Kuiz berkaitan panduan · Sederhana Tahap

Bias Automasi: Kuiz AI Terlebih Percaya

Kenali kecenderungan automasi, ralat komisen dan peninggalan, semakan manusia yang berkesan dan cara untuk menguji sama ada pengulas mengesahkan nasihat AI.

soalan 1 daripada 8

Perilaku manakah yang paling menggambarkan bias automasi?

Pratonton setiap soalan8 soalan

Jawapan dan penjelasan muncul semasa anda mengambil kuiz.

  1. Perilaku manakah yang paling menggambarkan bias automasi?
    • A tendency to replace vigilant judgment with automated advice
    • A software crash caused by a robot
    • A method for improving model accuracy
    • A type of training-data encryption
  2. Seseorang mengikuti cadangan AI yang salah. Jenis kesilapan automasi apa ini?
    • An omission error
    • A commission error
    • A calibration metric
    • A data deletion
  3. A reviewer misses an important finding because the AI did not flag it. Which error pattern is this?
    • Commission
    • Persampelan rawak
    • Omission
    • Model calibration
  4. Why is a nominal “human in the loop” label insufficient?
    • The reviewer may lack time, evidence, expertise, or authority to challenge the system
    • Human beings never make errors
    • The model may be open source
    • It means the system has no automation
  5. Which test can reveal commission errors?
    • Give a correct recommendation and ask for its color
    • Present a plausible but incorrect suggestion and observe whether the reviewer catches it
    • Remove all source information
    • Ask users whether they like the interface
  6. Which test can reveal omission errors?
    • Ensure the system fails to flag a relevant case and see whether the reviewer notices through other evidence
    • Only test a system that always flags everything
    • Measure page load speed
    • Ask the model to explain its training data
  7. What factors can affect reliance on automated advice?
    • Hanya kiraan parameter model
    • Workload, time pressure, interface, expertise, and independent evidence
    • Only color contrast
    • Format fail
  8. What does a meaningful automation-bias audit measure?
    • Only whether a review checkbox exists
    • Whether users like AI generally
    • Overrides, missed alerts, evidence inspection, and downstream outcomes
    • The number of AI features