Ananya Joshi
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
Ananya Joshi has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Alexander Rodríguez
University of Michigan, Ann Arbor, MI, USA
2 shared publications - Catalina Vajiac
Carnegie Mellon University, Pittsburgh, PA, USA
2 shared publications - Kai Wang
Georgia Institute of Technology, Atlanta, GA, USA
2 shared publications - Ajitesh Srivastava
Northeastern University, Charlotte, NC, USA
1 shared publication - Anil Vullikanti
University of Virginia
1 shared publication - Aparna Taneja
Google DeepMind, Gurugram, Haryana, India
1 shared publication
Latest Publications
- Open Access
- 28 October 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Sensitivity Analyses for Generative AI tasks in Psychiatry: LLM-Generated Hospitalization Risk Scores (Preprint)
BACKGROUND Generative artificial intelligence (AI), particularly large language models (LLMs), are increasingly studied for clinical reasoning and risk assessment tasks. However, AI outputs can be sensitive to prompt variations in the (1) contextual information provided and (2) prompt framing. The impacts of these variations remain insufficiently characterized in complex and inherently ambiguous domains such as psychiatry. OBJECTIVE We propose an approach for sensitivity analyses in downstream AI tasks by structuring evaluation around two sources of prompt variation -- prompt context and framing -- by (1) including clinically peripheral information (e.g. favorite color) and (2) changing the rhetorical prompt framing. We measure this sensitivity through the task of LLM-generated psychiatric hospitalization risk scores. METHODS We generated 50 structured synthetic patient profiles, each consisting of 15 clinically relevant baseline features (e.g. symptom severity) and up to 50 clinically peripheral (e.g., favorite color) features, where clinically peripheral is defined as features with limited established clinical relevance to hospitalization risk. Output differences based on both sources of variation were evaluated for all profiles using four LLM families across four rhetorical prompt framings (neutral, logical, human impact, and clinical judgment). RESULTS Our findings highlight the need for ongoing sensitivity assessments to ensure the reliability of LLM-based clinical decision-support systems. Adding clinically low-relevance features was associated with statistically significant changes in output stability in 8 of 16 model-prompt conditions suggesting that model sensitivity to clinically low-relevance information varied across models and prompt framings. Prompt framing also affected the trajectory of output stability, with effects varying across models. CONCLUSIONS Across model families, AI model outputs were differingly sensitive to these sources of variation, demonstrating the benefit of adding sensitivity analysis as evaluation criteria for clinically-facing AI systems, especially in complex or ambiguous domains like psychiatry.
- 28 October 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessSensitivity Analyses for Generative AI tasks in Psychiatry: LLM-Generated Hospitalization Risk Scores (Preprint)
BACKGROUND Generative artificial intelligence (AI), particularly large language models (LLMs), are increasingly studied for clinical reasoning and risk assessment tasks. However, AI outputs can be sensitive to prompt variations in the (1) contextual information provided and (2) prompt framing. The impacts of these variations remain insufficiently characterized in complex and inherently ambiguous domains such as psychiatry. OBJECTIVE We propose an approach for sensitivity analyses in downstream AI tasks by structuring evaluation around two sources of prompt variation -- prompt context and framing -- by (1) including clinically peripheral information (e.g. favorite color) and (2) changing the rhetorical prompt framing. We measure this sensitivity through the task of LLM-generated psychiatric hospitalization risk scores. METHODS We generated 50 structured synthetic patient profiles, each consisting of 15 clinically relevant baseline features (e.g. symptom severity) and up to 50 clinically peripheral (e.g., favorite color) features, where clinically peripheral is defined as features with limited established clinical relevance to hospitalization risk. Output differences based on both sources of variation were evaluated for all profiles using four LLM families across four rhetorical prompt framings (neutral, logical, human impact, and clinical judgment). RESULTS Our findings highlight the need for ongoing sensitivity assessments to ensure the reliability of LLM-based clinical decision-support systems. Adding clinically low-relevance features was associated with statistically significant changes in output stability in 8 of 16 model-prompt conditions suggesting that model sensitivity to clinically low-relevance information varied across models and prompt framings. Prompt framing also affected the trajectory of output stability, with effects varying across models. CONCLUSIONS Across model families, AI model outputs were differingly sensitive to these sources of variation, demonstrating the benefit of adding sensitivity analysis as evaluation criteria for clinically-facing AI systems, especially in complex or ambiguous domains like psychiatry.
- Open Access
- 22 August 2026
- Published by Taylor & Francis in International Review of Psychiatry
- p. 1-16
Precision medicine for mood disorders in the big data era
Advances in computation now allow large clinical datasets to be used for more precise diagnosis, ongoing monitoring, and personalized psychiatric care (Potash et al., 2025). In particular, applicat...
- 22 August 2026
- Published by Taylor & Francis in International Review of Psychiatry
- p. 1-16
Open AccessPrecision medicine for mood disorders in the big data era
Advances in computation now allow large clinical datasets to be used for more precise diagnosis, ongoing monitoring, and personalized psychiatric care (Potash et al., 2025). In particular, applicat...
- Open Access
- 20 August 2026
- Published by Wiley in AI Magazine
- Vol. 47 (3) , e70081
Reducing alert fatigue through AI ranking: A deployed public health data monitoring system
Public health experts need scalable methods to monitor large volumes of health data and identify data points that may indicate significant events, such as outbreaks or data quality issues. Experts then triage and analyze these data points so that they can prevent downstream errors in forecasting or policy. Still, traditional alert‐based data monitoring systems, used for decades in public health practice, may return tens of thousands of unhelpful alerts, making relevant events difficult to identify. We introduce a human‐in‐the‐loop AI system for public health data monitoring that uses a ranking‐based AI anomaly detection method. This system has since been deployed at a national public health organization under loads of up to 5 million data points daily. A three‐month longitudinal deployment evaluation revealed a significant improvement in system goals, including a 54x increase in data reviewer efficiency and increased engagement compared to traditional alert‐based methods.
- 20 August 2026
- Published by Wiley in AI Magazine
- Vol. 47 (3) , e70081
Open AccessReducing alert fatigue through AI ranking: A deployed public health data monitoring system
Public health experts need scalable methods to monitor large volumes of health data and identify data points that may indicate significant events, such as outbreaks or data quality issues. Experts then triage and analyze these data points so that they can prevent downstream errors in forecasting or policy. Still, traditional alert‐based data monitoring systems, used for decades in public health practice, may return tens of thousands of unhelpful alerts, making relevant events difficult to identify. We introduce a human‐in‐the‐loop AI system for public health data monitoring that uses a ranking‐based AI anomaly detection method. This system has since been deployed at a national public health organization under loads of up to 5 million data points daily. A three‐month longitudinal deployment evaluation revealed a significant improvement in system goals, including a 54x increase in data reviewer efficiency and increased engagement compared to traditional alert‐based methods.
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 13365-13369
Generative AI in Healthcare: Temporal Data Analysis and Sequential Decision-Making
Generative AI offers transformative solutions for public health and clinical practice. However, applying these models in high-stakes domains requires navigating multi-modal time series data, complex causality, and spurious correlations. This tutorial provides a comprehensive, data-to-deployment overview of generative AI in healthcare. We first examine causal analysis to ensure models are robust and grounded in true relationships rather than observational noise. We explore generative models for population health forecasting, tackling time-dependent data and uncertainty quantification. We then introduce advanced techniques—such as utilizing diffusion models for reinforcement learning and optimization—to build flexible clinical decision support systems. Finally, a real-world case study on maternal and child healthcare resource allocation provides a blueprint for responsibly deploying AI in critical care settings.
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 13365-13369
Open AccessGenerative AI in Healthcare: Temporal Data Analysis and Sequential Decision-Making
Generative AI offers transformative solutions for public health and clinical practice. However, applying these models in high-stakes domains requires navigating multi-modal time series data, complex causality, and spurious correlations. This tutorial provides a comprehensive, data-to-deployment overview of generative AI in healthcare. We first examine causal analysis to ensure models are robust and grounded in true relationships rather than observational noise. We explore generative models for population health forecasting, tackling time-dependent data and uncertainty quantification. We then introduce advanced techniques—such as utilizing diffusion models for reinforcement learning and optimization—to build flexible clinical decision support systems. Finally, a real-world case study on maternal and child healthcare resource allocation provides a blueprint for responsibly deploying AI in critical care settings.
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 13431-13432
epiDAMIK 2026: Workshop on Data-driven Decision Making for Public and Population Health
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 13431-13432
Open AccessepiDAMIK 2026: Workshop on Data-driven Decision Making for Public and Population Health
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 7511-7521
Ranking Anomalous Subsequences in Large Scale Streaming Data for Efficient Triage
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 7511-7521
Open AccessRanking Anomalous Subsequences in Large Scale Streaming Data for Efficient Triage
- Open Access
Constrained Process Maps for Multi-Agent Generative AI Workflows
Large language model (LLM)-based agents are increasingly used to perform complex, multi-step workflows in regulated settings such as compliance and due diligence. However, many agentic architectures rely primarily on prompt engineering of a single agent, making it difficult to observe or compare how models handle uncertainty and coordination across interconnected decision stages and with human oversight. We introduce a multi-agent system formalized as a finite-horizon Markov Decision Process (MDP) with a directed acyclic structure. Each agent corresponds to a specific role or decision stage (e.g., content, business, or legal review in a compliance workflow), with predefined transitions representing task escalation or completion. Epistemic uncertainty is quantified at the agent level using Monte Carlo estimation, while system-level uncertainty is captured by the MDP's termination in either an automated labeled state or a human-review state. We illustrate the approach through a case study in AI safety evaluation for self-harm detection, implemented as a multi-agent compliance system. Results demonstrate improvements over a single-agent baseline, including up to a 19\% increase in accuracy, up to an 85x reduction in required human review, and, in some configurations, reduced processing time.
Open AccessConstrained Process Maps for Multi-Agent Generative AI Workflows
Large language model (LLM)-based agents are increasingly used to perform complex, multi-step workflows in regulated settings such as compliance and due diligence. However, many agentic architectures rely primarily on prompt engineering of a single agent, making it difficult to observe or compare how models handle uncertainty and coordination across interconnected decision stages and with human oversight. We introduce a multi-agent system formalized as a finite-horizon Markov Decision Process (MDP) with a directed acyclic structure. Each agent corresponds to a specific role or decision stage (e.g., content, business, or legal review in a compliance workflow), with predefined transitions representing task escalation or completion. Epistemic uncertainty is quantified at the agent level using Monte Carlo estimation, while system-level uncertainty is captured by the MDP's termination in either an automated labeled state or a human-review state. We illustrate the approach through a case study in AI safety evaluation for self-harm detection, implemented as a multi-agent compliance system. Results demonstrate improvements over a single-agent baseline, including up to a 19\% increase in accuracy, up to an 85x reduction in required human review, and, in some configurations, reduced processing time.
- Open Access
- 24 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 6731-6732
Workshop on Discovering Drift Phenomena in Evolving Data Landscape (DELTA)
Automated systems must adapt to evolving environments, yet many struggle with drift phenomena affecting healthcare, finance, and cybersecurity domains. The DELTA workshop addresses this by distinguishing between data and concept drift, aiming to create a practical, human-centric framework for managing drift. The workshop seeks innovative drift detection, prediction, and analysis solutions by uniting researchers and practitioners. DELTA fosters collaboration to advance the understanding and management of drift in dynamic data landscapes by featuring keynotes, paper presentations, interactive sessions, and discussions.
- 24 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 6731-6732
Open AccessWorkshop on Discovering Drift Phenomena in Evolving Data Landscape (DELTA)
Automated systems must adapt to evolving environments, yet many struggle with drift phenomena affecting healthcare, finance, and cybersecurity domains. The DELTA workshop addresses this by distinguishing between data and concept drift, aiming to create a practical, human-centric framework for managing drift. The workshop seeks innovative drift detection, prediction, and analysis solutions by uniting researchers and practitioners. DELTA fosters collaboration to advance the understanding and management of drift in dynamic data landscapes by featuring keynotes, paper presentations, interactive sessions, and discussions.
- Open Access
- 2 March 2023
- Published by Association for Computing Machinery (ACM)
- p. 87-93
CS-JEDI
Computer science (CS) has historically struggled with issues related to diversity, equity, and inclusion (DEI). Based on how these issues were affecting PhD students in our department (the Carnegie Mellon University CS Department), we identified required DEI education for PhD students as a potentially high-impact approach to improving the PhD student experience in our program. Given that no existing curriculum met the desired criteria, we (PhD students)-alongside many members of the CMU community-developed and implemented CS-JEDI: Justice, Equity, Diversity, and Inclusion in Computer Science. CS-JEDI is a 6-week DEI curriculum that is now taken by all first-year PhD students in our department. This paper covers CS-JEDI's motivation and goals; describes how its evidence-based curriculum is tailored to these goals and to the CS PhD context; and gives a data-driven evaluation of the extent to which CS-JEDI's first offering, in Spring 2022, achieved these goals.
- 2 March 2023
- Published by Association for Computing Machinery (ACM)
- p. 87-93
Open AccessCS-JEDI
Computer science (CS) has historically struggled with issues related to diversity, equity, and inclusion (DEI). Based on how these issues were affecting PhD students in our department (the Carnegie Mellon University CS Department), we identified required DEI education for PhD students as a potentially high-impact approach to improving the PhD student experience in our program. Given that no existing curriculum met the desired criteria, we (PhD students)-alongside many members of the CMU community-developed and implemented CS-JEDI: Justice, Equity, Diversity, and Inclusion in Computer Science. CS-JEDI is a 6-week DEI curriculum that is now taken by all first-year PhD students in our department. This paper covers CS-JEDI's motivation and goals; describes how its evidence-based curriculum is tailored to these goals and to the CS PhD context; and gives a data-driven evaluation of the extent to which CS-JEDI's first offering, in Spring 2022, achieved these goals.