Викторина, связанная с руководством · Средний Уровень
Викторина о возрастной предвзятости в системах искусственного интеллекта
Сравнивает случай найма EEOC, биометрические оценки NIST и исследования эйджизма LLM, сохраняя при этом каждый результат привязанным к проверенной системе и образцу.
Вопрос 1 из 8
What did the EEOC allege about iTutorGroup’s application software?
Предварительный просмотр каждого вопроса8 вопросы
Ответы и пояснения появляются по мере прохождения теста.
- What did the EEOC allege about iTutorGroup’s application software?
- It rejected every applicant under 40
- It was used only to estimate age in photos
- It automatically rejected female applicants age 55+ and male applicants age 60+
- It had no automated screening feature
- What did NIST report about demographic effects in face-recognition algorithms?
- Every algorithm had the same error rates
- Age had no effect on performance
- Many tested algorithms showed demographic differentials, with magnitude varying by algorithm and application
- The test measured LLM attitudes only
- How do face recognition and age estimation differ?
- Recognition matches identity; age estimation predicts an age from a face image
- Age estimation verifies a person’s legal identity
- Both tasks always produce the same output
- Face recognition predicts a birth year only
- What factors did NIST say can affect age-estimation accuracy?
- Whether the image is in color
- Image quality, gender, region of birth, age and interactions among factors
- Only the model’s file size
- Only the person’s job title
- What did the 2024 LLM age-value study evaluate?
- Six selected LLMs using prompts about age-related value orientations
- Every AI product used by older adults
- Only face-recognition error rates
- A clinical diagnosis of cognitive impairment
- Why should NIST age-estimation error not be treated as a universal result for all systems?
- NIST measured every system ever deployed
- The evaluation covers submitted algorithms and datasets; results vary by model, group and image conditions
- An average error proves every group has identical error
- Age estimation and hiring are the same task
- Which hiring audit best tests age-related effects?
- Use age as the only quality measure
- Compare unrelated applicants for different jobs
- Test one applicant once
- Compare equivalent applications while varying age cues and keeping job qualifications constant
- Which broad conclusion is supported across the cited age-bias research?
- NIST’s results determine whether a hiring model is lawful
- Every AI system discriminates against older adults
- Older adults cannot use digital tools
- Age effects vary by task and model, so evaluate the actual use and age groups involved