Rehber bağlantılı sınav · Orta Seviye

İK Testinde Yapay Zeka

İş uygunluğunu, vekalet önyargısını, grup değerlendirmesini, kayıt düzeltmeyi, insan incelemesini, dokümantasyonu ve yargı alanına özgü yasayı test eder.

İlgili kılavuz yollarıAI İK
Soru 1 arasında 8

An HR model ranks candidates. What should the ranking be treated as?

Her soruyu önizleyin8 sorular

Testi çözdüğünüzde cevaplar ve açıklamalar görünür.

  1. An HR model ranks candidates. What should the ranking be treated as?
    • A final measure of a candidate’s worth
    • Proof that an applicant will succeed
    • A recommendation that requires job-related human review, not a neutral fact about a person
    • A legally complete decision record
  2. Why can removing race and gender fields fail to remove bias?
    • Models cannot use indirect signals
    • Removing two fields guarantees fairness
    • Bias only enters through labels
    • Other variables may act as proxies or reflect historical patterns
  3. Which question should precede using AI to screen applicants?
    • What job-related outcome is being predicted, and who makes the final decision?
    • How many applicants can be rejected automatically?
    • Can the vendor hide its thresholds?
    • Should prior hiring outcomes be copied unchanged?
  4. A model performs well overall but poorly for applicants needing an accommodation. What is needed?
    • Ignore the difference because aggregate accuracy is high
    • Evaluate the complete application and review process across relevant groups and accommodations
    • Remove accommodation requests from the records
    • Use the score as the only criterion
  5. A candidate sees an incorrect work-history record in an automated screening process. Which safeguard matters?
    • A higher model threshold only
    • A longer automated assessment
    • A second score from the same data
    • A meaningful way to correct inaccurate records
  6. How should prior hiring decisions be used as model labels?
    • Assume they are objective ground truth
    • Use them without review because they are historical
    • Check whether they reflect relevant qualifications and outcomes before reuse
    • Replace them with the model’s predictions
  7. What should a vendor evaluation document?
    • Only product branding
    • Data sources, model versions, thresholds, and decisions
    • Only the number of employees screened
    • A promise of unbiased results
  8. Which current legal practice is safest when using AI for hiring?
    • Consult applicable law for the jurisdiction and specific use
    • Assume one global rule applies everywhere
    • Rely on the model to determine legality
    • Ignore local accommodation requirements