arXiv subjects

Cleidson R. B. de Souza

Publications and source records attributed to Cleidson R. B. de Souza.

6 recordsLinked to original sources

arXiv2026

Toward Inclusive AI-Driven Development: Exploring Gender Differences in Code Generation Tool Interactions

The increasing reliance on Code Generation Tools (CGTs), such as Claude Code and GitHub Copilot, is revamping programming workflows and raising critical questions about fairness and inclusivity in human-AI collaboration. While CGTs offer potential productivity enhancements, their effectiveness across diverse user groups have not been sufficiently investigated. We hypothesized that developers' interactions with CGTs vary based on gender, influencing task outcomes and cognitive load, as prior research suggests that gender differences can affect technology use and cognitive processing. This study employed a mixed-subjects design with 39 participants, evenly divided by gender for a counterbalanced design. Participants completed two programming tasks of medium to high difficulty using two distinct treatments: only CGT assistance and only internet access. Task orders and conditions were counterbalanced to mitigate order effects. We collected cognitive load surveys, screen recordings, and task performance metrics such as completion time, code correctness, and CGT interaction behaviors. Our results indicate no statistically significant gender differences in cognitive load or performance outcomes when using CGTs compared to Internet-based workflows. CGTs reduce intrinsic and extraneous cognitive load compared to Internet based workflows, but the reduction was not statistically significantly. However, CGTs improved advanced code correctness. Our results suggest that CGTs can lower cognitive load and enhance performance on complex coding tasks without significantly affecting core correctness or completion time. These findings highlight how CGT usage can reduce cognitive burden and support more equitable programming experiences across users.

cs.SE↗

arXiv2026

When Who You Are Can Change the Code You Get: A Study of Persona-Induced Bias in LLM Code Generation

Large Language Models (LLMs) are widely used as programming assistants, yet it remains unclear whether and how user's demographic information impacts the technical quality of generated code. We conduct a large-scale empirical study of persona-induced bias in LLM-based code generation, focusing a proprietary model (Gemini 2.5 Pro) and an open-weight model (GPT-OSS-120B). Using 18 demographic personas spanning nationality, gender, and experience level, we compare persona-induced prompts against a neutral baseline. Across 35,000+ generated programs, we analyze demographic marker leakage in reasoning and responses, as well as differences in functional correctness, maintainability, code style, and security. Our results show that demographic cues are frequently reflected in LLM reasoning and outputs. Demographic markers appear in up to 65% of responses and 70% of reasoning traces, despite being semantically irrelevant to the tasks. On LiveCodeBench, persona prompting were associated with lower correctness scores of the Gemini model by an average of 1.54 percentage points, with one persona exhibiting a decrease of 3.6% (odds ratio = 0.51). In contrast, the accuracy of the GPT-OSS model improved by 3.4 - 5.7% across all personas (odds ratios = 1.8 - 3.0). Maintainability and code style metrics show statistically significant but negligible effect sizes (all Cliff's δ < 0.15), and security vulnerabilities exhibit no systematic persona-specific patterns. Overall, our results show that the presence of demographic information about users is associated with measurable variation in LLM reasoning and code quality even in purely technical tasks, and that these effects hold across models. Our work highlights an under-examined risk in LLM-assisted software development.

cs.SE↗

arXiv2025

Cracking CodeWhisperer: Analyzing Developers' Interactions and Patterns During Programming Tasks

The use of AI code-generation tools is becoming increasingly common, making it important to understand how software developers are adopting these tools. In this study, we investigate how developers engage with Amazon's CodeWhisperer, an LLM-based code-generation tool. We conducted two user studies with two groups of 10 participants each, interacting with CodeWhisperer - the first to understand which interactions were critical to capture and the second to collect low-level interaction data using a custom telemetry plugin. Our mixed-methods analysis identified four behavioral patterns: 1) incremental code refinement, 2) explicit instruction using natural language comments, 3) baseline structuring with model suggestions, and 4) integrative use with external sources. We provide a comprehensive analysis of these patterns .

cs.SE↗

arXiv2025

CodeWatcher: IDE Telemetry Data Extraction Tool for Understanding Coding Interactions with LLMs

Understanding how developers interact with code generation tools (CGTs) requires detailed, real-time data on programming behavior which is often difficult to collect without disrupting workflow. We present \textit{CodeWatcher}, a lightweight, unobtrusive client-server system designed to capture fine-grained interaction events from within the Visual Studio Code (VS Code) editor. \textit{CodeWatcher} logs semantically meaningful events such as insertions made by CGTs, deletions, copy-paste actions, and focus shifts, enabling continuous monitoring of developer activity without modifying user workflows. The system comprises a VS Code plugin, a Python-based RESTful API, and a MongoDB backend, all containerized for scalability and ease of deployment. By structuring and timestamping each event, \textit{CodeWatcher} enables post-hoc reconstruction of coding sessions and facilitates rich behavioral analyses, including how and when CGTs are used during development. This infrastructure is crucial for supporting research on responsible AI, developer productivity, and the human-centered evaluation of CGTs. Please find the demo, diagrams, and tool here: https://osf.io/j2kru/overview.

cs.SE↗

arXiv2024

Assisting Novice Developers Learning in Flutter Through Cognitive-Driven Development

Cognitive-Driven Development (CDD) is a coding design technique that helps developers focus on designing code within cognitive limits. The imposed limit tends to enhance code readability and maintainability. While early works on CDD focused mostly on Java, its applicability extends beyond specific programming languages. In this study, we explored the use of CDD in two new dimensions: focusing on Flutter programming and targeting novice developers unfamiliar with both Flutter and CDD. Our goal was to understand to what extent CDD helps novice developers learn a new programming technology. We conducted an in-person Flutter training camp with 24 participants. After receiving CDD training, six remaining students were tasked with developing a software management application guided by CDD practices. Our findings indicate that CDD helped participants keep code complexity low, measured using Intrinsic Complexity Points (ICP), a CDD metric. Notably, stricter ICP limits led to a 20\% reduction in code size, improving code quality and readability. This report could be valuable for professors and instructors seeking effective methodologies for teaching design practices that reduce code and cognitive complexity.

cs.SE↗

arXiv2020

Academic viewpoints and concerns on CSCW education and training in Latin America

Computer-Supported Cooperative Work, or simply CSCW, is the research area that studies the design and use of socio-technical technology for supporting group work. CSCW has a long tradition in interdisciplinary work exploring technical, social, and theoretical challenges for the design of technologies to support cooperative and collaborative work and life activities. However, most of the research tradition, methods, and theories in the field follow a strong trend grounded in social and cultural aspects from North America and Western Europe. Therefore, it is inevitable that some of the underlying, and established, knowledge in the field will not be directly transferrable or applicable to other populations. This paper presents the results of an interview study conducted with Latin American faculty on the feasability, viability, and prospect of a curriculum proposal for CSCW Education in Latin America: To this end, we conducted nine interviews with faculty currently based in six countries of the region, aiming to understand how a CSCW course targeted to undergraduate and/or graduate students in Latin America might be deployed. Our findings suggest that there are specific traits that need to be addressed in such a course, such as: tailoring foundational CSCW concepts to the diversity of local cultures, motivating the involvement of students by tackling relevant problems to their local communities, and revitalizing CSCW research and practice in the continent.

cs.HC↗