Ravi Shroff
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ravi Shroff has a total of 37 co-authors. Here are the top co-authors by number of shared publications:
- Johann D. Gaebler
Department of Statistics, Harvard University
3 shared publications - Sharad Goel
Kennedy School of Government, Harvard University
3 shared publications - Alan L. Mendelsohn
Department of Pediatrics, NYU Grossman School of Medicine, New York, NY
1 shared publication André Minor
University of California at San Diego
1 shared publication- Anne Martin
Department of Psychology, Georgetown University
1 shared publication - Ashleigh I. Aviles
Institute of Human Development and Social Change (AI Aviles), New York University, New York, NY
1 shared publication
Latest Publications
- Open Access
- 20 July 2026
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (30)
Measuring disparate impact in human and machine decisions
Empirical analyses have grown increasingly important in discrimination litigation with the greater availability of detailed data on individuals and decisions. A popular analytic strategy is to estimate disparities after adjusting for observed covariates, typically with a regression model, in hopes of ferreting out discriminatory intent. This approach, however, is ill-suited to auditing algorithms that are now commonly used to aid decisions, which typically do not include race or other legally protected factors as inputs. Motivated by legal understandings of disparate impact, we introduce an approach that aims to measure “unjustified” disparities in both human and machine decisions. Our method, which we call risk-adjusted regression, proceeds in three steps. In the first step, we combine all available information in a machine learning model to estimate the value, or inversely, the risk, of taking a certain action, such as approving a loan application or hiring a job candidate. Second, we measure disparities in decisions after adjusting for these risk estimates alone. Finally, in the third step, we assess the sensitivity of results to potential mismeasurement of risk. We demonstrate this approach on a detailed dataset of 2.2 million police stops of pedestrians in New York City, and show that traditional statistical tests of discrimination can substantially understate the magnitude of (risk-adjusted) racial disparities.
- 20 July 2026
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 123 (30)
Open AccessMeasuring disparate impact in human and machine decisions
Empirical analyses have grown increasingly important in discrimination litigation with the greater availability of detailed data on individuals and decisions. A popular analytic strategy is to estimate disparities after adjusting for observed covariates, typically with a regression model, in hopes of ferreting out discriminatory intent. This approach, however, is ill-suited to auditing algorithms that are now commonly used to aid decisions, which typically do not include race or other legally protected factors as inputs. Motivated by legal understandings of disparate impact, we introduce an approach that aims to measure “unjustified” disparities in both human and machine decisions. Our method, which we call risk-adjusted regression, proceeds in three steps. In the first step, we combine all available information in a machine learning model to estimate the value, or inversely, the risk, of taking a certain action, such as approving a loan application or hiring a job candidate. Second, we measure disparities in decisions after adjusting for these risk estimates alone. Finally, in the third step, we assess the sensitivity of results to potential mismeasurement of risk. We demonstrate this approach on a detailed dataset of 2.2 million police stops of pedestrians in New York City, and show that traditional statistical tests of discrimination can substantially understate the magnitude of (risk-adjusted) racial disparities.
- 14 December 2024
- Published by Elsevier in Social Science & Medicine
- Vol. 366, 117636
Economic hardships during COVID-19 and maternal mental health: Combining samples with low incomes across three cities
- 14 December 2024
- Published by Elsevier in Social Science & Medicine
- Vol. 366, 117636
Economic hardships during COVID-19 and maternal mental health: Combining samples with low incomes across three cities
- Open Access
- 3 June 2024
- Published by MIT Press in Neurobiology of Language
- Vol. 5 (2) , 432-453
Reactive Inhibitory Control Precedes Overt Stuttering Events
Research points to neurofunctional differences underlying fluent speech between stutterers and non-stutterers. Considerably less work has focused on processes that underlie stuttered vs. fluent speech. Additionally, most of this research has focused on speech motor processes despite contributions from cognitive processes prior to the onset of stuttered speech. We used MEG to test the hypothesis that reactive inhibitory control is triggered prior to stuttered speech. Twenty-nine stutterers completed a delayed-response task that featured a cue (prior to a go cue) signaling the imminent requirement to produce a word that was either stuttered or fluent. Consistent with our hypothesis, we observed increased beta power likely emanating from the right pre-supplementary motor area (R-preSMA)—an area implicated in reactive inhibitory control—in response to the cue preceding stuttered vs. fluent productions. Beta power differences between stuttered and fluent trials correlated with stuttering severity and participants’ percentage of trials stuttered increased exponentially with beta power in the R-preSMA. Trial-by-trial beta power modulations in the R-preSMA following the cue predicted whether a trial would be stuttered or fluent. Stuttered trials were also associated with delayed speech onset suggesting an overall slowing or freezing of the speech motor system that may be a consequence of inhibitory control. Post-hoc analyses revealed that independently generated anticipated words were associated with greater beta power and more stuttering than researcher-assisted anticipated words, pointing to a relationship between self-perceived likelihood of stuttering (i.e., anticipation) and inhibitory control. This work offers a neurocognitive account of stuttering by characterizing cognitive processes that precede overt stuttering events.
- 3 June 2024
- Published by MIT Press in Neurobiology of Language
- Vol. 5 (2) , 432-453
Open AccessReactive Inhibitory Control Precedes Overt Stuttering Events
Research points to neurofunctional differences underlying fluent speech between stutterers and non-stutterers. Considerably less work has focused on processes that underlie stuttered vs. fluent speech. Additionally, most of this research has focused on speech motor processes despite contributions from cognitive processes prior to the onset of stuttered speech. We used MEG to test the hypothesis that reactive inhibitory control is triggered prior to stuttered speech. Twenty-nine stutterers completed a delayed-response task that featured a cue (prior to a go cue) signaling the imminent requirement to produce a word that was either stuttered or fluent. Consistent with our hypothesis, we observed increased beta power likely emanating from the right pre-supplementary motor area (R-preSMA)—an area implicated in reactive inhibitory control—in response to the cue preceding stuttered vs. fluent productions. Beta power differences between stuttered and fluent trials correlated with stuttering severity and participants’ percentage of trials stuttered increased exponentially with beta power in the R-preSMA. Trial-by-trial beta power modulations in the R-preSMA following the cue predicted whether a trial would be stuttered or fluent. Stuttered trials were also associated with delayed speech onset suggesting an overall slowing or freezing of the speech motor system that may be a consequence of inhibitory control. Post-hoc analyses revealed that independently generated anticipated words were associated with greater beta power and more stuttering than researcher-assisted anticipated words, pointing to a relationship between self-perceived likelihood of stuttering (i.e., anticipation) and inhibitory control. This work offers a neurocognitive account of stuttering by characterizing cognitive processes that precede overt stuttering events.
- Open Access
The Measure and Mismeasure of Fairness
The field of fair machine learning aims to ensure that decisions guided by algorithms are equitable. Over the last decade, several formal, mathematical definitions of fairness have gained prominence. Here we first assemble and categorize these definitions into two broad families: (1) those that constrain the effects of decisions on disparities; and (2) those that constrain the effects of legally protected characteristics, like race and gender, on decisions. We then show, analytically and empirically, that both families of definitions typically result in strongly Pareto dominated decision policies. For example, in the case of college admissions, adhering to popular formal conceptions of fairness would simultaneously result in lower student-body diversity and a less academically prepared class, relative to what one could achieve by explicitly tailoring admissions policies to achieve desired outcomes. In this sense, requiring that these fairness definitions hold can, perversely, harm the very groups they were designed to protect. In contrast to axiomatic notions of fairness, we argue that the equitable design of algorithms requires grappling with their context-specific consequences, akin to the equitable design of policy. We conclude by listing several open challenges in fair machine learning and offering strategies to ensure algorithms are better aligned with policy goals.
Open AccessThe Measure and Mismeasure of Fairness
The field of fair machine learning aims to ensure that decisions guided by algorithms are equitable. Over the last decade, several formal, mathematical definitions of fairness have gained prominence. Here we first assemble and categorize these definitions into two broad families: (1) those that constrain the effects of decisions on disparities; and (2) those that constrain the effects of legally protected characteristics, like race and gender, on decisions. We then show, analytically and empirically, that both families of definitions typically result in strongly Pareto dominated decision policies. For example, in the case of college admissions, adhering to popular formal conceptions of fairness would simultaneously result in lower student-body diversity and a less academically prepared class, relative to what one could achieve by explicitly tailoring admissions policies to achieve desired outcomes. In this sense, requiring that these fairness definitions hold can, perversely, harm the very groups they were designed to protect. In contrast to axiomatic notions of fairness, we argue that the equitable design of algorithms requires grappling with their context-specific consequences, akin to the equitable design of policy. We conclude by listing several open challenges in fair machine learning and offering strategies to ensure algorithms are better aligned with policy goals.
- Open Access
- 1 July 2022
- Published by Cambridge University Press (CUP) in Judgment and Decision Making
- Vol. 17 (6) , 1176-1207
Pretrial release judgments and decision fatigue
Field studies in many domains have found evidence of decision fatigue, a phenomenon describing how decision quality can be impaired by the act of making previous decisions. Debate remains, however, over posited psychological mechanisms underlying decision fatigue, and the size of effects in high-stakes settings. We examine an extensive set of pretrial arraignments in a large, urban court system to investigate how judicial release and bail decisions are influenced by the time an arraignment occurs. We find that release rates decline modestly in the hours before lunch and before dinner, and these declines persist after statistically adjusting for an extensive set of observed covariates. However, we find no evidence that arraignment time affects pretrial release rates in the remainder of each decision-making session. Moreover, we find that release rates remain unchanged after a meal break even though judges have the opportunity to replenish their mental and physical resources by resting and eating. In a complementary analysis, we find that the rate at which judges concur with prosecutorial bail requests does not appear to be influenced by either arraignment time or a meal break. Taken together, our results imply that to the extent that decision fatigue plays a role in pretrial release judgments, effects are small and inconsistent with previous explanations implicating psychological depletion processes.
- 1 July 2022
- Published by Cambridge University Press (CUP) in Judgment and Decision Making
- Vol. 17 (6) , 1176-1207
Open AccessPretrial release judgments and decision fatigue
Field studies in many domains have found evidence of decision fatigue, a phenomenon describing how decision quality can be impaired by the act of making previous decisions. Debate remains, however, over posited psychological mechanisms underlying decision fatigue, and the size of effects in high-stakes settings. We examine an extensive set of pretrial arraignments in a large, urban court system to investigate how judicial release and bail decisions are influenced by the time an arraignment occurs. We find that release rates decline modestly in the hours before lunch and before dinner, and these declines persist after statistically adjusting for an extensive set of observed covariates. However, we find no evidence that arraignment time affects pretrial release rates in the remainder of each decision-making session. Moreover, we find that release rates remain unchanged after a meal break even though judges have the opportunity to replenish their mental and physical resources by resting and eating. In a complementary analysis, we find that the rate at which judges concur with prosecutorial bail requests does not appear to be influenced by either arraignment time or a meal break. Taken together, our results imply that to the extent that decision fatigue plays a role in pretrial release judgments, effects are small and inconsistent with previous explanations implicating psychological depletion processes.
- Top Cited PapersOpen Access
- 25 February 2022
- Published by Taylor & Francis in Statistics and Public Policy
- Vol. 9 (1) , 26-48
A Causal Framework for Observational Studies of Discrimination
In studies of discrimination, researchers often seek to estimate a causal effect of race or gender on outcomes. For example, in the criminal justice context, one might ask whether arrested individuals would have been subsequently charged or convicted had they been a different race. It has long been known that such counterfactual questions face measurement challenges related to omitted-variable bias, and conceptual challenges related to the definition of causal estimands for largely immutable characteristics. Another concern, which has been the subject of recent debates, is post-treatment bias: many studies of discrimination condition on apparently intermediate outcomes, like being arrested, that themselves may be the product of discrimination, potentially corrupting statistical estimates. There is, however, reason to be optimistic. By carefully defining the estimand—and by considering the precise timing of events—we show that a primary causal quantity of interest in discrimination studies can be estimated under an ignorability condition that may hold approximately in some observational settings. We illustrate these ideas by analyzing both simulated data and the charging decisions of a prosecutor’s office in a large county in the United States.
- 25 February 2022
- Published by Taylor & Francis in Statistics and Public Policy
- Vol. 9 (1) , 26-48
Top Cited PapersOpen AccessA Causal Framework for Observational Studies of Discrimination
In studies of discrimination, researchers often seek to estimate a causal effect of race or gender on outcomes. For example, in the criminal justice context, one might ask whether arrested individuals would have been subsequently charged or convicted had they been a different race. It has long been known that such counterfactual questions face measurement challenges related to omitted-variable bias, and conceptual challenges related to the definition of causal estimands for largely immutable characteristics. Another concern, which has been the subject of recent debates, is post-treatment bias: many studies of discrimination condition on apparently intermediate outcomes, like being arrested, that themselves may be the product of discrimination, potentially corrupting statistical estimates. There is, however, reason to be optimistic. By carefully defining the estimand—and by considering the precise timing of events—we show that a primary causal quantity of interest in discrimination studies can be estimated under an ignorability condition that may hold approximately in some observational settings. We illustrate these ideas by analyzing both simulated data and the charging decisions of a prosecutor’s office in a large county in the United States.
- 1 September 2017
- Published by SAGE Publications in Big Data
- Vol. 5 (3) , 189-196
Predictive Analytics for City Agencies: Lessons from Children's Services
Many municipal agencies maintain detailed and comprehensive electronic records of their interactions with citizens. These data, in combination with machine learning and statistical techniques, offer the promise of better decision making, and more efficient and equitable service delivery. However, a data scientist employed by an agency to implement these techniques faces numerous and varied choices that cumulatively can have significant real-world consequences. The data scientist, who may be the only person at an agency equipped to understand the technical complexity of a predictive algorithm, therefore, bears a good deal of responsibility in making judgments. In this perspective, I use a concrete example from my experience of working with New York City's Administration for Children's Services to illustrate the social and technical tradeoffs that can result from choices made in each step of data analysis. Three themes underlie these tradeoffs: the importance of frequent communication between the data scientist, agency leadership, and domain experts; the agency's resources and organizational constraints; and the necessity of an ethical framework to evaluate salient costs and benefits. These themes inform specific recommendations that I provide to guide agencies that employ data scientists and rely on their work in designing, testing, and implementing predictive algorithms.
- 1 September 2017
- Published by SAGE Publications in Big Data
- Vol. 5 (3) , 189-196
Predictive Analytics for City Agencies: Lessons from Children's Services
Many municipal agencies maintain detailed and comprehensive electronic records of their interactions with citizens. These data, in combination with machine learning and statistical techniques, offer the promise of better decision making, and more efficient and equitable service delivery. However, a data scientist employed by an agency to implement these techniques faces numerous and varied choices that cumulatively can have significant real-world consequences. The data scientist, who may be the only person at an agency equipped to understand the technical complexity of a predictive algorithm, therefore, bears a good deal of responsibility in making judgments. In this perspective, I use a concrete example from my experience of working with New York City's Administration for Children's Services to illustrate the social and technical tradeoffs that can result from choices made in each step of data analysis. Three themes underlie these tradeoffs: the importance of frequent communication between the data scientist, agency leadership, and domain experts; the agency's resources and organizational constraints; and the necessity of an ethical framework to evaluate salient costs and benefits. These themes inform specific recommendations that I provide to guide agencies that employ data scientists and rely on their work in designing, testing, and implementing predictive algorithms.
- Open Access
- 1 October 2014
- Published by International Press of Boston in Arkiv för Matematik
- Vol. 52 (2) , 301-327
CR singular images of generic submanifolds under holomorphic maps
Project Euclid - mathematics and statistics online
- 1 October 2014
- Published by International Press of Boston in Arkiv för Matematik
- Vol. 52 (2) , 301-327
Open AccessCR singular images of generic submanifolds under holomorphic maps
Project Euclid - mathematics and statistics online