引導式測驗 · 中等 等級

人工智慧在法律研究和判例法搜尋測驗中的應用

測試合法人工智慧工具如何使用檢索增強生成、聊天機器人為何偽造引文以及馬塔訴阿維安卡案製裁的教訓。

問題 1 的 8

為什麼 Mata v. Avianca(2023)案中的律師會受到懲戒?

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. 為什麼 Mata v. Avianca(2023)案中的律師會受到懲戒?
    • They cited cases that ChatGPT had invented and stood by them when questioned
    • They used a Westlaw account without a valid license
    • They missed a filing deadline because an AI calendar tool failed
    • They pasted confidential client documents into a public chatbot
  2. 檢索增強生成(RAG)在法律研究工具中起到什麼作用?
    • It trains a new model from scratch on each firm's past briefs
    • It lets the model answer from memory without any database
    • It searches a curated database of real authorities first, then has the model answer from the retrieved passages
    • It replaces citators by predicting whether a case is still good law
  3. 為什麼通用型聊天機器人能夠產生看似令人信服、但實際上並不存在的判例引文?
    • It deliberately inserts fake cases to test users
    • It predicts plausible text by default instead of looking cases up in a database
    • It only has access to overruled cases
    • Court databases block chatbots, so it has to guess the case names
  4. 当合法的人工智能答案“有误”时,这意味着什么?
    • The cited case does not exist anywhere
    • The tool refused to answer the question
    • The answer came from a different country's law database
    • The cited authority is real but does not support the claim it is cited for
  5. 混合检索中,什么关键词索引(如BM25)特别擅长捕捉?
    • Passages that mean the same thing in completely different words
    • Exact terms of art, statute numbers and party names
    • Whether the judge who wrote the opinion is still on the bench
    • The emotional tone of a court's opinion
  6. What role do citators such as KeyCite and Shepard's play in AI-assisted research?
    • They write first drafts of briefs
    • They convert cases into embeddings
    • They flag negative treatment such as reversal or overruling
    • They translate opinions into plain English
  7. Which statement about retrieval-based legal AI tools is accurate?
    • Retrieval reduces fabricated citations, but tools can still misstate holdings or cite cases for points they do not support
    • Retrieval eliminates all hallucinations, so verification is unnecessary
    • Retrieval-based tools hallucinate more often than general chatbots
    • Retrieval tools only work for criminal law
  8. Under professional rules, who is responsible for the accuracy of citations in a filed brief?
    • The AI vendor that supplied the tool
    • 受理案件存檔的法院書記官
    • Opposing counsel, who must spot any errors
    • The lawyer who signs the filing