arXiv subjects

Brittany Johnson

Publications and source records attributed to Brittany Johnson.

At least 19 recordsLinked to original sources

arXiv2026

What Makes Software Issue Resolution Tasks Difficult for Agents?

Background. Advances in agentic systems are simultaneously, and rapidly, saturating benchmarks. Despite this often discussed phenomena, benchmark scores remain difficult to interpret due to the lack of control and characterization of task difficulty. More specifically, we currently have little understanding of what makes one task harder than another, and to what extent task difficulty is predictable from static task properties. Aims. We propose a measurement framework to investigate and systematically quantify what structural properties of software tasks correspond to agent success rates for issue resolution tasks. Method. We conducted a large scale empirical study on CoderForge-Preview, the largest open dataset of coding agent trajectories to date, by extracting features across task patch, repository and prompt. We evaluated the predictive power of each feature against task outcomes using ensemble methods, SHAP attribution, and effect size analysis. Results We found that task difficulty is substantially predictable from static features (AU C = 0.863) and is largely driven by patch fragmentation and repository scale. Prompt linguistic features become visible among top contributors for tasks in the mid-band, revealing a layered structure of difficulty. Conclusion. The difficulty of an issue resolution task is encoded in its structure. This enables static, pre-hoc difficulty estimation and lays the groundwork for difficulty-controlled benchmark construction for evaluation of agents.

cs.SE↗

arXiv2025

Towards A Cultural Intelligence and Values Inferences Quality Benchmark for Community Values and Common Knowledge

Large language models (LLMs) have emerged as a powerful technology, and thus, we have seen widespread adoption and use on software engineering teams. Most often, LLMs are designed as "general purpose" technologies meant to represent the general population. Unfortunately, this often means alignment with predominantly Western Caucasian narratives and misalignment with other cultures and populations that engage in collaborative innovation. In response to this misalignment, there have been recent efforts centered on the development of "culturally-informed" LLMs, such as ChatBlackGPT, that are capable of better aligning with historically marginalized experiences and perspectives. Despite this progress, there has been little effort aimed at supporting our ability to develop and evaluate culturally-informed LLMs. A recent effort proposed an approach for developing a national alignment benchmark that emphasizes alignment with national social values and common knowledge. However, given the range of cultural identities present in the United States (U.S.), a national alignment benchmark is an ineffective goal for broader representation. To help fill this gap in this US context, we propose a replication study that translates the process used to develop KorNAT, a Korean National LLM alignment benchmark, to develop CIVIQ, a Cultural Intelligence and Values Inference Quality benchmark centered on alignment with community social values and common knowledge. Our work provides a critical foundation for research and development aimed at cultural alignment of AI technologies in practice.

cs.SE↗

arXiv2025

Why Do We Code? A Theory on Motivations and Challenges in Software Engineering from Education to Practice

Motivations and challenges jointly shape how individuals enter, persist, and evolve within software engineering (SE), yet their interplay remains underexplored across the transition from education to professional practice. We conducted 15 semi-structured interviews and employed the Gioia Methodology, an adapted grounded theory methodology from organizational behavior, to inductively derive taxonomies of motivations and challenges, and build the Exposure-Pursuit-Evaluation (EPE) Process Model. Our findings reveal that impactful early exposure triggers intrinsic motivations, while non-impactful exposure requires an extrinsic push (e.g., career/ personal goals, external validation). We identify curiosity and avoiding alternatives as a distinct educational drivers, and barriers to belonging as the only challenge persisting across education and career. Our findings show that career progression challenges (e.g., navigating the corporate world) constrain extrinsic fulfillment while technical training challenges, barriers to belonging and threats to motivation constrain intrinsic fulfillment. The theory shows how unmet motivations and recurring challenges influence persistence, career shifts, or departure from the field. Our results provide a grounded model for designing interventions that strengthen intrinsic fulfillment and reduce systemic barriers in SE education and practice.

cs.SE↗

arXiv2025

Explaining Code Risk in OSS: Towards LLM-Generated Fault Prediction Interpretations

Open Source Software (OSS) has become a very important and crucial infrastructure worldwide because of the value it provides. OSS typically depends on contributions from developers across diverse backgrounds and levels of experience. Making safe changes, such as fixing a bug or implementing a new feature, can be challenging, especially in object-oriented systems where components are interdependent. Static analysis and defect-prediction tools produce metrics (e.g., complexity,coupling) that flag potentially fault-prone components, but these signals are often hard for contributors new or unfamiliar with the codebase to interpret. Large Language Models (LLMs) have shown strong performance on software engineering tasks such as code summarization and documentation generation. Building on this progress, we investigate whether LLMs can translate fault-prediction metrics into clear, human-readable risk explanations and actionable guidance to help OSS contributors plan and review code modifications. We outline explanation types that an LLM-generated assistant could provide (descriptive, contextual, and actionable explanations). We also outline our next steps to assess usefulness through a task-based study with OSS contributors, comparing metric-only baselines to LLM-generated explanations on decision quality, time-to-completion, and error rates

cs.SE↗

arXiv2025

From Commits to Confidence: Towards Stability-Informed Risk Assessment in Open Source Software

Open source software (OSS) generates trillions of dollars in economic value and has become essential to the technical infrastructures that power organizations worldwide. As these systems increasingly depend on OSS, understanding the evolution of these projects is critical. While existing metrics provide insights into project health, one dimension remains understudied: project resilience, or the ability to return to normal operations after disturbances such as contributor departures,security vulnerabilities and bug report spikes. We hypothesize that stable commit patterns may serve as an indicator of underlying project characteristics such as mature governance, sustained contributors, and robust development processes, factors that existing research associates with resilience. Our findings reveal that only 2% of repositories exhibit daily stability, 29% achieve weekly stability, and 50\% demonstrate monthly stability, while the remaining half are unstable across all levels of granularity. Analysis of the 50 unstable repositories indicate that 86% of activity is concentrated among a few maintainers, with the top 3 contributors accounting for over 50% of commits in the past 5 years. In contrast, the 50 stable repositories distribute work more evenly, with the top 3 contributors representing less than 50% of commits. Our insights thus far indicate the fragile and multi-dimensional nature of OSS project stability, suggesting a need to go beyond commits to understand how our understanding of stability can be enriched with other considerations such as community engagement metrics and issue or pull request churn. Though our efforts only identified two repositories that achieved stability at all three temporal commit granularities, further investigation into their processes and policies can provide insights and foundations for stability-informed risk assessment in practice.

cs.SE↗

arXiv2025

Towards Bridging Language Gaps in OSS with LLM-Driven Documentation Translation

While open source communities attract diverse contributors across the globe, only a few open source software repositories provide essential documentation, such as ReadMe or CONTRIBUTING files, in languages other than English. Recently, large language models (LLMs) have demonstrated remarkable capabilities in a variety of software engineering tasks. We have also seen advances in the use of LLMs for translations in other domains and contexts. Despite this progress, little is known regarding the capabilities of LLMs in translating open-source technical documentation, which is often a mixture of natural language, code, URLs, and markdown formatting. To better understand the need and potential for LLMs to support translation of technical documentation in open source, we conducted an empirical evaluation of translation activity and translation capabilities of two powerful large language models (OpenAI ChatGPT 4 and Anthropic Claude). We found that translation activity is often community-driven and most frequent in larger repositories. A comparison of LLM performance as translators and evaluators of technical documentation suggests LLMs can provide accurate semantic translations but may struggle preserving structure and technical content. These findings highlight both the promise and the challenges of LLM-assisted documentation internationalization and provide a foundation towards automated LLM-driven support for creating and maintaining open source documentation.

cs.SE↗

arXiv2025

An Empirical Validation of Open Source Repository Stability Metrics

Over the past few decades, open source software has been continuously integrated into software supply chains worldwide, drastically increasing reliance and dependence. Because of the role this software plays, it is important to understand ways to measure and promote its stability and potential for sustainability. Recent work proposed the use of control theory to understand repository stability and evaluate repositories' ability to return to equilibrium after a disturbance such as the introduction of a new feature request, a spike in bug reports, or even the influx or departure of contributors. This approach leverages commit frequency patterns, issue resolution rate, pull request merge rate, and community activity engagement to provide a Composite Stability Index (CSI). While this framework has theoretical foundations, there is no empirical validation of the CSI in practice. In this paper, we present the first empirical validation of the proposed CSI by experimenting with 100 highly ranked GitHub repositories. Our results suggest that (1) sampling weekly commit frequency pattern instead of daily is a more feasible measure of commit frequency stability across repositories and (2) improved statistical inferences (swapping mean with median), particularly with ascertaining resolution and review times in issues and pull request, improves the overall issue and pull request stability index. Drawing on our empirical dataset, we also derive data-driven half-width parameters that better align stability scores with real project behavior. These findings both confirm the viability of a control-theoretic lens on open-source health and provide concrete, evidence-backed applications for real-world project monitoring tools.

cs.SE↗

arXiv2025

Exploring Fairness Interventions in Open Source Projects

The deployment of biased machine learning (ML) models has resulted in adverse effects in crucial sectors such as criminal justice and healthcare. To address these challenges, a diverse range of machine learning fairness interventions have been developed, aiming to mitigate bias and promote the creation of more equitable models. Despite the growing availability of these interventions, their adoption in real-world applications remains limited, with many practitioners unaware of their existence. To address this gap, we systematically identified and compiled a dataset of 62 open source fairness interventions and identified active ones. We conducted an in-depth analysis of their specifications and features to uncover considerations that may drive practitioner preference and to identify the software interventions actively maintained in the open source ecosystem. Our findings indicate that 32% of these interventions have been actively maintained within the past year, and 50% of them offer both bias detection and mitigation capabilities, mostly during inprocessing.

cs.SE↗

arXiv2025

An Investigation into Maintenance Support for Neural Networks

As the potential for neural networks to augment our daily lives grows, ensuring their quality through effective testing, debugging, and maintenance is essential. This is especially the case as we acknowledge the prospects of negative impacts from these technologies. Traditional software engineering methods, such as testing and debugging, have proven effective in maintaining software quality; however, they reveal significant research and practice gaps in maintaining neural networks. In particular, there is a limited understanding of how practitioners currently address challenges related to understanding and mitigating undesirable behaviors in neural networks. In our ongoing research, we explore the current state of research and practice in maintaining neural networks by curating insights from practitioners through a preliminary study involving interviews and supporting survey responses. Our findings thus far indicate that existing tools primarily concentrate on building and training models. While these tools can be beneficial, they often fall short of supporting practitioners' understanding and addressing the underlying causes of unexpected model behavior. By evaluating current procedures and identifying the limitations of traditional methodologies, our study aims to offer a developer-centric perspective on where current practices fall short and highlight opportunities for improving maintenance support in neural networks.

cs.SE↗

arXiv2025

What Makes a Fairness Tool Project Sustainable in Open Source?

As society becomes increasingly reliant on artificial intelligence, the need to mitigate risk and harm is paramount. In response, researchers and practitioners have developed tools to detect and reduce undesired bias, commonly referred to as fairness tools. Many of these tools are publicly available for free use and adaptation. While the growing availability of such tools is promising, little is known about the broader landscape beyond well-known examples like AI Fairness 360 and Fairlearn. Because fairness is an ongoing concern, these tools must be built for long-term sustainability. Using an existing set of fairness tools as a reference, we systematically searched GitHub and identified 50 related projects. We then analyzed various aspects of their repositories to assess community engagement and the extent of ongoing maintenance. Our findings show diverse forms of engagement with these tools, suggesting strong support for open-source development. However, we also found significant variation in how well these tools are maintained. Notably, 53 percent of fairness projects become inactive within the first three years. By examining sustainability in fairness tooling, we aim to promote more stability and growth in this critical area.

cs.SE↗

arXiv2025

A Preliminary Framework for Intersectionality in ML Pipelines

Machine learning (ML) has become a go-to solution for improving how we use, experience, and interact with technology (and the world around us). Unfortunately, studies have repeatedly shown that machine learning technologies may not provide adequate support for societal identities and experiences. Intersectionality is a sociological framework that provides a mechanism for explicitly considering complex social identities, focusing on social justice and power. While the framework of intersectionality can support the development of technologies that acknowledge and support all members of society, it has been adopted and adapted in ways that are not always true to its foundations, thereby weakening its potential for impact. To support the appropriate adoption and use of intersectionality for more equitable technological outcomes, we amplify the foundational intersectionality scholarship--Crenshaw, Combahee, and Collins (three C's), to create a socially relevant preliminary framework in developing machine-learning solutions. We use this framework to evaluate and report on the (mis)alignments of intersectionality application in machine learning literature.

cs.LG↗

arXiv2025

The Evolution of Information Seeking in Software Development: Understanding the Role and Impact of AI Assistants

About 32% of a software practitioners' day involves seeking and using information to support task completion. Although the information needs of software practitioners have been studied extensively, the impact of AI-assisted tools on their needs and information-seeking behaviors remains largely unexplored. To addresses this gap, we conducted a mixed-method study to understand AI-assisted information seeking behavior of practitioners and its impact on their perceived productivity and skill development. We found that developers are increasingly using AI tools to support their information seeking, citing increased efficiency as a key benefit. Our findings also amplify caveats that come with effectively using AI tools for information seeking, especially for learning and skill development, such as the importance of foundational developer knowledge that can guide and inform the information provided by AI tools. Our efforts have implications for the effective integration of AI tools into developer workflows as information retrieval systems and learning aids.

cs.SE↗

arXiv2025

Causality-Driven Neural Network Repair: Challenges and Opportunities

Deep Neural Networks (DNNs) often rely on statistical correlations rather than causal reasoning, limiting their robustness and interpretability. While testing methods can identify failures, effective debugging and repair remain challenging. This paper explores causal inference as an approach primarily for DNN repair, leveraging causal debugging, counterfactual analysis, and structural causal models (SCMs) to identify and correct failures. We discuss in what ways these techniques support fairness, adversarial robustness, and backdoor mitigation by providing targeted interventions. Finally, we discuss key challenges, including scalability, generalization, and computational efficiency, and outline future directions for integrating causality-driven interventions to enhance DNN reliability.

cs.LG↗

arXiv2025

Exploring Culturally Informed AI Assistants: A Comparative Study of ChatBlackGPT and ChatGPT

In recent years, we have seen an influx in reliance on AI assistants for information seeking. Given this widespread use and the known challenges AI poses for Black users, recent efforts have emerged to identify key considerations needed to provide meaningful support. One notable effort is the development of ChatBlackGPT, a culturally informed AI assistant designed to provide culturally relevant responses. Despite the existence of ChatBlackGPT, there is no research on when and how Black communities might engage with culturally informed AI assistants and the distinctions between engagement with general purpose tools like ChatGPT. To fill this gap, we propose a research agenda grounded in results from a preliminary comparative analysis of outputs provided by ChatGPT and ChatBlackGPT for travel-related inquiries. Our efforts thus far emphasize the need to consider Black communities' values, perceptions, and experiences when designing AI assistants that acknowledge the Black lived experience.

cs.HC↗

arXiv2025

An Evaluation of LLMs for Detecting Harmful Computing Terms

Detecting harmful and non-inclusive terminology in technical contexts is critical for fostering inclusive environments in computing. This study explores the impact of model architecture on harmful language detection by evaluating a curated database of technical terms, each paired with specific use cases. We tested a range of encoder, decoder, and encoder-decoder language models, including BERT-base-uncased, RoBERTa large-mnli, Gemini Flash 1.5 and 2.0, GPT-4, Claude AI Sonnet 3.5, T5-large, and BART-large-mnli. Each model was presented with a standardized prompt to identify harmful and non-inclusive language across 64 terms. Results reveal that decoder models, particularly Gemini Flash 2.0 and Claude AI, excel in nuanced contextual analysis, while encoder models like BERT exhibit strong pattern recognition but struggle with classification certainty. We discuss the implications of these findings for improving automated detection tools and highlight model-specific strengths and limitations in fostering inclusive communication in technical domains.

cs.CL↗

arXiv2025

An Investigation of Experiences Engaging the Margins in Data-Centric Innovation

Data-centric technologies provide exciting opportunities, but recent research has shown how lack of representation in datasets, often as a result of systemic inequities and socioeconomic disparities, can produce inequitable outcomes that can exclude or harm certain demographics. In this paper, we discuss preliminary insights from an ongoing effort aimed at better understanding barriers to equitable data-centric innovation. We report findings from a survey of 261 technologists and researchers who use data in their work regarding their experiences seeking adequate, representative datasets. Our findings suggest that age and identity play a significant role in the seeking and selection of representative datasets, warranting further investigation into these aspects of data-centric research and development.

cs.HC↗

arXiv2025

Towards Decoding Developer Cognition in the Age of AI Assistants

Background: The increasing adoption of AI assistants in programming has led to numerous studies exploring their benefits. While developers consistently report significant productivity gains from these tools, empirical measurements often show more modest improvements. While prior research has documented self-reported experiences with AI-assisted programming tools, little to no work has been done to understand their usage patterns and the actual cognitive load imposed in practice. Objective: In this exploratory study, we aim to investigate the role AI assistants play in developer productivity. Specifically, we are interested in how developers' expertise levels influence their AI usage patterns, and how these patterns impact their actual cognitive load and productivity during development tasks. We also seek to better understand how this relates to their perceived productivity. Method: We propose a controlled observational study combining physiological measurements (EEG and eye tracking) with interaction data to examine developers' use of AI-assisted programming tools. We will recruit professional developers to complete programming tasks both with and without AI assistance while measuring their cognitive load and task completion time. Through pre- and post-task questionnaires, we will collect data on perceived productivity and cognitive load using NASA-TLX.

cs.HC↗

arXiv2024

"For Us By Us": Intentionally Designing Technology for Lived Black Experiences

HCI research to date has only scratched the surface of the unique approaches racially minoritized communities take to building, designing, and using technology systems. While there has been an increase in understanding how people across racial groups create community across different platforms, there is still a lack of studies that explicitly center on how Black technologists design with and for their own communities. In this paper, we present findings from a series of semi-structured interviews with Black technologists who have used, created, or curated resources to support lived Black experiences. From their experiences, we find a multifaceted approach to design as a means of survival, to stay connected, for cultural significance, and to bask in celebratory joy. Further, we provide considerations that emphasize the need for centering lived Black experiences in design and share approaches that can empower the broader research community to conduct further inquiries into design focused on those in the margins.

cs.HC↗