Rethinking Hiring Tests: Qualitative Analysis for Autism

According to the PLOS One study (Whelpley et al., 2026), commonly used hiring tests can reproduce the same adverse outcomes for autistic applicants that earlier work attributed to interviews.

• Whelpley et al., 2026 collected responses from June 6-9, 2022, with 107 autistic and 99 neurotypical respondents.
• Whelpley et al., 2026 reported GMA means of 5.96 for neurotypical and 4.73 for autistic respondents in June 2022 (Cohen’s d = 0.52).
• Whelpley et al., 2026 reported SJT means of 4.24 for neurotypical and 0.57 for autistic respondents in June 2022 (Cohen’s d = 0.84).
• Whelpley et al., 2026 concluded that "commonly used selection tools... may nonetheless result in adverse impact for autistic applicants if used in hiring decisions, " calling for assessment redesign rather than blind tool substitution.

Direct answer: Whelpley et al., 2026 compared autistic and neurotypical adults on three standard selection tools and found consistent subgroup differences unfavorable to autistic respondents.

Whelpley et al., 2026 recruited two samples via Amazon Mechanical Turk, one neurotypical sample (n = 99, mean age 37.8) and one autistic sample (n = 107, mean age 35.1), with data collected from June 6, 2022, to June 9, 2022, as reported in PLOS One.

Whelpley et al., 2026 administered a 50-item Five-Factor personality inventory, the 16-item ICAR-16 general mental ability (GMA) measure, and a 12-stem situational judgment test (SJT) validated for customer service; the study reports internal reliabilities and mean differences in PLOS One.

Whelpley et al., 2026 used standardized mean comparisons and multivariate regressions to test whether SJT differences persisted after controlling for GMA, personality, gender, and age; the regression showed autism remained a significant negative predictor of SJT score in the PLOS One results.

Limitations noted by Whelpley et al., 2026 in PLOS One include lower-than-expected reliability for personality scales in the autistic sample and the SJT being developed and validated in a predominantly neurotypical workforce.

Direct answer: Organizations should not assume standard tests automatically increase equity for autistic applicants; instead, selection design must target job-relevant skills and avoid conflating social inference with task ability.

Whelpley et al., 2026 in PLOS One show that personality inventories, GMA tests, and SJTs were each associated with subgroup differences unfavorable to autistic applicants, indicating that replacing interviews with these tools can reproduce adverse impact.

Whelpley et al., 2026 recommend emphasizing task-based assessments such as work samples and assessment centers for roles where interpersonal inference is not central, because those approaches evaluate observable job performance rather than normative social judgments.

Practitioner testimony cited in Whelpley et al., 2026 (for example, SAP’s hiring programs) highlights why companies frame neurodiverse talent as strategic: SAP reportedly said “innovation comes from the edges, ” which organizations cited as part of a project-based evaluation approach.

Whelpley et al., 2026 in PLOS One provide clear quantitative evidence that widely used selection tools can produce subgroup differences unfavorable to autistic applicants and that SJT disparities can persist after standard statistical controls.

AI-enabled qualitative research turns those numeric findings into actionable fixes by identifying which item framings, social cues, or normative assumptions cause misalignment.

If your team wants to pair quantitative adverse-impact checks with rapid qualitative diagnostics and redesign, Try Evidano for free to pilot thematic analysis on your selection items and candidate responses.