Lab research and project pages are also hosted at riseilab.org — the dedicated RISEI Lab site.
Research NEW: AI Measurement Disability Employment Wage Policy & Pay Equity Labor Economics Vocational Rehabilitation & ROI Inclusive Education Education Economics Human Capital Cost-Benefit Analysis Racial Disparities Accessibility & Universal Design

New Working Paper & Policy Brief

AI occupational exposure scores are unstable — and the policies built on them are too

New — April 2026

How (un)Stable Are LLM Occupational Exposure Scores?

Yin, M., Vu, H., & Persico, C. | RISEI Working Paper, April 2026

Every major forecast about which jobs AI will eliminate comes from asking AI to rate itself. We replicated the dominant rubric with three frontier models on 18,797 tasks. The scores diverge 3.6-fold, agreement is as low as 57%, and employment estimates flip sign depending on which model is used. The companion policy brief proposes the Adaptive Precision Framework.

3.6×
Score Divergence
57%
Worst Agreement
2.4×
Estimate Variation
±
Sign Flips

Disability Employment & Wage Policy

Eliminating subminimum wages and advancing competitive integrated employment

18 States
have eliminated 14(c)
Flagship Project

Subminimum Wage Elimination & Virginia EPIC

$4.29M | U.S. Department of Education (2022-2028)

First national quasi-experimental analysis of Section 14(c) elimination. No aggregate job loss, 12.4% decline in welfare dependence. Virginia EPIC transitions workers to competitive integrated employment.

0
Job Loss
−12.4%
Welfare Dependence
~2,000
Workers/State
129
VA Workers Left

Related Publications

Active Project

Maine Pathways to Partnerships

$2.98M | Federal Model Project (2023-2028)

Developing interagency collaboration models for youth with disabilities transitioning from education to employment. Features the ROI of vocational rehabilitation and interagency coordination.

$2.55
ROI per $1 (3.5yr)
$21.54
ROI Lifetime
2x
Employment w/ Coordination

Related Publications

Young professionals in workplace
Student working on digital assessment
Assessment Research

NAEP Process Data & Digital Assessment Accessibility

Research Initiative | National Assessment of Educational Progress

Using large-scale NAEP log data to understand how students solve math problems and who benefits from extended time and other accommodations in digital assessments.

72%
Didn't use full time
24%
Still active at timeout
3
Solving Strategies

Related Publications (3 papers)

RANDOMIZED EXPERIMENT Introductory microeconomics course R RANDOM ASSIGNMENT LIVE LECTURE In-person, same instructor & content INTERNET LECTURE Recorded same lectures, identical materials FINDING Live instruction modestly outperforms internet-only Strongest effects: Hispanic, male, and lower-achieving students
Signature Paper · Education Economics

Is It Live or Is It Internet? Experimental Estimates of the Effects of Online Instruction on Student Learning

Figlio, Rush & Yin | Journal of Labor Economics, 31(4), 2013 | NBER WP #16089

First randomized experimental comparison of live vs. internet instruction in higher education. Students in a large introductory microeconomics course were randomly assigned to live lectures or to watch the same lectures online, with all other factors held constant. Finds modest evidence that live instruction dominates internet-only — with stronger effects for Hispanic, male, and lower-achieving students.

1st
Experimental Study
JLE
Journal of Labor Econ.
3
Subgroups w/ Strongest Effects
Inclusion Research

Special Olympics Unified Champion Schools

Global Evaluation | 170+ countries, 7,800+ schools

Evaluating how whole-school inclusion programs affect academic and behavioral outcomes for all students, with 11 years of student-level data and rigorous econometric methods.

11 yrs
Student Data
1M+
Students
314
UCS Schools

Related Publications

  • Yin & Seo (2024). Global Evaluation of the Special Olympics UCS Program in Six Countries (Argentina, Egypt, India, Pakistan, Romania, Rwanda). Final Report
  • Yin, Siwach & Orellana (2025). AERA Open
  • Yin, Siwach & Belyakova (2020). AERJ
Diverse group of young people together

Disability Economics & Pay Equity

Quantifying the economic impact of disability employment disparities

Hidden Market

$490B

After-tax income of working-age adults with disabilities. A significant, underserved consumer market.

Uneven Playing Field

$141B

GDP impact if workers with disabilities were paid comparably. Gap widens with education level.

One Size Does Not Fit All

18-56%

Labor force participation varies dramatically by disability type. Policy must be tailored.

Leading the Way, or Falling Behind?

Boston pay gap: ~$24,000 for workers with disabilities, nearly $10K above the national average.

View Research →

Racial Disparities in VR

Identifying systemic inequities in access to and outcomes from vocational rehabilitation services.

View Research →