Guide-linked quiz · Medium Level

Facial Expression Recognition and Its Limits Quiz

Test the distinction between visible facial actions and inferred emotion, plus annotation bias, participant-level evaluation, and consent.

Question 1 of 8

A classifier labels a smile as “happy.” What has it directly measured?

Preview every question8 questions

Answers and explanations appear as you take the quiz.

  1. A classifier labels a smile as “happy.” What has it directly measured?
    • The subject’s self-reported feeling at that moment
    • A direct biological measurement of happiness
    • An image pattern that its dataset associates with that label
    • The person’s reason for making the movement
  2. Why can forced-choice emotion labels exaggerate what a benchmark proves?
    • Annotator agreement among supplied words does not verify the subject’s actual feeling
    • High agreement verifies the subject’s inner state
    • A restricted label list eliminates context effects
    • Every annotator sees the same meaning in a cropped face
  3. A dataset has many frames per participant. Which split better tests use on new people?
    • Randomly divide consecutive frames from each person
    • Keep one frame per person in training and nearby frames in test
    • Tune category thresholds on the final test participants
    • Keep all frames from each participant on one side of the split
  4. Which claim is supported by the Barrett and colleagues review?
    • A fixed movement uniquely identifies happiness across contexts
    • Facial movements do not map reliably and uniquely to emotion categories across contexts
    • The same expression has one meaning when lighting is controlled
    • Culture and situation are irrelevant once the face is visible
  5. What additional evidence helps study whether a task was frustrating?
    • More cropped face frames from the same session alone
    • A higher classifier confidence threshold alone
    • Participant self-report and task context alongside observable behavior
    • Labels from viewers who never saw the task context alone
  6. Why is a model confidence score not the probability that a person feels an emotion?
    • It scores a dataset category under the model’s learned labeling scheme
    • It is calibrated against each subject’s private experience
    • A larger score directly measures stronger felt intensity
    • It measures biological emotion rather than training labels
  7. A school wants to punish a student because software says their face looks angry. Which concern is central?
    • A good score on cropped photos proves the disciplinary use
    • More emotion classes remove uncertainty about intent
    • An automated score is sufficient without context or appeal
    • The decision infers intent from uncertain facial cues in a consequential setting
  8. Which project has the narrower and more testable goal?
    • Infer honesty from a job applicant’s brief face crop
    • Detect whether an explicitly defined eyebrow movement occurs in consented video
    • Classify a patient’s diagnosis from smile intensity
    • Score motivation without asking the learner

Keep testing yourself

More quizzes picked for your level.

View all quizzes