Skip to Main Content

AI for Legal Research: AI & Academic Writing

Practice Prompting

Lateral Reading

Professional Conduct regarding AI and like-technologies

The use of AI, including LLMs, can be beneficial in certain aspects of academic work, however, you must adhere to the Student Conduct Code when using AI. The following are violations of the Code:

  • 2.1 (L) Submitting plagiarized work in an academic pursuit. This conduct consists of the appropriation of the literary composition or other creative work of another (including an artificial intelligence source), or passages or ideas thereof, and passing them off as the product of one's own mind.
  • 2.1 (G) Consulting electronic materials or using an electronic device during an examination except as authorized by the examining professor, or consulting electronic sources (including artificial intelligence sources) during an examination. 

Plagiarism can be avoided by citing any sources (AI or otherwise), see the Avoiding Plagiarism section.

The use of artificial intelligence can be alluring for many lawyers with the "promise of efficiency and accuracy of legal services," according to the American Bar Association Resolution 112 about the ethical and legal issues about the usage of AI in the practice of law. The resolution lists 4 ethical duties:

  • 3.1 (A) Duty of Competence.
    • They are required to understand how AI technology produces results.
  • 3.1 (B) Duty to Communication
    • A lawyer's duty of communication under Rule 1.4 includes discussing with his or her client the decision to use AI in providing legal services. A lawyer should obtain approval from the client before using AI, and this consent must be informed. The discussion should include the risks and limitations of the AI tool. 
  • 3.1 (C) Duty of Confidentiality
    • The use of some AI tools may require client confidences to be "shared" with third-party vendors. As a result, lawyers must take appropriate stems to ensure that their clients' information appropriately safeguarded. Appropriate communication with the client is also necessary.
  • 3.1 (D) Duty to Supervise
    • Lawyers are obligated to supervise the work of AI utilized in the provision of legal services, and understand the technology well enough to ensure compliance with the lawyer's ethical duties. this includes making sure that the work product produced by AI is accurate and complete and does not create a risk of disclosing client confidential information.

Prompting

Prompting refers to the questions, topics, or asks posed to an AI tool. "A prompt can be as simple as a phrase or as complex as multiple sentences and paragraphs" (MIT Management, 2024). 

When using LLMs for Legal Research, chose an LLM that is designed for legal research and writing. There are a few tips to get positive interactions rather than responses that are generic or unhelpful. Users should include examples in their, ask for refinement in multiple iterations, and if possible, upload documents. Prompts for legal research should always:

  • Provide context 
  • Be specific in detail about what is needed
  • Have multiple iterations (building on the results or conversation) 

Prompting involved asking a question or giving a topic to an LLM. Many LLMs learn from user-interactions, unless otherwise stated in the data and privacy policies of companies. Some LLMs help pages offer tips on prompting such as refining different iterations. Basic principles are generally consistent across all LLMs.

To read more about prompts: https://mitsloanedtech.mit.edu/ai/basics/effective-prompts/

Zero-shot v. Few-Shot

There are two types of prompting frameworks, one is known as few-shot learning, or n-shot learning and zero-shot learning. Few-shot learning is used when data is scarce and aims to emulate the human ability to learn from a few examples or interactions. When interacting with AI models in a legal research context, providing examples of what you want or examples that are similar to expectations can help the machine learn and model what is being asked for.  

Zero-shot learning is a framework in machine learning where there are no examples to learn from, only previous observations. The machine learns to categorize based off of previous examples and knowledge rather than a set of examples. 

Prompt Iterations:

Prompt iterations involve restating and refining prompts to provide context and specific details in order to gain desired results from GenAI. Iterations often benefit from rewording, adding examples, and giving constraints. 

Alphabet, What are AI hallucinations?, Google Cloud (2024), https://cloud.google.com/discover/what-are-ai-hallucinations?hl=en.

 

Step-by-Step

A step-by-step  approach is a significant part of legal reasoning and demonstrates how the author(s) reached their conclusion. Including key phrases such as: "think step-by-step", "answer this query thinking step by step", or "show your work" in a prompt or iterations of a prompt can aid in verification and allows the user to see how the AI reached the final result. 

Source Verification

There are some widely known instances of legal practitioners using AI to assist them with legal research and creating documents that are used in court. Unfortunately, these practitioners did not verify their sources and have been cited as examples of malpractice. AI hallucinations can create false sources, statements, or veer from reality. It's important for users to verify that the sources and statements are correct. The best way to address  hallucinations is to independently verify anything that comes from an AI product, including products from Lexis, Westlaw, and other legal AI tools. Verifying sources is foundational in legal research and practice: lateral reading,

Westcheck and Shepard's Brief Check are a few tools that can be utilized. For further examples view the PowerPoint presentation created by Dean Peoples.

Turnitin.com

Turnitin is a plagiarism detection tool utilized by Oklahoma City University.  Faculty members have the option on D2L to let students submit drafts of their papers and view the plagiarism score before the final submission. 

To avoid accusations of plagiarism in essays and assignments, be sure to cite all sources, including those generated by AI. Refer to the PowerPoint for more information.