A Cursory Look at Lexis+ AI

Guest post by Justin Tung [1]

What Lexis+ AI is, and What it Promises

On Friday, November 17, 2023, Lexis sent an email announcing access to its new generative Artificial Intelligence (AI) tool called Lexis+ AI. It promises “the fastest legal generative AI with conversational search, drafting, summarization, document analysis, and linked hallucination-free legal citations.”[2] AI and legal research have had a rocky relationship since AI’s advent and rise to prominence, with instances of attorneys relying on AI to their detriment. However, given its increasing industrial and social ubiquity, many believe that AI is a tool to be refined and improved, and it’s only a matter of time before there’s widespread adoption in the field of legal practice.

Although many smaller companies have launched AI legal-research tools, Lexis, a legal research platform giant, is the first company of its size to do so. It promises to leverage the immense database of legal material it already commands and improve accessibility to it through the integration of an AI tool. Lexis also presumably has a deep understanding of what attorneys who use their platforms want and expect and would potentially be able to offer an AI tool that can meet those expectations.

This post examines an instance of an interaction with Lexis+ AI show a sample of its functionality. Hopefully, analysis of this functionality can show the priorities and abilities of the AI and assess the AI’s utility for various research needs.

(Note also that when initiated, a chat session with the AI has a drafting tool, which was not examined.)

The Prompt

When asked, “Is it legal to keep chickens in Austin, Texas?” Lexis+ AI generated the following response (formatting slightly altered for coherence):

“The following suggestions might be helpful in assisting you in your task.

The legality of keeping chickens in Austin, Texas depends on the specific location and circumstances [State ex rel. Burks v. Stovall, Ex parte Naylor, McKee v. Mt. Pleasant.] Some key points from the provided information:

– An ordinance in Houston made it illegal to keep chickens and other fowl within 100 feet of residences, though some exceptions existed for commercial purposes [State ex rel. Burks v. Stovall]. A similar ordinance in Corpus Christi allowed chickens with a permit more than 100 feet from residences, up to a limit of 12 [Ex parte Naylor].

– In Mt. Pleasant, the city sought to enjoin a resident from keeping over 100 chickens, alleging it created a nuisance [McKee v. Mt. Pleasant]. This suggests cities may restrict chickens deemed a nuisance.

– The provided sources generally indicate Texas cities can regulate keeping chickens through ordinances, including proximity to residences, permitting requirements, and nuisance restrictions. Specific regulations likely vary by city.

– For Austin specifically, the provided sources do not directly address the legality. The city’s regulations would need to be consulted to determine if chickens are allowed and any restrictions that apply.

Content Supporting AI-Generated Response:

State ex rel. Burks v. Stovall | Court of Criminal Appeals | 168 Tex. Crim. 207 | Jun 10, 1959 | Texas

Ex parte Naylor | Court of Criminal Appeals | 157 Tex. Crim. 355 | Jan 30, 1952 | Texas

McKee v. Mt. Pleasant | 6th Court of Appeals | 328 S.W.2d 224 | Aug 25, 1959 | Texas”

Critical Analysis of the Generated Response

Initial Impressions

Initially, one big red flag is the dates of these cases. Except for historical research, legal research often requires attention to the currency of authorities. Tools like citators, which are some of the most valuable tools legal research platforms can offer, exist for researchers to determine whether the authority they might rely on is still currently good law.

This may indicate a kind of mistranslation by the AI. If someone were to ask a friend or attorney, “Is it legal to keep chickens in Austin, Texas?” that interlocular would likely believe that the speaker is interested in raising chickens in Austin or is having issues with chickens being raised in Austin, e.g. by a hostile neighbor. The AI, however, has felt it appropriate to give answers to this question which are historical, and likely not representative of the current laws or even the current law’s immediate predecessor. While more recent cases are certainly available, its preference for a historical answer may be indicative of the fact that it believes that earlier cases best answer this prompt.

Analysis of the Response’s Ordinance-Based Issues

Several legal issues are mentioned in this response:

• Commercial purpose vs. noncommercial purpose
• Distance to residences
• Limit on number of chickens
• Authority of ordinances as a means of regulation
• Nuisance
• Permit requirements

While these legal issues are certainly relevant, it is unclear if these are the priorities and considerations of the opinions it cites in its response, or whether they are necessarily current considerations for current chicken keeping.

Besides the excerpt of the local ordinances in the opinion, the texts of the ordinances were exceedingly difficult to find. However, current ordinances for these jurisdictions were very easy to locate through a quick web search. They showed that in addition to number, distance, nuisance, commercial purposes, and permitting requirements, there were also many other considerations:

• Appropriate structures for confinement[3]
• Sanitation/dropping removal[4]
• Visibility[5]
• Single vs multifamily dwellings[6]
• Food storage requirements[7]
• Watering requirements[8]
• Death or illness[9]
• Sound[10]
• Coloring[11]
• Gender restrictions[12]
• Breed-specific restrictions[13]
• Clipped wings[14]

If one were to rely on this answer to distinguish considerations to account for in raising chickens, they would be significantly underinformed. However, these are the legal considerations in the case law the AI cites. Given the age of the cases, it is no surprise that they are quite different from modern ordinances. It is possible that this AI first picks the cases it believes are relevant, then analyses them and generates a response from its analysis, as opposed to determining the topics it ought to cover, then finding authorities to support it.

A slight irony here is that information about the modern ordinances could not be found on the Lexis platform itself, but rather by running a web search for the ordinances of those jurisdictions. If someone were able to do this, they could simply have just googled for the relevant ordinances in Austin, and not bother with the other jurisdictions. Discussion where citing caselaw could have been helpful— the judicial or administrative interpretation of the text of the ordinances— is largely absent from the AI’s response.

Analysis of the Response’s Other Legal Issues

The AI generated response also included other aspects of legality which the current ordinances did not address or did not address in the same way. These issues are, first, the authority of local ordinances to regulate the keeping of chickens, and second, nuisance. How the AI handles these two legal issues may show certain content and generative preferences.

Firstly, in the current year, the issue as to whether the keeping of chickens can be regulated by the authority of a city ordinance is not controversial. Except for certain complications such as a constitutional amendment by the recent “right to farm” ballot measure,[15] the authority of local ordinances in this matter is uncontested by most. While answering this question may be significant for someone doing legal historical research, an attorney who is asked by a client whether they may raise chickens should probably not respond with a discussion about the legitimacy of the authority of ordinances.

Secondly, the discussion and cases provided by the AI are not very helpful either in scope or depth. Although the current ordinances do, at times, mention nuisance, they do not go much beyond saying that the keeping of chickens may not be done in a way as to constitute a nuisance. This may reflect a deference to common law on this issue.

While cases are usually a good source of common law, in this situation, the provided cases are not. Ex parte Naylor[16] discusses nuisance, but the chicken owner was also slaughtering them, and the court found that this was a health hazard and nuisance in a residential area. McKee v. City of Mt. Pleasant[17]discusses nuisance, but the owner was running a farm operation with over 100 chickens, including game cocks. Additionally, a lot of the citations to this case are for procedural reasons, not substantive ones. State ex rel. Burks v. Stovall[18] does not discuss nuisance at all.

Content Types, Preferences, and Exclusion

When asked this question, the AI responded with only cases. This may be a limitation or a preference of the AI as well. A human researcher may recognize that certain questions are best answered or approached with certain content types. To get the AI to suggest something other than cases, it needed to be specifically prompted in this chat session. Even then, when asked, “Where can I find Austin ordinances on whether it’s legal to keep chickens?” it returned two links to the Texas Water Code, with no mention that it could not retrieve an internal link to ordinances.

The difference between the AI’s comprehension of statutes and cases is quixotic but may reflect the Lexis’ difference in comprehension between statues and cases. Whereas cases are deeply summarized, interlinked, citated, and Shepardized, statutes are typically not analyzed to such depth.

The Possible Utility of this Response

For someone who is interested in keeping chickens in Austin or is worried about keeping chickens in Austin, the response generated by this AI is categorically worse than a google search would have been. The response discusses things like an ordinance’s legal authority, which would not be of concern to someone who is looking practically into the current legality of chicken raising. However, it is important to remember that such people are not a demographic targeted to be Lexis’ client base, so it is not surprising that an answer is not tailored to them.

For a legal practitioner looking for reliable authorities, the provided cases do not offer that much utility. The largest reason is that none of them discusses the current ordinances. Any argument about the ordinances themselves made on such a basis would be tangential and attenuated at best, and irrelevant at worst.

For a legal scholar looking to investigate a topic such as the development of ordinances as a means of agricultural regulation, these cases could possibly suit the purposes, but there are likely much better ways of doing this research, such as with reliable secondary sources, than the response given by the AI.

Conclusion

While Lexis+ AI may be useful or some things, at this time, it is difficult to see how it would be better than non-AI methods of conducting legal research. While it seemed to not have any issues with hallucinations and the cases retrieved are relevant and not inaccurately represented, the AI model does not seem equipped to understand a user’s needs before generating its response. It is still a much more a search engine in a conversational wrapper than any artifice approximating intelligence.

A hallmark of all AI models has been secrecy. Part of their allure and mystique has been that one can center a prompt and receive a response that has not been generated before. However, there are always tradeoffs. A book’s contents are non-responsive to user input, but it can make strong guarantees that its contents are factually accurate. An AI like ChatGPT can very responsively generate replies, but a great portion of it is going to be made by a non-human entity, often with a dubious relationship to truth. Legal AI tools have been faced with the challenge of generating responses that are both true and responsive to the query, and there will be inevitable compromises on either front.

While Lexis+ AI is markedly worse at holding a conversation than, for instance, ChatGPT, the limitation of Lexis+ AI is exemplified by its limit on queries. After five queries, the AI tells the user to start a new session. This limit is arbitrary, and one can’t help but wonder why it exists. In the end, since the algorithm is a black box, one cannot help but think that, like a person who ducks out of a short conversation, they are hiding something.

[1] Reference Associate at the University of Texas School of Law

[2] LexisNexis, Lexis+ AI, https://www.lexisnexis.com/en-us/products/lexis-plus-ai.page (last visited Nov. 22, 2023).

[3] City of Mount Pleasant, Tex., Code of Ordinances § 97.07(A) (2022), City of Houston, Tex., Code of Ordinances § 6-35 (2022), City of Corpus Christi, Tex., Code of Ordinances § 6-154(c)(4) (2021)

[4] City of Mount Pleasant, Tex., Code of Ordinances § 97.07(C) (2022), City of Houston, Tex., Code of Ordinances §§ 6-36, 6-33 (2022), City of Corpus Christi, Tex., Code of Ordinances § 6-154(c)(7) (2021)

[5] City of Mount Pleasant, Tex., Code of Ordinances § 97.07(B)(2) (2022), City of Houston, Tex., Code of Ordinances § 6-34 (2022)

[6] City of Mount Pleasant, Tex., Code of Ordinances § 97.07(B)(3) (2022)

[7] City of Mount Pleasant, Tex., Code of Ordinances § 97.07(C)(2) (2022)

[8] City of Mount Pleasant, Tex., Code of Ordinances § 97.07(C)(3) (2022)

[9] City of Mount Pleasant, Tex., Code of Ordinances § 97.07(C)(4) (2022), City of Corpus Christi, Tex., Code of Ordinances § 6-154(c)(8) (2021)

[10] City of Mount Pleasant, Tex., Code of Ordinances § 97.07(D) (2022)

[11] City of Houston, Tex., Code of Ordinances § 6-39 (2022)

[12] City of Houston, Tex., Code of Ordinances § 6-38 (2022), City of Corpus Christi, Tex., Code of Ordinances § 6-154(c)(9) (2021)

[13] City of Houston, Tex., Code of Ordinances § 6-37 (2022)

[14] City of Corpus Christi, Tex., Code of Ordinances § 6-154(c)(3) (2021)

[15] The New York Times, Election Results for Texas Proposition 1 “Right to Farm and Ranch” (Nov. 7, 2023), https://www.nytimes.com/interactive/2023/11/07/us/elections/results-texas-proposition-1-right-to-farm-and-ranch.html.

[16] Ex parte Naylor, 157 Tex. Crim. 355 (1952).

[17] McKee v. City of Mt. Pleasant, 328 S.W.2d 224 (Tex.Civ.App.-Texarkana 1959, no writ).

[18] State ex rel. Burks v. Stovall, 168 Tex. Cr. 207, 324 S.W.2d 874 (1959).