引導式測驗 · 中等 等級
臉部表情辨識與其限制測驗
測試可見面部動作和推斷情緒之間的區別,以及註釋偏差、參與者層級的評估和同意。
問題 1 的 8
A classifier labels a smile as “happy.” What has it directly measured?
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。
- 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
- 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
- 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
- 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
- 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
- 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
- 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
- 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