Thomas LaToza

I study how humans interact with code and design new ways to build software. I work at the intersection of software engineering and human-computer interaction.

I use behavioral methods to study developers as users of their programming tools, understand what makes designing, implementing, and debugging software hard, and re-envision the relationship between developers and code. A key focus of my work has been designing new techniques to view and manipulate code. I have pioneered the design of crowdsourced programming environments, which reify developer’s design knowledge, strategies, and mental models in explicit forms which can be manipulated by program analysis tools and connected to code.

News

Our paper on How omniscient debuggers impact debugging behavior will appear at VL/HCC 2025.

Our paper on The Evolution of Information Seeking in Software Development: Understanding the Role and Impact of AI Assistants will appear at HumanAISE 2025.

Our paper on OurCode: Experiences Transitioning University Research into a Developer Tools Startup will appear as a FSE 2025 Industry Paper.

I gave a keynote at ICPC 2025 on Theories of Program Comprehension in the Age of LLMs.

I joined the editorial board of Empirical Software Engineering.

Our paper on Advancing HCI with Neuromorphic Technology: Guidelines for Designing User-Friendly Developer Tools for Neuromorphic Development appeared at CHI 2025.

My students, Emad Aghayi, David Samudio, and Sahar Mehrpour, successfully defended their dissertations.

Our paper on How many pomodoros do professional engineers need to complete a microtask of programming? appeared as a FSE 2024 Industry Paper.

I was honored to be recognized as a George Mason University Teacher of Distinction.

PhD Students

Ruochen Wang

Mainul Hossain

PhD Alumni

Research Projects

Programming Strategies

Developers solve problems by applying programming strategies, describing sequences of steps to debug, reuse, design, test, and understand code. Experienced developers work better by using more effective strategies. We've been exploring ways to capture and make explicit the strategies developers use, helping uncover how developers work differently and to create new ways to help developers benefit from the hard-earned wisdom of experienced developers.
[ICSE23] [CHI22] [VL/HCC21] [EMSE20] [SIGCSE19]

Debugging Mental Models

As developers debug, they use their mental models to formulate hypotheses about the cause of a defect. Unfortunately, when their hypotheses are wrong, developers can waste time and get stuck gathering information that does not lead them any closer to the cause of the defect. We've conducted studies to understand how developers debug and the role of hypotheses in debugging and are exploring new forms of debugging tools which enable developers to find, use, and test relevant hypotheses from similar defects.
[UIST23] [EMSE23] [VL/HCC22] [PLATEAU21] [VL/HCC21] [VL/HCC20]

Microtask Programming

Onboarding on a new project is often a long, hard process, dissuading casual contributors from ever starting. What if you could contribute code to a new software project in a few minutes? How would enabling many transient contributors change software development? We've been exploring these questions through building web-based programming environments which enable microtask programming and conducting studies to understand how they change software development work.
[ASE/IS24] [EMSE23] [JSS21] [FSE-IP20] [TSE18] [SOFTWARE16] [VL/HCC15] [ICSE15] [UIST2014] [CHASE13b]

Working with Design Decisions in Code

Good documentation has long been argued to be the key to helping developers work more quickly and consistently with design decisions. But our studies have found that documentation is left largely disconnected from code, making it hard to write and update and causing it to become out of date and untrusted. This leaves developers to instead reverse engineer design decisions from code, causing rationale questions about design decisions to be some of the most challenging to answer. We've conducted studies to explore the nature of this problem and have invented new techniques to make documentation active and bidirectionally synchronized with code.
[FSE-IP25] [CSUR23] [SOFTWARE23] [EMSE23b] [PLATEAU21] [FSE20] [VL/HCC19a] [VL/HCC19b] [VL/HCC15] [PLATEAU10] [FSE07] [ICSE06]

Information Needs in Software Development

What does this do? Will this work? Who uses this? Asking and answering questions is central to programming, software design, and coordination in programming teams. Our work has examined the most frequent questions developers ask and the challenges that they face.
[CHI25] [VL/HCC22] [SOFTWARE20] [TSE15] [CHASE13a] [PLATEAU10] [CHASE10b] [FSE07] [ICSE06]

Methods for Developer Experience Design

How can we understand what software developers do? Do developer tools really help developers? In our work, we have developed methods for using field observations, lab studies, interviews, surveys, and controlled experiments to learn how developers interact with their development environments and programming languages.
[FSE-IP25] [TOSEM23] [SOFTWARE20] [CHASE19] [VL/HCC19] [COMPUTER16] [EMSE15] [PLATEAU11] [CHASE10a]

Answering Reachability Questions

What can make debugging in large and complex codebases so hard and time-consuming? We found that developers ask reachability questions, searching across control flow for statements matching search criteria. For example, developers debugging a deadlock may reconstruct control flow describing where resources are acquired and released. We designed Reacher, which uses a static analysis to automate control flow searches, enabling developers to query for statements on paths and interactively construct an interprocedural control flow visualization.
[VL/HCC11][ICSE10]

Resources

What do we know about human aspects of software development? Bibliography

How has software engineering research evolved over the past 4 decades? Tag Clouds

Evaluating programming languages and tools in studies with human participants Paper, Slides from tutorial