引導式測驗 · 中等 等級

膚色偏見與 Monk 膚色尺度測驗

測驗量表的目的和起源、種族和臨床分類的區別,以及有關納入和註釋不確定性的研究證據。

問題 1 的 8

How many shades are in the Monk Skin Tone Scale?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. How many shades are in the Monk Skin Tone Scale?
    • 十
    • 六
    • Forty
    • 一百
  2. Which task is the MST Scale intended to support?
    • A legal determination of race
    • A diagnosis of skin cancer
    • A test of ultraviolet sensitivity
    • Skin-tone representation and evaluation in technology research and products
  3. Which researcher developed the scale in partnership with Google?
    • Ziad Obermeyer
    • Ellis Monk
    • Alexandra Chouldechova
    • Matt Kusner
  4. Does an MST category identify a person’s race or ethnicity?
    • No, it only measures eye color
    • Yes, it determines ancestry exactly
    • No; it is a visual skin-tone reference, not a race classifier
    • Yes, each shade maps to one ethnic group
  5. What did the 2,214-person comparison study find about the Fitzpatrick scale?
    • It was universally preferred across every group
    • Participants perceived it as less inclusive than the Fenty and Monk scales, especially among some marginalized groups
    • It measured clinical diagnoses perfectly
    • It showed skin-tone scales are unnecessary
  6. Why can annotators disagree when assigning a skin-tone label to an image?
    • The MST has only one shade
    • Images contain no color information
    • Skin tone is subjective and image appearance varies with capture and viewing conditions
    • Annotators must infer the person’s legal identity
  7. Why is the Fitzpatrick scale not a direct substitute for an image-skin-tone scale?
    • It measures only hair color
    • It contains ten shades
    • It was designed as an AI fairness law
    • It was designed around burning and tanning response, not universal visible skin color
  8. Which practice improves an AI skin-tone audit using MST labels?
    • Use varied lighting and cameras, multiple annotators, and report coverage and disagreement
    • Treat the label as proof of fairness
    • Assume the shade label is objective and complete
    • Use only one image of one shade