Generative AI & Legal Research

Rombauer Framework:

1. Preliminary Analysis:

Preliminary analysis is the most important step in solving any legal research problem. Here, you need to write down all the pertinent information to the problem and analyze what you know, what you need to learn, and your research plan. 

Your preliminary assessment of a problem should include the following factors:

• Identify relevant and material facts.
• Select appropriate keywords for your research,
• Identify preliminary issues and formulate search queries. 
• Identify the jurisdiction(s) involved. 
• Identify what you know about the area of the law.
• Identify possible gaps in your knowledge.

The preliminary analysis stage is the foundation for your research. Spend adequate time laying down your basis and establishing a list of key terms and research queries. Further, gather foundational information that helps you better understand the research question. This can include simple Google searches and reading secondary materials such as Cause of Actions, American Law Reports, or Law Review Articles. Now is your time to build your understanding of the situation and gain competency in the research question.

Building your bibliography

As you read through secondary sources, note any statutes, cases, or regulations mentioned in a bibliography of sources. Include the citation to the source, the currentness of the source when you found it, and a brief description of the source. This will set you up for the following steps.  

2. Search for Statutes and Regulations:

Here, you should check for relevant statutes in annotated statutory codes--if you can access annotated codes. If you find adequate secondary sources during your preliminary analysis, these secondary sources should lead you to the relevant statute. If you are unsure there is a statute on point, be diligent in searching and refer back to your secondary sources to see if you missed anything. It may be a common law issue if you cannot find any statutes on point.

Next, check for any relevant regulations. Again, these can be found during your preliminary analysis, or by citing references from statutes. Include the citations to relevant regulations in your bibliography. 

Before moving on, check the currentness of the statute to ensure you have an accurate version of the statute and/or regulation. 

3. Mandatory Precedent:

If you could use an annotated statute, use the citing references from the statute to find cases covering this area of law. Both print and online versions include citations to cases covering the statute. Westlaw further has the West Key Number system, which can be used to find similar cases covering a specific topic. You can add more cases to the list you created during your preliminary analysis using these sources. As you read through these cases, remove irrelevant cases and add cases frequently referenced by the courts. Prioritize cases from the courts of last resort (state supreme courts or the supreme court) to find the mandatory precedent. 

4. Persuasive Precedent

Persuasive authority can be found using the same research channels as mandatory precedent. Look for contrary cases from other jurisdictions, or, if there is a split in the courts, the different precedents provided by the lower courts that created the schism. As before, we include persuasive authority in our bibliography, exclude irrelevant cases, and focus on cases frequently mentioned to find the most relevant sources. 

5. Refinement and Currentness

The last step in the Rombauer process is refinement and currentness. Remove any unnecessary sources from your bibliography, and then use a citator (such as KeyNote or Shepherds) to check the currentness of your sources. This ensures that you have relevant and accurate information at the end of your research process and that none of your statutes have been changed or mandatory precedents overturned. 

Generative AI tools such as ChatGPT, Google Gemini, and Microsoft CoPilot can help with your legal research. The use of these tools neatly fits within the Rombauer Method. Further, the Rombauer Method helps prevent the possible pitfalls of AI hallucinations and increases the trustworthiness of information found using Generative AI. 

 

Take time to Think.

Generative AI is an excellent research tool that can increase your research efficiency, but it is still imperative to take a few minutes to locate your key terms and access your base knowledge. This will help you formulate prompts to ask the AI and have a structure for your research. 

I recommend thinking of AI similarly to Google. AI is an excellent tool for gaining contextual knowledge; however, secondary sources, such as Causes of Action, American Law Reports, and American Jurisprudence, can provide superior in-depth analysis and explanation for complex legal issues. 

Imagine you had a question about how World War I started. You open the Wikipedia article, read it, and now have a basic understanding of the events, major players, and timeline. Compare this with reading a treatise by a Historian explaining the international diplomatic relations that set up the tensions before World War One. Without the contextual knowledge from the Wikipedia article, much of the book would seem nonsensical and difficult to digest. The Wikipedia article laid out the contextual information you needed to understand the treatise, and the treatise then explained the minutiae of a complex historical event. 

AI and legal secondary sources have a similar relationship. AI can provide an overview of an area of law, while the secondary sources dive into the subtleties of the matter. Thus, utilizing AI is a supplemental step in your research process, not an all-out replacement for good research methods. 

 

How AI can help with legal research:

During your preliminary analysis, you can use AI to increase your efficiency and understanding of a topic in several ways, for example:

• Brainstorming other relevant issues (preliminary analysis)

• Confirming understanding of basic terms and ideas (any point)

• Looking at possible arguments and counterarguments (preliminary analysis/refinement)

• Context and other foundational matters (preliminary analysis)

• Generate more key terms and possible queries (preliminary analysis/refinement)

• Find relevant statutes and cases (preliminary analysis/finding precedent/refinement)

This list is not exhaustive, and considering how to use AI to help with your legal issue is part of the first step in performing your preliminary analysis. Legal issues are diverse, and research questions can be simple or complex. AI can help you navigate the process, but only if used well. 

 

Brainstorming:

If you are stuck on approaching a legal problem, you can interface with the AI and ask it queries to help you brainstorm possible approaches to the problem, issues that need to be raised, and ideas for a course of action. It is important to note that AI systems can be accessed by others and are not secure. Client confidentiality must be maintained. Do not place any personal information about a client or their case into any AI system--ever–period. Instead, use legal concepts and general questions.

Here is an example of brainstorming using Google’s Gemini program:

Prompt: 

Help me brainstorm a research problem. I need to learn about premise liability in Washington State. I need to know what duty of care a private university owes to visiting students from a high school. 

Response:

Brainstorming research problem: Premise liability of private universities in Washington towards visiting high school students

Main topic: Duty of care owed by a private university in Washington State to visiting high school students.

Possible angles:

1. Scope of the duty of care:

• Based on student status:

  • Are high school students considered invitees, licensees, or trespassers?

  • Does the standard of care change depending on the specific circumstances (e.g., attending a tour, participating in a competition, unsupervised exploration)?

• Nature of the dangerous condition:

  • Does the type of hazard (e.g., broken stairs, faulty equipment, criminal activity) affect the university's duty of care?

• Foreseeability of harm:

  • Was the university aware of, or should it have been aware of, the potential for harm?

  • Did the university take reasonable steps to mitigate the risk?

[rest of response omitted]

 

Here, the AI responded by offering factors to consider, key terms to look into–invitees, licensees, or trespassers– and elements of premise liability to investigate. Further, it asks you questions for consideration. Using AI to brainstorm can help you take complex ideas and find specific areas to focus your attention and research. 

 

Confirming understanding of basic terms and ideas:

Throughout your research process, you can use AI to ensure you understand legal terms, ideas, and principles. Generative AI is a quick resource for looking up unknown terms and ideas. In the same way, you might have previously Googled a word to get its definition; now, you can ask a Generative AI for that exact term, with examples of its use and explanations. This is an invaluable asset for building understanding and knowledge while researching and is something that researchers should utilize frequently to confirm their competency of the legal issue. 

For example:

Prompt:

What does the term licensee mean for premise liability in Washington State?

Response (ChatGPT):

In Washington State, a licensee refers to someone who enters another person's property for their purposes, with the property owner's permission, but without necessarily conferring any benefit to the property owner. In terms of premises liability, the duty of care owed to licensees by property owners is generally to warn them of known dangers that are not obvious and that the licensee is unlikely to discover on their own. This duty requires the property owner to take reasonable steps to ensure that licensees are aware of potential hazards on the property.

 

This may or may not be the technical legal term (something that must be verified with either statute or case law), but it provides a simple explanation and contextual information that can help with your research. 

You can use AI to test your understanding of new terms, ideas, or concepts. This is especially helpful if you have a new legal topic that you need to learn. AI can quickly provide simple explanations for new terms and ideas–that MUST be verified throughout your research– helping speed up the research process.

 

Developing Arguments and Counterarguments:

As you research, you should consciously think of possible arguments --and counterarguments-- that can be made for/against your issue. As a subset of your research notes, it is important to keep a list of possible arguments and counterarguments you will refine once you fully understand your issue. AI can help with this process by Generating possible arguments/counterarguments. Again, these arguments may not be the most sophisticated or even the best, but they provide a foundation for consideration and inspection. 

For example (Using Microsoft's Copilot) 

Prompt:

During a campus visit, a high school student slipped on ice at a private university. The student broke his ankle from the fall. The school had laid salt on the sidewalks and stairs and shoveled but did not salt the area where the boy fell. The school had warning signs that it might be slippery. The student who slipped was on a part of the campus where he was not supposed to be, as instructed by the tour guide. Write three legal arguments on why the university is liable for the student's injury and three counter arguments for why the school is not liable to the student.

Response:

Certainly! Let’s examine the legal arguments for and against the university’s liability in this scenario:

Arguments for University Liability:

1. Premises Liability:

   • The university has a duty to maintain a safe environment for visitors, including prospective students during campus tours.
   • By not salting the specific area where the student fell, the university may have breached this duty.
   • The presence of warning signs does not absolve the university from its responsibility to ensure safety.

2. Foreseeability:

   • The university should have reasonably foreseen that icy conditions could pose a risk to visitors.
   • Since they took precautions elsewhere (salting sidewalks and stairs), it suggests awareness of potential hazards.
   • Failing to address the specific area where the student slipped may be considered negligence.

3. Invitee Status:

   • The injured student was an invitee, meaning they were on campus with the university’s permission.
   • Invitees are owed a higher duty of care compared to trespassers.
   • The university’s failure to address the icy spot where the student fell could be seen as a breach of this duty.

Counter Arguments Against University Liability:

1. Assumption of Risk:

   • The student knowingly entered an area where they were not supposed to be, as instructed by the tour guide.
   • By doing so, the student assumed some level of risk associated with their actions.
   • The university may argue that the student’s own choices contributed to the injury.

2. Contributory Negligence:

   • The university might claim that the student’s decision to disregard instructions and venture into restricted areas contributed significantly to the accident.
   • If the student’s actions were reckless or negligent, it could reduce the university’s liability.

3. Limited Duty Outside Designated Areas:

   • The university’s duty of care may be limited to designated tour areas.
   • If the icy spot was outside the usual tour route, the university might argue that they were not responsible for maintaining it.
   • The student’s presence in an unauthorized area could weaken the university’s liability.

These arguments provide a good framework for your research, points to investigate, and areas of the law to consider. As you learn more through your research, you can refine your questions and have the AI generate more arguments for consideration. 

 

Context and other Foundational information:

Perhaps Generative AI's most useful function is its ability to provide context. This is especially useful for situations where you may know the law but not know anything about the subject. For example, you could have a solid understanding of patent law but have no idea what the technical terms in a patent mean. You can use AI to find definitions, straightforward explanations, and examples to understand the subject matter better.

Here is an example using ChatGPT:

Prompt:

Pretend you are an expert in mechanics. Explain how a pully system works for a high school class. Include examples.

Response:

Absolutely! Let's delve into the fascinating world of pulley systems.

Imagine you have to lift a heavy object, like a crate filled with books, to a higher floor. Doing it by hand would be quite difficult, right? That's where a pulley system comes to the rescue!

At its core, a pulley system consists of a wheel with a groove around its circumference and a rope or belt that fits into this groove. The wheel can either be fixed in place or able to move. Now, let's break down how it works using some examples:

1. **Fixed Pulley:**
   - Imagine a flagpole. When you tie a rope to the flag and pull down, you're using a fixed pulley system. The pulley is attached to a stationary point, like the top of the pole. As you pull down, the pulley redirects the force you apply upward, making it easier to lift the flag.

2. **Movable Pulley:**
   - Now, let's think about lifting that crate of books using a movable pulley system. You attach one end of the rope to the ceiling and pass it through a pulley attached to the crate. When you pull down on the other end of the rope, the pulley attached to the crate moves upward, lifting the crate with it. Here, you're effectively halving the force needed to lift the crate because the weight is distributed between you pulling down and the pulley system lifting up.

3. **Compound Pulley:**
   - Combining multiple pulleys can further increase mechanical advantage. For instance, if you have two movable pulleys attached to the crate and one fixed pulley attached to the ceiling, you'd have a compound pulley system. This setup can make it even easier to lift heavy loads as each additional pulley increases the mechanical advantage.

In essence, pulley systems allow us to trade distance for force. You may have to pull the rope a longer distance, but you're applying less force to achieve the same work. This makes tasks like lifting heavy objects much more manageable.

By understanding the principles behind pulley systems, you can solve various real-world problems, from lifting weights in a gym to operating complex machinery in industries.

You will need to confirm this information independently, but if you are trying to gain a foothold understanding of the subject matter, AI can help you gather the basic information necessary to understand the topic at hand without needing to look in multiple places. 

Finding Statutes and Case Law:

Not all AI are created equal when finding statutes and case law. Generative AIs are prone to hallucinating case law. Google Gemini and Microsoft CoPilot provide links to cited cases you can immediately validate, while ChatGPT does not. If you have access, Lexis+AI is the most reliable AI for finding case law and statutes. Note that currently, all Gonzaga Law students have access to Lexis+AI. If you cannot access Lexis+AI, I suggest using CoPilot or Gemini for finding cases. For statutes, ChatGPT will provide statute numbers you can independently verify online either using a paid service such as Westlaw or Lexis or on the state legislature website. 

Further, Generative AI does not currently have a way to evaluate the relevance and usefulness of a case or statute in a specific area of law. For research purposes, I suggest using legal secondary sources to build your bibliography through frequently cited sources and key terms. Lexis+AI helps find cases on the Lexis database. However, it may be more useful to use Lexis+AI to find secondary sources, and from there build your bibliography. 

Here is an example using Lexis+AI:

Prompt:

List cases from the Washington Supreme Court on Premise Liability and attractive nuisance.

Response:

Meyer v. GE Co., 46 Wn.2d 251: The court held that the ditch did not qualify as an attractive nuisance because it had natural characteristics and no hidden dangers. The court found the primary duty was on the parents and a watercourse does not become an attractive nuisance due to parental neglect.

Deffland v. Spokane Portland Cement Co., 26 Wn.2d 891: The court held the property owner did not have a duty to the child under the attractive nuisance doctrine because the essential elements were not established. The child was a licensee and the property owner only owed a duty not to willfully/wantonly injure.

Degel v. Majestic Mobile Manor, 129 Wn.2d 43: In Degel v. Majestic Mobile Manor, the court held that the natural bodies of water exception to attractive nuisance does not apply when a child is invited onto the landowner's property. The court found the landowner owed a duty of reasonable care to make the property safe for invited child guests.

Sherman v. Seattle, 57 Wn.2d 233: In Sherman v. Seattle, a 3-year-old child was severely injured by a lift device operated by the city on a dam project site. The court held the attractive nuisance doctrine did not apply because the child was not attracted to the lift itself, but the city still owed the child a duty of reasonable care regardless of the child's status on the property, since the presence of young children near the lift was foreseeable.

Note how each case has a summary and links for verification. However, whether these are all the relevant cases is yet to be determined. This does provide a way for researchers to find cases that they can use to find more cases through citing resources and other tools. 

Refinement and Verification

Do not trust the information you gathered from AI. Verify everything. Verification is easy if you use AI as a legal research tool instead of a source of information. You can verify your context, facts, and understanding gleaned from AI as you find statutes, case law, and quality secondary sources. Just as you would check your sources' currentness and reliability when performing traditional legal research, you must do the same with AI. Again, think of the Wikipedia v. Academic treatise analogy. A Wikipedia article is helpful but not a reliable source for a citation in an academic paper. Just as you would test the facts provided by the Wikipedia article with the information from the peer-reviewed treatise, you must do the same with information gathered from an AI with professional legal materials. 

Final thoughts:

Generative AI is a research tool, not a substitute for quality research methods. Use AI to supplement your research process, but do not rely on it as a substitute for secondary sources, treatises, or citators. AI can help you understand materials, build context, reinforce understanding, develop arguments, and more. Still, it does not have all the nuanced information you need to comprehend a legal matter fully.