The Simulation That Screened Out Half the Women

Compute the rates. Men: 74/96 = 77.1%. Women: 26/48 = 54.2%. The ratio is 54.2/77.1 = 0.70, below the four-fifths threshold in the Uniform Guidelines. Overall pass rate is 108/156 = 69.2%, but a favorable bottom line does not cure a disparate impact at a component stage - that defense was rejected in Connecticut v. Teal.

Check significance. Pooled rate = 100/144 = 69.4%; standard error of the difference = square root of [0.694 x 0.306 x (1/96 + 1/48)] = 0.0815. The 22.9-point gap is about 2.8 standard errors - statistically significant by the two-standard-deviation convention, though the sample is small enough that a larger pool should be examined.

Assess validity. The vendor has supplied no validity study. A task list from a job posting plus three dispatchers reviewing items is not evidence of content validity: there is no job analysis linking critical work behaviors to test content, no documentation of item development, no reliability data, and no justification for the pass/fail cut score. Under the Guidelines, the user - not the vendor - bears the burden of showing job-relatedness.

Options, in order of preference: (1) suspend the simulation pending a proper validation study; (2) if retained, use it as one weighted component rather than a pass/fail gate, and only with criterion- or content-validity evidence and a defensible cut score; (3) re-open the eligibility list and reprocess affected candidates. Notify counsel, preserve all test records for the required retention period, and notify the vendor in writing that its validation package is inadequate.

The denial cannot rest on the vendor's statement. The ADA requires an individualized interactive process and a reasonable accommodation unless it imposes undue hardship or fundamentally alters the job requirements - and a vendor's assertion that a test 'cannot be modified' is a commercial claim, not a legal conclusion.

Steps:

1. Reopen each request and document the interactive dialogue: what functional limitation is at issue, what documentation supports it, what barriers the test creates.
2. Ask the vendor what alternatives exist - extended time, a practice version, a separate room, a human-administered or lower-fidelity version, altered pacing, or a different assessment altogether. Extra time on a dispatch simulation is frequently reasonable because the underlying construct is judgment and prioritization, not raw speed.
3. Analyze whether speed is an essential function. If response time is genuinely essential (it usually is in some measure), the question becomes whether the accommodation changes the essential requirement or merely the testing modality. Removing the time limit entirely may not be reasonable; extending it moderately usually is.
4. Decide in writing with reasons, and offer an alternative assessment where the accommodation is denied.

Also check Section 504 and state law if the center receives federal funds or is a public entity under Title II. The eight denials should be reviewed as a group - a pattern of blanket denials is itself evidence of a failure to engage in the interactive process.

The retaliation claim has all three elements: protected activity (the internal complaint), adverse action (status changed from eligible to inactive, removing her from consideration), and temporal proximity (immediately after the complaint). Proximity alone is not proof, but it shifts the burden to the employer to articulate a legitimate, non-retaliatory reason - and there is no obvious one on this record.

The supervisor's comment that the simulation was designed to 'weed out people who can't keep up' is worse than stray: it is direct evidence of intent that will be read against the employer in any charge, and it undermines the job-relatedness defense for the whole instrument.

Immediate actions:

• Issue a litigation hold; preserve test data, scoring records, emails, the vendor contract, and the complaint file.
• Remove the supervisor from all involvement in this selection process and from any contact with the complainant about her application.
• Restore the complainant's eligibility status pending review unless there is a documented, independent reason not to.
• Assign an investigator outside the communications department; interview the colleague who heard the comment.
• Issue the standard no-retaliation notice to everyone involved in the process.
• Have counsel assess whether the entire eligibility list must be re-derived.

Separately, review whether the complainant's identity and medical information were shared; accommodation records must be kept confidential and apart from the personnel file.