arXiv subjects

Venkatesh Sivaraman

Publications and source records attributed to Venkatesh Sivaraman.

At least 19 recordsLinked to original sources

arXiv2026

"I Know Where to Look," But Does the LLM? Charting the Gaps Between Clinical Expert Needs and Unstructured Data Abstraction Tools

Clinical data abstraction, the process of distilling structured information from patient records, plays a key role in advancing knowledge about diseases such as cancer. Information extraction (IE) with large language models (LLMs) could accelerate this process, but it is unclear whether current frameworks effectively support clinical researchers without AI expertise. To address this, we co-designed an interactive LLM-based abstraction system called Libretto with seven cancer research teams, then evaluated the system's ability to help them answer real-world research questions. We found that while clinicians knew where and how to annotate complex concepts in patient notes, in twelve of fourteen tasks they faced barriers to replicating those intuitions with LLMs. Contextual note reliability judgments, difficulties in steering vibe-coded prompts, and inflexible evaluation strategies necessitated fundamental changes to the IE workflow. Our results highlight open problems for HCI research to bridge the gaps between AI data work tools and clinical users' needs.

cs.HC↗

arXiv2026

Expert Evaluation of Clinical AI Tools on Real Point-of-Care Clinical Queries

Physicians now pose millions of clinical questions to AI tools each week, yet these tools are evaluated largely on hypothetical or exam-style questions, not those actually asked in practice. We report a blinded evaluation built on 620 Real-world Point-Of-Care Queries (Real-POCQi) submitted to the OpenEvidence (OE) platform by physicians spanning 30 specialties, as well as 187 questions from HealthBench. 149 practicing physicians across 36 states made head-to-head comparisons between answers from three frontier general-purpose models (Claude Opus 4.8, Gemini 3.1 Pro, and GPT-5.5) and a specialized clinical tool (OE), with graders matched to each question's specialty. When comparing answers along five dimensions relevant to clinical decision support -- accuracy, clinical utility, source quality, verifiability, & completeness -- physicians scored the specialized tool highest on all axes; in the primary analysis on Real-POCQi, win differences (margins between win and loss rates) ranged from 25 to 39 percentage points (p<0.001). Results remained consistent in sensitivity analyses stratifying by citation display, answer length, OE-user status, and Real-POCQi versus HealthBench. In parallel, LLM judges were found to systematically differ from expert judges, though both generally agreed on the best model. These findings underscore two conclusions: (i) AI tool evaluations should reflect real-world query distributions and use expert judges that mirror the specialization defining modern medicine and (ii) the consistent advantage of the specialized tool over general-purpose models does not necessarily mean that the latter cannot serve similar purposes, but that targeted engineering and customization can yield meaningful gains in performance for its users. We release Real-POCQi as a public benchmark, as well as the prespecified statistical analysis for reproducing results of this study.

cs.AI↗

arXiv2026

Adaptive auditing of AI systems with anytime-valid guarantees

A major bottleneck in characterizing the failure modes of generative AI systems is the cost and time of annotation and evaluation. Consequently, adaptive testing paradigms have gained popularity, where one opportunistically decides which cases and how many to annotate based on past results. While this framework is highly practical, its extreme flexibility makes it difficult to draw statistically rigorous conclusions, as it violates classical assumptions: the number of observations is typically limited (often 10 to 50 cases) and decisions regarding sampling and stopping are made in the midst of data collection rather than based a pre-specified rule. To characterize what statistical inferences can be drawn from highly adaptive audits, we introduce a hypothesis testing framework from two 'dueling' perspectives: (i) the model's null that asserts there is no failure mode with performance below a target threshold versus (ii) the auditor's null that asserts they have a sampling strategy that will uncover a failure mode. Leveraging Safe Anytime-Valid Inference (SAVI), we formalize the auditor as conducting 'testing by betting', which translates into simultaneous e-processes for testing the dueling null hypotheses. Furthermore, if the auditor is sufficiently powerful, we prove that these two hypotheses are asymptotically inverses of each other, in that passage of a stringent audit does in fact certify the AI system as being globally robust. Empirically, we demonstrate that our proposed testing procedures maintain anytime-valid type-I error control, outperform pre-specified testing methods, and can reach statistically rigorous conclusions sometimes with as few as 20 observations.

cs.AI↗

arXiv2026

More Than "Means to an End": Supporting Reasoning with Transparently Designed AI Data Science Processes

Generative artificial intelligence (AI) tools can now help people perform complex data science tasks regardless of their expertise. While these tools have great potential to help more people work with data, their end-to-end approach does not support users in evaluating alternative approaches and reformulating problems, both critical to solving open-ended tasks in high-stakes domains. In this paper, we reflect on two AI data science systems designed for the medical setting and how they function as tools for thought. We find that success in these systems was driven by constructing AI workflows around intentionally-designed intermediate artifacts, such as readable query languages, concept definitions, or input-output examples. Despite opaqueness in other parts of the AI process, these intermediates helped users reason about important analytical choices, refine their initial questions, and contribute their unique knowledge. We invite the HCI community to consider when and how intermediate artifacts should be designed to promote effective data science thinking.

cs.HC↗

arXiv2026

Intelligent Reasoning Cues: A Framework and Case Study of the Roles of AI Information in Complex Decisions

Artificial intelligence (AI)-based decision support systems can be highly accurate yet still fail to support users or improve decisions. Existing theories of AI-assisted decision-making focus on calibrating reliance on AI advice, leaving it unclear how different system designs might influence the reasoning processes underneath. We address this gap by reconsidering AI interfaces as collections of intelligent reasoning cues: discrete pieces of AI information that can individually influence decision-making. We then explore the roles of eight types of reasoning cues in a high-stakes clinical decision (treating patients with sepsis in intensive care). Through contextual inquiries with six teams and a think-aloud study with 25 physicians, we find that reasoning cues have distinct patterns of influence that can directly inform design. Our results also suggest that reasoning cues should prioritize tasks with high variability and discretion, adapt to ensure compatibility with evolving decision needs, and provide complementary, rigorous insights on complex cases.

cs.HC↗

arXiv2025

TempoQL: A Readable, Precise, and Portable Query System for Electronic Health Record Data

Electronic health record (EHR) data is an essential data source for machine learning for health, but researchers and clinicians face steep barriers in extracting and validating EHR data for modeling. Existing tools incur trade-offs between expressivity and usability and are typically specialized to a single data standard, making it difficult to write temporal queries that are ready for modern model-building pipelines and adaptable to new datasets. This paper introduces TempoQL, a Python-based toolkit designed to lower these barriers. TempoQL provides a simple, human-readable language for temporal queries; support for multiple EHR data standards, including OMOP, MEDS, and others; and an interactive notebook-based query interface with optional large language model (LLM) authoring assistance. Through a performance evaluation and two use cases on different datasets, we demonstrate that TempoQL simplifies the creation of cohorts for machine learning while maintaining precision, speed, and reproducibility.

cs.HC↗

arXiv2025

Over-Relying on Reliance: Towards Realistic Evaluations of AI-Based Clinical Decision Support

As AI-based clinical decision support (AI-CDS) is introduced in more and more aspects of healthcare services, HCI research plays an increasingly important role in designing for complementarity between AI and clinicians. However, current evaluations of AI-CDS often fail to capture when AI is and is not useful to clinicians. This position paper reflects on our work and influential AI-CDS literature to advocate for moving beyond evaluation metrics like Trust, Reliance, Acceptance, and Performance on the AI's task (what we term the "trap" of human-AI collaboration). Although these metrics can be meaningful in some simple scenarios, we argue that optimizing for them ignores important ways that AI falls short of clinical benefit, as well as ways that clinicians successfully use AI. As the fields of HCI and AI in healthcare develop new ways to design and evaluate CDS tools, we call on the community to prioritize ecologically valid, domain-appropriate study setups that measure the emergent forms of value that AI can bring to healthcare professionals.

cs.HC↗

arXiv2025

Tempo: Helping Data Scientists and Domain Experts Collaboratively Specify Predictive Modeling Tasks

Temporal predictive models have the potential to improve decisions in health care, public services, and other domains, yet they often fail to effectively support decision-makers. Prior literature shows that many misalignments between model behavior and decision-makers' expectations stem from issues of model specification, namely how, when, and for whom predictions are made. However, model specifications for predictive tasks are highly technical and difficult for non-data-scientist stakeholders to interpret and critique. To address this challenge we developed Tempo, an interactive system that helps data scientists and domain experts collaboratively iterate on model specifications. Using Tempo's simple yet precise temporal query language, data scientists can quickly prototype specifications with greater transparency about pre-processing choices. Moreover, domain experts can assess performance within data subgroups to validate that models behave as expected. Through three case studies, we demonstrate how Tempo helps multidisciplinary teams quickly prune infeasible specifications and identify more promising directions to explore.

cs.HC↗

arXiv2025

Divisi: Interactive Search and Visualization for Scalable Exploratory Subgroup Analysis

Analyzing data subgroups is a common data science task to build intuition about a dataset and identify areas to improve model performance. However, subgroup analysis is prohibitively difficult in datasets with many features, and existing tools limit unexpected discoveries by relying on user-defined or static subgroups. We propose exploratory subgroup analysis as a set of tasks in which practitioners discover, evaluate, and curate interesting subgroups to build understanding about datasets and models. To support these tasks we introduce Divisi, an interactive notebook-based tool underpinned by a fast approximate subgroup discovery algorithm. Divisi's interface allows data scientists to interactively re-rank and refine subgroups and to visualize their overlap and coverage in the novel Subgroup Map. Through a think-aloud study with 13 practitioners, we find that Divisi can help uncover surprising patterns in data features and their interactions, and that it encourages more thorough exploration of subtypes in complex data.

cs.HC↗

arXiv2025

Static Algorithm, Evolving Epidemic: Understanding the Potential of Human-AI Risk Assessment to Support Regional Overdose Prevention

Drug overdose deaths, including those due to prescription opioids, represent a critical public health issue in the United States and worldwide. Artificial intelligence (AI) approaches have been developed and deployed to help prescribers assess a patient's risk for overdose-related death, but it is unknown whether public health experts can leverage similar predictions to make local resource allocation decisions more effectively. In this work, we evaluated how AI-based overdose risk assessment could be used to inform local public health decisions using a working prototype system. Experts from three health departments, of varying locations and sizes with respect to staff and population served, were receptive to the potential benefits of algorithmic risk prediction and of using AI-augmented visualization to connect across data sources. However, they also expressed concerns about whether the risk prediction model's formulation and underlying data would match the state of the overdose epidemic as it evolved in their specific locations. Our findings extend those of other studies on algorithmic systems in the public sector, and they present opportunities for future human-AI collaborative tools to support decision-making in local, time-varying contexts.

cs.HC↗

arXiv2024

Counterpoint: Orchestrating Large-Scale Custom Animated Visualizations

Custom animated visualizations of large, complex datasets are helpful across many domains, but they are hard to develop. Much of the difficulty arises from maintaining visualization state across many animated graphical elements that may change in number over time. We contribute Counterpoint, a framework for state management designed to help implement such visualizations in JavaScript. Using Counterpoint, developers can manipulate large collections of marks with reactive attributes that are easy to render in scalable APIs such as Canvas and WebGL. Counterpoint also helps orchestrate the entry and exit of graphical elements using the concept of a rendering "stage." Through a performance evaluation, we show that Counterpoint adds minimal overhead over current high-performance rendering techniques while simplifying implementation. We provide two examples of visualizations created using Counterpoint that illustrate its flexibility and compatibility with other visualization toolkits as well as considerations for users with disabilities. Counterpoint is open-source and available at https://github.com/cmudig/counterpoint.

cs.GR↗

arXiv2024

Compress and Compare: Interactively Evaluating Efficiency and Behavior Across ML Model Compression Experiments

To deploy machine learning models on-device, practitioners use compression algorithms to shrink and speed up models while maintaining their high-quality output. A critical aspect of compression in practice is model comparison, including tracking many compression experiments, identifying subtle changes in model behavior, and negotiating complex accuracy-efficiency trade-offs. However, existing compression tools poorly support comparison, leading to tedious and, sometimes, incomplete analyses spread across disjoint tools. To support real-world comparative workflows, we develop an interactive visual system called Compress and Compare. Within a single interface, Compress and Compare surfaces promising compression strategies by visualizing provenance relationships between compressed models and reveals compression-induced behavior changes by comparing models' predictions, weights, and activations. We demonstrate how Compress and Compare supports common compression analysis tasks through two case studies, debugging failed compression on generative language models and identifying compression artifacts in image classification models. We further evaluate Compress and Compare in a user study with eight compression experts, illustrating its potential to provide structure to compression workflows, help practitioners build intuition about compression, and encourage thorough analysis of compression's effect on model behavior. Through these evaluations, we identify compression-specific challenges that future visual analytics tools should consider and Compress and Compare visualizations that may generalize to broader model comparison tasks.

cs.HC↗

arXiv2024

How Consistent are Clinicians? Evaluating the Predictability of Sepsis Disease Progression with Dynamics Models

Reinforcement learning (RL) is a promising approach to generate treatment policies for sepsis patients in intensive care. While retrospective evaluation metrics show decreased mortality when these policies are followed, studies with clinicians suggest their recommendations are often spurious. We propose that these shortcomings may be due to lack of diversity in observed actions and outcomes in the training data, and we construct experiments to investigate the feasibility of predicting sepsis disease severity changes due to clinician actions. Preliminary results suggest incorporating action information does not significantly improve model performance, indicating that clinician actions may not be sufficiently variable to yield measurable effects on disease progression. We discuss the implications of these findings for optimizing sepsis treatment.

cs.LG↗

arXiv2024

Sketching AI Concepts with Capabilities and Examples: AI Innovation in the Intensive Care Unit

Advances in artificial intelligence (AI) have enabled unprecedented capabilities, yet innovation teams struggle when envisioning AI concepts. Data science teams think of innovations users do not want, while domain experts think of innovations that cannot be built. A lack of effective ideation seems to be a breakdown point. How might multidisciplinary teams identify buildable and desirable use cases? This paper presents a first hand account of ideating AI concepts to improve critical care medicine. As a team of data scientists, clinicians, and HCI researchers, we conducted a series of design workshops to explore more effective approaches to AI concept ideation and problem formulation. We detail our process, the challenges we encountered, and practices and artifacts that proved effective. We discuss the research implications for improved collaboration and stakeholder engagement, and discuss the role HCI might play in reducing the high failure rate experienced in AI innovation.

cs.HC↗

arXiv2023

Ignore, Trust, or Negotiate: Understanding Clinician Acceptance of AI-Based Treatment Recommendations in Health Care

Artificial intelligence (AI) in healthcare has the potential to improve patient outcomes, but clinician acceptance remains a critical barrier. We developed a novel decision support interface that provides interpretable treatment recommendations for sepsis, a life-threatening condition in which decisional uncertainty is common, treatment practices vary widely, and poor outcomes can occur even with optimal decisions. This system formed the basis of a mixed-methods study in which 24 intensive care clinicians made AI-assisted decisions on real patient cases. We found that explanations generally increased confidence in the AI, but concordance with specific recommendations varied beyond the binary acceptance or rejection described in prior work. Although clinicians sometimes ignored or trusted the AI, they also often prioritized aspects of the recommendations to follow, reject, or delay in a process we term "negotiation." These results reveal novel barriers to adoption of treatment-focused AI tools and suggest ways to better support differing clinician perspectives.

cs.HC↗

arXiv2022

Extended Analysis of "How Child Welfare Workers Reduce Racial Disparities in Algorithmic Decisions"

This is an extended analysis of our paper "How Child Welfare Workers Reduce Racial Disparities in Algorithmic Decisions," which looks at racial disparities in the Allegheny Family Screening Tool, an algorithm used to help child welfare workers decide which families the Allegheny County child welfare agency (CYF) should investigate. On April 27, 2022, Allegheny County CYF sent us an updated dataset and pre-processing steps. In this extended analysis of our paper, we show the results from re-running all quantitative analyses in our paper with this new data and pre-processing. We find that our main findings in our paper were robust to changes in data and pre-processing. Particularly, the Allegheny Family Screening Tool on its own would have made more racially disparate decisions than workers, and workers used the tool to decrease those algorithmic disparities. Some minor results changed, including a slight increase in the screen-in rate from before to after the implementation of the AFST reported our paper.

cs.HC↗

arXiv2022

LSTM-RASA Based Agri Farm Assistant for Farmers

The application of Deep Learning and Natural Language based ChatBots are growing rapidly in recent years. They are used in many fields like customer support, reservation system and as personal assistant. The Enterprises are using such ChatBots to serve their customers in a better and efficient manner. Even after such technological advancement, the expert advice does not reach the farmers on timely manner. The farmers are still largely dependent on their peers knowledge in solving the problems they face in their field. These technologies have not been effectively used to give the required information to farmers on timely manner. This project aims to implement a closed domain ChatBot for the field of Agriculture Farmers Assistant. Farmers can have conversation with the Chatbot and get the expert advice in their field. Farmers Assistant is based on RASA Open Source Framework. The Chatbot identifies the intent and entity from user utterances and retrieve the remedy from the database and share it with the user. We tested the Bot with existing data and it showed promising results.

cs.CL↗

arXiv2022

Improving Human-AI Partnerships in Child Welfare: Understanding Worker Practices, Challenges, and Desires for Algorithmic Decision Support

AI-based decision support tools (ADS) are increasingly used to augment human decision-making in high-stakes, social contexts. As public sector agencies begin to adopt ADS, it is critical that we understand workers' experiences with these systems in practice. In this paper, we present findings from a series of interviews and contextual inquiries at a child welfare agency, to understand how they currently make AI-assisted child maltreatment screening decisions. Overall, we observe how workers' reliance upon the ADS is guided by (1) their knowledge of rich, contextual information beyond what the AI model captures, (2) their beliefs about the ADS's capabilities and limitations relative to their own, (3) organizational pressures and incentives around the use of the ADS, and (4) awareness of misalignments between algorithmic predictions and their own decision-making objectives. Drawing upon these findings, we discuss design implications towards supporting more effective human-AI decision-making.

cs.HC↗