Ying Yee (Isabella) Yu

FairPlay Lab — equitable playing time in school soccer

An Auditable Mobile Decision-Support System for Equitable Playing Time and Team Cohesion in School Soccer · Ying Yee (Isabella) Yu, Kyler Harris · SoftCom Lab, Cal Poly Pomona

FairPlay Lab is a cross-platform app that turns records a volunteer coach can keep — training attendance, a seven-day RPE average, match minutes, and tallied pitch events — into three inspectable indicators: a squad fairness index, a hero-ball versus team-ball cohesion split, and a per-athlete academic crash-risk band. Every substitution suggestion opens an itemized explanation card. The scoring is rule-based, not learned.

Accepted at the 4th International Conference on Education, AI & Information Technology (EDUAIT 2026, Toronto). A model audit showed that RPE outweighed playing-time history, and exposed a high-load threshold discontinuity that inverted the intended penalty. Piloted across three schools. Live on the App Store; U.S. provisional patent 64/164,480 (29 Sep 2026).

Paper GitHub App Store Product site

AI detector bias on ELL school essays

False Positives on Free and Open AI Writing Detectors: A Replication–Extension Study of English Language Learner Essays in School Writing Corpora · Ying Yee (Isabella) Yu · mentored by Dr. Mani, Carnegie Mellon University

Liang et al. (2023) found high false-positive rates when GPT detectors scored non-native TOEFL essays. I tested whether the same ELL disadvantage appears on free and open detectors applied to school writing (ASAP 2.0 / ELLIPSE), using ZeroGPT and an open RoBERTa classifier on two samples of ~200 human-written essays each.

Contrary to my hypothesis, ELL essays were not flagged more often (ZeroGPT 1% vs 1% and 5% vs 0%; RoBERTa 26% vs 20% and 14% vs 9%; all p > .05). False positives tracked shorter, less lexically diverse text rather than ELL status. Schools should not treat a single free detector score as evidence of misconduct.

Paper GitHub

School resources, not school type

Basis Independent Brooklyn senior research

I’ll investigate whether differences in academic and college outcomes between New York City public and private school students can be explained by specific school resources rather than school type alone. Building on my previous research, I want to look beyond broad labels like “public” and “private” and explore which specific factors actually matter.

The effect of school choice on participants

Quantitative Evaluation for Public Policy · July 2026 · Irene Gong, Kristen Gumagay, Ying Yee (Isabella) Yu · presenting Cullen, Jacob & Levitt (2006)

Group 9’s final project walks through the causal design in Cullen, Jacob, and Levitt’s Econometrica study of Chicago Public Schools lotteries. Winning a seat is random, so differences between winners and losers identify an intent-to-treat effect; using the lottery as an instrument identifies LATE for students who actually enroll.

Across 194 lotteries and 14,434 students, winners attended objectively better schools, but test scores, graduation, attendance, and credits barely moved. Behavioral outcomes did: fewer disciplinary incidents and lower arrest and incarceration rates.

The exercise trained me to separate ITT from LATE, to name the identifying assumptions, and to treat a null academic result as a finding rather than a failure of the design.

Final project (PDF)

The most sustainable Super Bowl host

Team 17156 · Problem B · 2025 · Ying Yee (Isabella) Yu

We built an AHM–PCA–PHM hybrid weighting model plus TOPSIS and GM(1,1) forecasts to rank host-city sustainability on 24 indicators across carbon, energy, transport, water, waste, and lodging. Transportation (weight 0.32) and energy (0.26) dominated. Minneapolis scored 0.780 and is the recommended 2029 host.

International honorable mention, top 20% of HiMCM papers.

Paper (PDF)

Should we be polite to ChatGPT?

John Locke Institute Essay Competition · Science & Technology · 2026 · Ying Yee (Isabella) Yu

The essay argues that politeness to AI chatbots is not empty courtesy: it can improve answer quality, preserve civility as a practiced habit, and hedge against longer-run loss of control. I used prompt-politeness studies alongside moral-psychology accounts of digital manners.

Shortlisted in Science & Technology, top 17.5%.

Essay (PDF) Essay (Word)