, i.e., that no confounders affect both treatment assignment as well outcome, can lead to introduce harm rather than benefit when some confounding is present, generally case with data. Instead, since policy value and regret may not be point-identifiable, we a method minimizes worst-case estimated...
Estimating the causal effects of an intervention on outcomes is crucial. But often in domains such as healthcare and social services, this critical information about documented by unstructured text, e.g. clinical notes or case services. For example, street outreach to homeless populations a common services intervention, with ambiguous hard-to-measure outcomes. Outreach workers compile note records which are informative Although experts can succinctly extract relevant from notes, it costly...
CD4 T cells are critical for control of persistent infections; however, the key signals that regulate help during chronic infection remain incompletely defined. While several studies have addressed role inhibitory receptors and soluble factors such as PD-1 IL-10, significantly less work has cell co-stimulatory molecules viral infection. Here we show a with lymphocytic choriomeningitis virus (LCMV) clone 13, mice lacking glucocorticoid-induced tumor necrosis factor receptor related protein...
We study the problem of policy evaluation and learning from batched contextual bandit data when treatments are continuous, going beyond previous work on discrete treatments. Previous for treatment/action spaces focuses inverse probability weighting (IPW) doubly robust (DR) methods that use a rejection sampling approach equivalent weighted classification learning. In continuous setting, this reduction fails as we would almost surely reject all observations. To tackle case treatments, extend...
Recent work in fairness machine learning has proposed adjusting for by equalizing accuracy metrics across groups and also studied how datasets affected historical prejudices may lead to unfair decision policies. We connect these lines of study the residual unfairness that arises when a fairness-adjusted predictor is not actually fair on target population due systematic censoring training data existing biased This scenario particularly common same applications where concern. characterize...
Abstract The TNFR superfamily member 4-1BB is important in the establishment of tissue-resident memory T cells (Trm) lung tissue following influenza infection. Moreover, supraphysiological boosting airways during boost phase a prime-boost immunization regimen increases long-lived Trm population, correlating with increased protection against heterotypic challenge. However, little known about how contributes to population. In this study, we show that effects on accumulation are already...
We study the interplay of fairness, welfare, and equity considerations in personalized pricing based on customer features. Sellers are increasingly able to conduct price personalization predictive modeling demand conditional covariates: setting customized interest rates, targeted discounts consumer goods, subsidies scarce resources with positive externalities like vaccines bed nets. These different application areas may lead concerns around objectives: burdens consumers, envy, firm revenue,...
Risk assessment instrument (RAI) datasets, particularly ProPublica's COMPAS dataset, are commonly used in algorithmic fairness papers due to benchmarking practices of comparing algorithms on datasets prior work. In many cases, this data is as a benchmark demonstrate good performance without accounting for the complexities criminal justice (CJ) processes. However, we show that pretrial RAI can contain numerous measurement biases and errors, disparities discretion deployment, applied limited...
Where machine-learned predictive risk scores inform high-stakes decisions, such as bail and sentencing in criminal justice, fairness has been a serious concern. Recent work characterized the disparate impact that can have when used for binary classification task. This may not account, however, more diverse downstream uses of their non-binary nature. To better account this, this paper, we investigate from point view bipartite ranking task, where one seeks to rank positive examples higher than...
Abstract The costimulatory TNFR family member GITR can provide important survival signals for CD8 T cells. However, little is known about the regulation of this pathway during a chronic infection. In study, we show that ligand (GITRL) maximally induced on APCs at day 2 post–lymphocytic choriomeningitis virus (LCMV) clone 13 infection, but downregulated to below baseline levels by 8 postinfection (p.i.), and remains so stage At its peak, GITRL expression highest macrophages, with lower...
We study the problem of learning personalized decision policies from observational data while accounting for possible unobserved confounding. Previous approaches, which assume unconfoundedness, that is, no confounders affect both treatment assignment as well outcome, can lead to introduce harm rather than benefit when some confounding is present generally case with data. Instead, because policy value and regret may not be point-identifiable, we a method minimizes worst-case estimated...
Two quinone methide (QM) metabolites of the phenolic antioxidant butylated hydroxytoluene (BHT), 2,6-di-tert-butyl-4-methylenecyclohexa-2,5-dienone (BHT−QM) and tert-butyl-hydroxylated derivative (BHTOH−QM), are believed to be responsible for promoting lung tumor formation in mice treated with BHT. QMs strongly electrophilic undergo Michael type additions nucleophiles at exocyclic methylene form benzylic thioether adducts. Our goal was identify intracellular protein targets these order gain...