Chigozirim Ifebi

Chigozirim Ifebi

Biography

ChiChi is a rising senior studying Applied and Computational Mathematics with a minor in Information and Data Science at Caltech. Having grown up in vibrant, multiethnic communities in New York, she is passionate about centering underrepresented communities in the creation and evaluation of AI systems. At MIT, she works under Prof. Danielle Wood to investigate how computing identity has shifted among middle schoolers due to AI usage in Zero Robotics. Beyond research, ChiChi is committed to making STEM spaces more equitable. She is the BSU President and has led student advocacy efforts at Caltech to expand Black Studies and increase support for underrepresented students. After graduation, she hopes to pursue a PhD in Human-Centered AI and ultimately shape AI policy that keeps pace with technological innovation while centering ethics and justice. Outside of work, she enjoys reading, playing Animal Crossing: New Leaf, browsing Pinterest, and trying new restaurants with friends.

Who Is The Programmer?: Impact of AI-Assisted Coding Usage on Identity
Formation in Zero Robotics
Chigozirim Ifebi¹, Alissa Chavalithumrong², Danielle Wood^2
1Department of Computing + Mathematical Sciences, California Institute of Technology
²Program in Media Arts & Sciences, Massachusetts Institute of Technology

This project aims to measure the usage of Artificial Intelligence (AI)-powered code generation tools over the past decade (2016–2026) in the Zero Robotics Middle School Competition and examine how this usage relates to middle school students’ perceptions of their identities as programmers. Research has proven K-12 robotics programming effective in improving STEM learning outcomes and fostering interest in engineering careers. However, a gap remains in research investigating the impact of AI coding assistants on K-12 students’ learning outcomes and computing identity — the “belief in one’s performance/competence, interest, and recognition in computing” — especially in informal environments like robotics programs. We finetune and evaluate CodeBERT, a pre-trained text classifier, on a custom labeled dataset of 300 Zero Robotics code submissions (2016–2021) and 300 corresponding scripts generated by GPT-4o mini, assessing performance via accuracy, precision, recall, and F1-score. In parallel, we use qualitative worksheets and interviews from 50 Zero Robotics middle schoolers to analyze computing identity and AI usage. Applying the computing identity framework, we identify themes of perceived competence, interest, and recognition in student responses. Results aim to help educators and program designers responsibly navigate AI integration in youth robotics competitions in ways that promote, rather than undermine, computing identity.