← Back to Dashboard

Methodology & Sources

The Risk Score Formula

Every degree receives a score from 0 (safest) to 100 (most dangerous). The score is computed from five weighted factors:

score = (
0.20 × unemployment_normalized +
0.25 × automation_risk_normalized +
0.35 × debt_salary_ratio_normalized +
0.20 × acceleration_normalized
) × 100
Unemployment (20%) - Recent graduate unemployment rate, normalized against a 15% ceiling.
Automation Risk (25%) - Probability of significant AI automation within 2 years, on a 0-100 scale.
Debt-to-Salary Ratio (35%) - Average total debt divided by median starting salary, normalized so that a 2:1 ratio (debt = 2x salary) hits maximum. This is the highest-weighted factor because it directly measures the personal financial trap.
AI Acceleration (20%) - The rate of unemployment increase since ChatGPT launched (November 2022). A field that went from 2% to 6% unemployment is in freefall. A field sitting at 6% for a decade is stable. This factor captures the velocity of disruption, not just the snapshot.

AGI / Singularity Mode

The toggle in the bottom-right corner applies a multiplier (1.5x–5x) to all AI-related factors: automation risk scores and the AI-driven portion of unemployment increases. At 2x, you're seeing what happens if AI capabilities double their current displacement rate. At 5x, you're modeling a near-AGI scenario where most cognitive tasks become automatable within 2 years. The baseline data stays untouched - only the AI amplification changes.

Major Data Sources

Unemployment Rates

Automation Risk

Student Debt

Starting Salaries

Trend / Historical Data

How We Calculate Automation Risk

Step 1: Start with Frey & Osborne's automation probability scores for 702 occupations.

Step 2: Map each degree to its top 3-5 career outcomes using BLS occupation-by-degree data.

Step 3: Apply a 2-year acceleration multiplier based on: current AI coding/writing/analysis benchmarks, enterprise AI adoption rates (McKinsey 2025: 72% of companies deploying), and rate of improvement in AI capabilities (GPT-3.5 → GPT-4 → o3 trajectory).

Step 4: Weight by what percentage of graduates actually enter those roles (not just what the degree enables, but where people end up).

Result: A defensible, citable 2-year automation risk score per degree.

The Acceleration Factor

Most tools look at unemployment as a snapshot. We measure the derivative - how fast it's changing. Specifically, the ratio of current unemployment to pre-ChatGPT unemployment (before November 2022).

Computer Science: 2.8% → 6.1% = 2.18x increase in ~3 years. Software Engineering: 2.5% → 7.5% = 3.0x increase. Nursing: 1.6% → 1.9% = 1.19x (barely moved).

This is the signal that existing tools miss entirely. A field with stable 5% unemployment is categorically different from a field where unemployment tripled in 3 years - even if the current number is the same.

School Worth-It Score

Every university receives a score from 0 (worst value) to 100 (best value). Note: this is inverted from the major risk score - higher means the school is a better bet. The score is computed from four weighted factors:

school_score = (
0.35 x debt_salary_value +
0.20 x ai_readiness +
0.20 x network_premium +
0.25 x post_ai_employability
) x 100
Debt-to-Salary Value (35%) - Average debt at graduation divided by median earnings 10 years post-entry. Lower ratio = higher score. Capped at 1.5x ratio. Source: College Scorecard.
AI Readiness (20%) - Composite of: percentage of degrees in STEM fields (30% weight), whether the school has a dedicated AI/ML program (40% weight), and Carnegie research classification R1/R2 (30% weight). Schools with strong AI programs and research output score highest.
Network Premium (20%) - Alumni salary uplift above national median (60% weight) combined with startup density per 1,000 alumni (40% weight). Startup data from PitchBook University Rankings. Measures whether the school's brand and alumni network provide tangible career advantages beyond education.
Post-AI Employability (25%) - Inverse of the weighted average automation risk across the school's most popular majors. Schools whose graduates cluster in AI-resilient fields score higher. In AGI mode, this factor gets amplified.

School Data Sources

Core Financial Data

AI Programs & Research

Network & Startup Data

Limitations & Disclaimers

These are estimates and extrapolations, not official statistics. Automation risk in particular is inherently speculative - nobody knows exactly when AI will be able to do X. Our scores represent informed projections based on current capability trajectories, not certainties.

Unemployment data varies by source and methodology. The NY Fed measures recent graduates differently than BLS measures all workers. We cite our specific source for each data point.

This is not financial advice. Individual outcomes vary enormously by school, location, specialization, internship experience, and luck. These scores describe averages across all graduates of a given major.

All data and methodology are open source on GitHub. If you find an error or have better data, submit a pull request.

← Major RankingsSchool RankingsGitHub Repository