Skip to Main Content
Banner Image

Artificial Intelligence [AI] and Legal Research

This guide is intended to introduce new law students, as well as law students who need a refresher on the topic, to the concepts of researching using AI.

6 Things to Know About AI

Definitions

  1. Algorithm:
    1. A rule set that an LRAT must follow when performing calculations or other problem-solving operations.
  2. Artificial Intelligence:
    1. Extractive A.I. [Ex. AI]:
      1. An algorithm that finds, or “extracts,” answers by analyzing data points from specified datasets it has been trained on.
    2. Generative A.I. [Gen. AI]:
      1. An algorithm that analyzes both past datasets and the query presented to it to generate novel, or original, data content.
    3. General A.I. [does not exist]
      1. Anything to do with:
        1. Sentient robots
        2. Computer consciousness
        3. Eternal life
        4. Machines that "think" like humans"
    4. Narrow A.I. [does exist]
      1. A.I. that works by having algorithms analyze an existing dataset, identify patterns/probabilities in that dataset, and codify those patterns/probabilities into an LLM such that what has been analyzed, identified, and codified may be output to an LRAT to answer a query.
  3. Information Literacy:
    1. A set of abilities requiring researchers to recognize when information is needed and to have the ability to locate, evaluate, and effectively use that needed information.
  4. Large Language Model [LLM]:
    1. A technological combination of large[r] datasets, neural networks, and transformers that can be utilized for text generation by taking input from a source and repeatedly predicting the next word.
  5. Legal Research Algorithmic Tools [LRATs]:
    1. A shorthand way to refer to legal A.I. tools.
  6. Prompt Engineering:
    1. A researcher’s structuring of their search question when using an A.I. tool using (1) context clues, (2) question structuring, and (3) defined search limitations.
  7. Tokenization:
    1. Tokenization is the process of converting input text into small units, or, "tokens," such as words or sub-words. By breaking text down into tokens, AI-powered system can process, analyze, and interpret text more accurately and efficiently.
    2. Once the input text has been tokenized, the LRAT's algorithm weighs the tokens against (1) the question asked, (2) the resources available in the database, and (3) then uses the token weights from the prompt to ensure that the answer given in the response answers the question asked.