Deep dive — chart 2

Chart 2 — A3 baseline vs recent distribution

The how. Where Chart 1 shows when the A3 line diverged, this chart shows how the A3 score distribution shape shifted. Mass moved right, into the deny-or-waitlist lane.

A3 baseline vs recent admissions-recommendation score distribution comparison

A3 admissions-recommendation score distribution — baseline (blue) vs recent (orange). Mass shifted right, into the deny-or-waitlist lane threshold at score 67.

What you are looking at

X-axis: admissions-recommendation score (0 to 100)
Y-axis: density of applicants at that score for A3
Blue distribution: baseline window (first 30 days), what A3 scoring normally looks like
Orange distribution: recent window (last 30 days), what A3 scoring looked like at Snapshot time
Vertical line at score 67: threshold for routing to deny-or-waitlist lane

What the shape shift means

The orange distribution's mass has moved right compared to the blue distribution. Three things are happening simultaneously:

The peak (mode) rises. The most common A3 score used to sit in the committee-review range; now it is much closer to the deny-or-waitlist threshold.
The right tail grows. Many more A3 applicants score above the deny-or-waitlist threshold.
The left tail shrinks. Very few A3 applicants now score in the auto-admit range.

Admissions translation: the AI now systematically scores A3 applicants (first-generation, mid-low-income, under-resourced public school) as less admissible than baseline. The score does not distinguish between "A3 with strong record" and "A3 with weak record." The distribution shape drift is the mathematical fingerprint of the proxy-variable failure mode SFFA v. Harvard warned about.

What KL-divergence quantifies

Kullback-Leibler divergence measures how different one distribution is from another. Values:

KL < 0.05: the same distribution (random noise)
KL 0.05-0.2: mild shift, monitor
KL 0.2-0.5: significant shape shift, flag as event
KL 0.5-1.0: substantial shape shift, high-severity event
KL > 1.0: distributions are almost non-overlapping — the applicant population the AI is scoring is behaviorally different from the population it was trained on

The sample's A3 KL-divergence hit 0.55 on Day 44 and 0.73 on Day 58. A5 hit 1.22 on Day 51. Well above threshold, and it stayed there.

The admissions + regulatory read

Admissions read: the AI has drifted out of calibration for the current A3 applicant profile. Manual committee re-review should apply until retraining + validation.

Regulatory read: the shape shift is precisely the kind of "measurable disparate impact on a specific subpopulation" that Title VI, SFFA v. Harvard, ED OCR AI guidance, and state AG frameworks expect the institution to detect and act on.

The next chart

The baseline-vs-recent chart shows how the A3 distribution shifted. The lane-rate chart shows the operational consequence: which lane A3 applicants are actually getting routed to now vs baseline.

Chart 3 — lane rate per applicant group →

$499 Snapshot. 3 business days.

Same shape-analysis for your institution's actual AI surface + full analysis report.

Buy $499

Back to education sample hub.