AI w badaniach prawnych i quizie dotyczącym wyszukiwania orzecznictwa

Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.

1. Dlaczego prawnicy w sprawie Mata przeciwko Avianca (2023) zostali ukarani?
   • 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. Do czego służy generacja wspomagana wyszukiwaniem (RAG) w prawnym narzędziu badawczym?
   • 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. Dlaczego chatbot ogólnego przeznaczenia może generować przekonujące cytaty na przypadki, które nie istnieją?
   • 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. Co to znaczy, że prawna odpowiedź AI jest „błędna”?
   • 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. Który indeks słów kluczowych (taki jak BM25) jest szczególnie skuteczny w pobieraniu hybrydowym?
   • 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
   • The court clerk who accepts the filing
   • Opposing counsel, who must spot any errors
   • The lawyer who signs the filing