About

Kurt Luther is an Associate Professor and Associate Director for Research at the Center for Human-Computer Interaction at Virginia Tech. His research interests include crowdsourcing, social computing, and human-AI collaboration. He holds a Ph.D. in human-centered computing from Georgia Tech, earned in 2012, and a B.S. in computer graphics technology from Purdue University, obtained in 2006. Luther is based at the Institute for Advanced Computing in Alexandria, Virginia, and is involved in advancing understanding and development in the fields of social computing and human-AI interaction.

Research topics

• Computer Science
• Sociology
• Social Science
• Political Science
• Artificial Intelligence
• Public relations
• Ecology
• Engineering ethics
• Art
• Engineering
• Visual arts
• Medicine
• Law
• Data science

Selected publications

• Supplemental Materials - Leveraging Generative AI for Hazard Mitigation Planning: Insights from the Co-Design and Evaluation of Hazard Helper

  Open MIND · 2026-02-25

  datasetOpen accessSenior author

  These are the Supplemental Materials for the study "Leveraging Generative AI for Hazard Mitigation Planning: Insights from the Co-Design and Evaluation of Hazard Helper." This study is an initial, exploratory phase in a broader research program on responsible integration of Generative Artificial Intelligence (GenAI) into hazard mitigation planning. The goal is not to introduce a new AI model or evaluate algorithmic performance. Instead, the contribution is empirical, and suggestive of a replicable process (i.e., the three phases designed in the participatory co-design approach)[MS6.1][MS6.2][SC6.3][RP6.4]. We examine how emergency management professionals perceive and use a customized GenAI system within planning workflows. The research documents a co-design process, identifies institutional, ethical, and tensions in early adoption, and specifies conditions under which GenAI can support or impede planning. The findings underscore the importance of human-centered, co-designed, and governance-focused analysis of AI in the public sector.

  OA PDFDOI

• The Influence of Distributed AI in Trust and Collaboration for Search-and-Rescue Teams

  2026-04-13 · 1 citations

  articleOpen access

  Artificial intelligence (AI) is increasingly deployed in high-stakes domains such as search-and-rescue (SAR), where detections or classifications can shape how teams share information, build trust, and make time-critical decisions. This paper investigates how teams of SAR professionals incorporate AI into their teamwork, highlighting both benefits and challenges. To support this study, we developed the Council of Wizards, a multi-agent Wizard-of-Oz technique that simulates distributed AI systems, enabling scalable and controlled evaluation of collaborative dynamics. Using this novel method, we conducted an experiment with 24 subject-matter experts (SMEs) who reviewed SAR video footage as small teams and made group decisions, with or without AI support. Quantitative results showed that AI-assisted teams reached consensus faster than controls. Qualitative feedback revealed how participants interpreted trust cues, adapted strategies, and sometimes struggled with overload or conflicting detections. Findings illustrate how AI shapes teamwork in SAR and provide design implications for trustworthy distributed human-AI interactions.

  PublisherDOI

• PerceptiSync: Trustworthy Object Detection using Crowds-in-the-Loop for Cyber-Physical Systems

  ACM Transactions on Cyber-Physical Systems · 2025-07-01 · 2 citations

  article

  Establishing reliable object detection in distributed environments is challenging, particularly when trust depends on results from multiple computer vision systems. In this manuscript, we introduce PerceptiSync, a novel and trustworthy Embodied-AI (EAI) framework. It is designed for shared perception across distributed Cyber-Physical Systems (CPS) that utilize object detection. This includes applications in Connected Autonomous Vehicles, drone swarms, and CCTV camera networks. PerceptiSync is designed around a Crowds-in-the-Loop (CITL) concept to enhance system reliability by incorporating four individual user configurations and the Dirichlet-Categorical trust model. PerceptiSync undergoes a two-stage evaluation. First, it is assessed using a benchmark Computer Vision (CV) dataset to track performance over time. Second, it is tested with integrated user configurations to evaluate trust accuracy and mitigation capabilities against false positives. The results show that PerceptiSync outperforms existing AI-only trust frameworks, achieving a higher mean Kendall’s Tau coefficient of 0.228 compared to 0.051, demonstrating successful performance over time.

  PublisherDOI

• OSINT Clinic: Co-designing AI-Augmented Collaborative OSINT Investigations for Vulnerability Assessment

  2025-04-24 · 3 citations

  articleOpen accessSenior author

  Small businesses need vulnerability assessments to identify and mitigate cyber risks. Cybersecurity clinics provide a solution by offering students hands-on experience while delivering free vulnerability assessments to local organizations. To scale this model, we propose an Open Source Intelligence (OSINT) clinic where students conduct assessments using only publicly available data.We enhance the quality of investigations in the OSINT clinic by addressing the technical and collaborative challenges. Over the duration of the 2023-24 academic year, we conducted a three-phase co-design study with six students. Our study identified key challenges in the OSINT investigations and explored how generative AI could address these performance gaps. We developed design ideas for effective AI integration based on the use of AI probes and collaboration platform features. A pilot with three small businesses highlighted both the practical benefits of AI in streamlining investigations, and limitations, including privacy concerns and difficulty in monitoring progress.

  PublisherDOI

• Virtual healthcare bot (VHC-Bot): a Person-centered AI chatbot for transforming patient care and healthcare workforce dynamics

  Network Modeling Analysis in Health Informatics and Bioinformatics · 2025-06-17 · 5 citations

  articleOpen access

  Abstract This study addresses the growing role of Virtual Healthcare (VHC) in mitigating the global shortage of skilled healthcare workers and explores how Artificial Intelligence (AI) can empower clinicians by providing rapid, reliable information at the point of care. However, the proliferation of AI in healthcare poses risks of potential deskilling of clinicians’ judgment and the inability of some non-AI platforms to deliver Person-Centered (PC) care. These shortcomings may lead to unsafe self-diagnosis practices. This paper introduces VHC-Bot, an AI-driven PC VHC platform designed to strike a balance between patient autonomy, healthcare worker expertise, and AI support to deliver accurate, efficient, and personalized care. It emphasizes collaborative decision-making, effective communication, and knowledge-sharing to enhance clinical skills. The study leverages advanced AI models to design the VHC-Bot platform and integrates PC care principles. Key components include natural language processing for effective communication, diagnostic algorithms for precise symptom evaluation, and machine learning models to adapt to individual patient needs. Performance evaluation methods include clinical simulation testing, patient satisfaction surveys, and workflow efficiency analysis. Results indicate significant improvements in diagnostic accuracy, consultation times, and clinician-patient communication using the platform, which fosters collaboration among healthcare professionals, enhancing their clinical judgment and maintaining decision-making authority. Furthermore, patient satisfaction scores demonstrated marked improvement due to the personalized and accessible care provided by VHC-Bot. VHC-Bot delivers high-quality, efficient care while safeguarding human expertise in clinical judgment. This approach ensures accessible healthcare, efficient, and human-centred, setting a benchmark for future AI-integrated VHC systems.

  PublisherOA PDFDOI

• Factors influencing the consistency in crowdsourced interpretations of aerial photographs to measure tree canopy cover

  Ecological Informatics · 2025-07-15

  articleOpen access

  Machine learning models are typically data-hungry algorithms that require large data inputs for training. When they produce wall-to-wall remote sensing products, model validation also requires large sets of temporally harmonized field observations. Crowdsourcing may offer a potential solution for the collection of photointerpretations for the training and validation of spatial models of tree canopy cover (TCC), as it harnesses the power of a large anonymous crowd in the completion of repetitive discrete analyses or human intelligence tasks (HITs). This study explores the factors that determine the consistency of TCC interpretations collected by an anonymous crowd to those collected by a control group. The crowd interpretations were obtained through an anonymous platform with a task-reward framework, while those collected by the control group were collected by known interpreters in a more traditional setting. Both groups carried out this task using an interface developed for Amazon’s Mechanical Turk platform. We collected multiple interpretations at sample plot locations from both crowd and control interpreters, and sampled these data in a Monte Carlo framework to estimate a classification model predicting the consistency of each crowd interpretation with control interpretations. Using this model, we identified the most important variables in estimating the relationship between a location’s characteristics and interpretation behaviors which affect consistency in interpretations between crowd workers our control group. Overall, we show low agreement between crowdsourced and control interpretations, as well as interpretations from individual control group members. This warrants caution in considering the crowdsourced photointerpretation of TCC as a data source for model training and validation without adequate interpreter training as well as significant quality control measures and consistency standards. We show that the number of plots interpreted was the strongest indicator of the reliability of an individual’s interpretations, further evidenced by apparent fatigue effects in crowd interpretations. The second most important variable related to the use of the false color display during interpretation followed by a variable related to the use of the natural color display during interpretation, reflecting the differences in interpretation methodologies used by crowd workers and control group interpreters and the impact display has on the interpretation of tree canopy cover. Finally, we discuss recommendations for further study and future implementations of crowdsourced photointerpretation. These include the enhanced use of existing mechanisms within Mechanical Turk such as worker qualifications to identify and reward more attentive workers, as well as enhanced attention to quality control measures throughout the data collection process and measures to increase intrinsic motivation. For our study we also recommend a minimum time on task or other measures to reduce the punishment of access to HITs for workers who took their time providing detailed interpretations. We also recommend using optimized default interface settings instead of providing a variety of options to the interpreter. • We evaluated the photointerpretation of tree canopy cover (a continuous variable) based on application utilization and various worker behaviors. • Photointerpretation of tree canopy cover can result in variability among control interpreters as well as crowd workers, and so should be used with caution. • Statistical tests show that reliability of crowd photointerpretation of tree canopy cover is related most significantly to interpreter fatigue, followed by interface utilization and time-money motivations. • For studies with similar applications and methods, we recommend robust quality control measures, evaluation of fair wages, mechanisms to identify and reward attentive workers, and interface design for optimal interpreter performance.

  PublisherDOI

• Reexamining Technological Support for Genealogy Research, Collaboration, and Education

  Proceedings of the ACM on Human-Computer Interaction · 2025-05-02 · 1 citations

  preprintOpen accessSenior author

  Genealogy, the study of family history and lineage, has seen tremendous growth over the past decade, fueled by technological advances such as home DNA testing and mass digitization of historical records. However, HCI research on genealogy practices is nascent, with the most recent major studies predating this transformation. In this paper, we present a qualitative study of the current state of technological support for genealogy research, collaboration, and education. Through semi-structured interviews with 20 genealogists with diverse expertise, we report on current practices, challenges, and success stories around how genealogists conduct research, collaborate, and learn skills. We contrast the experiences of amateurs and experts, describe the emerging importance of standardization and professionalization of the field, and stress the critical role of computer systems in genealogy education. We bridge studies of sensemaking and information literacy through this empirical study on genealogy research practices, and conclude by discussing how genealogy presents a unique perspective through which to study collective sensemaking and education in online communities.

  PublisherDOI

• KHAIT: K-9 Handler Artificial Intelligence Teaming for Collaborative Sensemaking

  2025-03-19 · 3 citations

  preprintOpen access

  In urban search and rescue (USAR) operations, communication between handlers and specially trained canines is crucial but often complicated by challenging environments and the specific behaviors canines are trained to exhibit when detecting a person. Since a USAR canine often works out of sight of the handler, the handler lacks awareness of the canine's location and situation, known as the 'sensemaking gap.' In this paper, we propose KHAIT, a novel approach to close the sensemaking gap and enhance USAR effectiveness by integrating object detection-based Artificial Intelligence (AI) and Augmented Reality (AR). Equipped with AI-powered cameras, edge computing, and AR headsets, KHAIT enables precise and rapid object detection from a canine's perspective, improving survivor localization. We evaluate this approach in a real-world USAR environment, demonstrating an average survival allocation time decrease of 22%, enhancing the speed and accuracy of operations.

  PublisherDOI

• Factors Influencing the Consistency in Crowdsourced Interpretations of Aerial Photographs to Measure Tree Canopy Cover

  SSRN Electronic Journal · 2025-01-01

  preprintOpen access
  PublisherDOI

• What Lies Beneath? Exploring the Impact of Underlying AI Model Updates in AI-Infused Systems

  2025-04-24 · 2 citations

  articleOpen accessSenior author

  AI models are constantly evolving, with new versions released frequently. Human-AI interaction guidelines encourage notifying users about changes in model capabilities, ideally supported by thorough benchmarking. However, as AI systems integrate into domain-specific workflows, exhaustive benchmarking can become impractical, often resulting in silent or minimally communicated updates. This raises critical questions: Can users notice these updates? What cues do they rely on to distinguish between models? How do such changes affect their behavior and task performance? We address these questions through two studies in the context of facial recognition for historical photo identification: an online experiment examining users’ ability to detect model updates, followed by a diary study exploring perceptions in a real-world deployment. Our findings highlight challenges in noticing AI model updates, their impact on downstream user behavior and performance, and how they lead users to develop divergent folk theories. Drawing on these insights, we discuss strategies for effectively communicating model updates in AI-infused systems.

  PublisherDOI

Recent grants

• CHS: Small: Supporting Crowdsourced Sensemaking in Big Data with Dynamic Context Slices

  NSF · $532k · 2015–2019

• GraphCrowd: Using Crowdsourced Design to Visualize Effects of Environmental Chemicals on Signaling Networks

  NIH · $615k · 2016–2019

• CAREER: Transforming Investigative Science and Practice with Expert-Led Crowdsourcing

  NSF · $579k · 2017–2023

Frequent coauthors

• Daria Cybulska

  14 shared

• Pip Willcox

  National Archives

  14 shared

• Meghan Ferriter

  Library of Congress

  14 shared

• Ben Brumfield

  14 shared

• Denise Burgher

  14 shared

• Michael Haley Goldman

  14 shared

• Brendon Wilkins

  14 shared

• Austin Mast

  14 shared

Education

• Ph.D., School of Interactive Computing

  Georgia Institute of Technology

  2012

• B.S., Department of Computer Graphics Technology

  Purdue University

  2006