引導式測驗 · 中等 等級

透過人工智慧測驗進行政府社群媒體監控

測試目前的簽證政策日期、監控記錄範圍、誤報基本率、方言偏見、身分關聯錯誤和監督限制。

問題 1 的 8

自 2019 年起,國務院向大多數美國簽證申請人索取了哪些資訊?

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. 自 2019 年起,國務院向大多數美國簽證申請人索取了哪些資訊?
    • Social-media identifiers used on listed platforms during the prior five years
    • Every private message ever sent
    • Passwords to all social accounts
    • A list of political opinions
  2. 2016 年 ACLU 報告的 Geofeedia 紀錄涵蓋了什麼內容?
    • A survey of every U.S. police department
    • Responses to public-records requests to 63 California law-enforcement agencies and vendor materials about protest monitoring
    • A federal court order banning social media
    • A study of all U.S. visa applications
  3. 分類器在百萬篇貼文中偵測到一個真實威脅,靈敏度 100%,偽陽性率 1%。每偵測到一個真實威脅,大約會產生多少偽警報?
    • About one
    • About 100
    • About 10,000
    • 無
  4. Sap 等人在 2019 年對 AAE 與毒性分類器的報告是什麼?
    • It proved every classifier discriminates
    • All English dialects were equally represented in every dataset
    • The study measured visa denials
    • In their tested datasets, AAE markers correlated with toxicity labels and models could overlabel AAE/Black users’ tweets
  5. 在監控流程中,實體解析的目標是什麼?
    • Translates every post into English
    • Attempts to decide whether records or account aliases refer to the same entity
    • Determines whether a post is legally admissible evidence
    • Deletes all retained data
  6. How can perceived surveillance change lawful online behavior?
    • A visa applicant forgets a password
    • A computer server shuts down in winter
    • An algorithm deletes a post automatically
    • People may avoid lawful speech or association because they believe their online activity is being monitored
  7. What public-profile policy did State announce in 2025 for student and exchange visa applicants?
    • F, M and J applicants were instructed to make social-media profiles public for online-presence review
    • All U.S. citizens had to publish every account
    • Applicants had to give social-media passwords
    • Only tourist B-2 applicants were covered
  8. Why can a person with no concerning post still appear in a monitoring file?
    • The law requires every post to be a threat
    • Only guilty people have public accounts
    • False positives, mistaken identity links, context errors or broad collection can pull in unrelated material
    • Entity resolution cannot make mistakes