It’s a pleasant Sunday morning and you decide to go out for coffee. Unsure of what cafes are open, you ask your Alexa device for some guidance. You pop the name of the cafe into your phone’s navigation system and begin your stroll. After ordering coffee, you unlock your phone using facial recognition to access your touch-to-pay app. Later, your social media is flooded with pictures of latté art and even promotions for the cafe you just visited. All of these are examples of artificial intelligence (AI) that individuals frequently use every day.
AI represents a class of computer products that provide data interpretation in multiple public- and private-sector settings, including health care, social media, and the legal profession. AI presents unique opportunities for data analysis that is faster and often more accurate than human decisions, opening endless opportunities in multiple settings. If there is any question about the potential force of AI in the legal community, one AI program recently passed the Uniform Bar Exam without human intervention, raising the distinct possibility that AI will “outthink lawyers,” even those with many years of training and experience.1
As a starting point, the discovery and admissibility of AI-based evidence raise ethical obligations for attorneys under the Wisconsin Rules of Professional Conduct. Under the rules,2 lawyers must have the requisite skill and knowledge, including understanding the benefits and risks of the technology involved, to perform the tasks (either by themselves or in collaboration with an experienced counsel or consultant) involving AI, such as 1) assisting clients in identifying sources (including custodians) of relevant electronically stored information (ESI); 2) engaging in meaningful meet-and-confer sessions with opposing counsel concerning an eDiscovery plan that targets AI as a data source; and 3) advising clients about the proper method to preserve, process, and collect responsive ESI in a manner that preserves the integrity of ESI for evidentiary purposes.3
AI also presents strategic opportunities when considered in the context of discovery and modern trial practice. As with all technology, AI is grounded in specific technical features that will require the law to adjust – law that is in a state of flux given the advantages and disadvantages associated with AI technology as it continues to evolve and expand. This two-part article explains AI, its unique technical features, and the legal principles that frame its discovery and admissibility. The article concludes with tips and strategies for lawyers who are confronted with discovery and admissibility issues involving AI and its various applications.
What is AI?
AI is a technology that enables computers to simulate human intelligence and problem-solving capabilities. At the most basic level, AI comprises algorithms that “learn” and solve problems that require human intelligence by applying “inferential reasoning, decision-making based on incomplete or uncertain information, classification, optimization, and perception.”4 Machine learning is an AI procedure in which a “machine has been ‘trained’ through exposure to a large quantity of data and infers a rule from the patterns it observes.”5 Machine learning algorithms can sift through enormous amounts of data, identify patterns, and draw conclusions with minimal human intervention.6
Timothy D. Edwards, Wayne State 1989, is the owner of Edwards ESI LLC, Fitchburg, where he provides e-discovery consulting services and litigates construction, employment, and business disputes. As an adjunct lecturer at the University of Wisconsin Law School, he teaches electronic discovery, civil procedure, and pre-trial advocacy. He is a member of the State Bar of Wisconsin’s Intellectual Property & Technology Law Section and Labor & Employment Law Section. He is a Fellow of the Wisconsin Law Foundation.
Hayley Rich, U.W. 2024, is an attorney in the litigation practice at Godfrey & Kahn S.C., focusing primarily on complex civil litigation. While in law school, she was a clinical student with the Restorative Justice Project and represented the law school as a competitor at the 31st Willem C. Vis International Commercial Arbitration Moot in Vienna, Austria. She also competed in New York at the 31st Annual Duberstein Bankruptcy Moot Court Competition. She is a member of the State Bar of Wisconsin’s Young Lawyers Division.
By definition, AI is a type of ESI, drawing from “the hypothetical ability of a computer to match or exceed a human’s performance in tasks requiring cognitive abilities, such as perception, language understanding and synthesis, reasoning, creativity, and emotion.”7 One commentator explained the breadth of AI as follows:
“We use search engines to find much of the information we need for work and leisure, navigate our way to work using Waze or Google Maps, bank electronically without even the thought of entering an actual bank, instruct voice-activated personal assistants like Alexa or Siri to help us in countless ways, and socialize online without the inconvenience of having to actually be social. Soon, we hear, our cars will be driving themselves, and it is only a matter of time before airplanes will be able to fly themselves from one place to another without the need for human pilots. Software applications, powered by seemingly omniscient and omnipotent ‘artificial intelligence’ algorithms, are used to diagnose and treat patients, evaluate applicants for employment or promotion, determine who is a good risk for a bank loan or credit card, determine where police departments should deploy officers to most effectively prevent and respond to crime, recognize faces in a photograph or video and match them to a real person, forecast which offenders will recidivate, and even predict an attorney’s chance of winning a lawsuit by analyzing data gathered about the presiding judge and opposing counsel.”8
Despite its current popularity, AI is not necessarily a “new” technology. In the ESI community, attorneys and IT support personnel have already harnessed the power of AI, primarily in information governance and ESI processing. “In addition to the use of data analytics and technology-assisted review in electronic discovery, …, machine-learning technologies have also been used for contract management and for due-diligence reviews in mergers and acquisitions, for public disclosure analytics, for natural-language legal research inquiries, for legal brief analytics, for drafting of legal memoranda and pleadings, for litigation forecasting for purposes of litigation funding, for review of legal billing, and even in bots employed to analyze claims and to complete forms to improve access to justice.”9 AI software is even used to analyze opposing counsel or judges and for online adjudication.10
Editor’s Note: Covering Artificial Intelligence in the Law
By Joe Forward
From the telephone (1870s) to the personal computer (1970s), from the internet (1990s) to smartphones (2000s), technological advances will always test the bounds of human innovation.
Today, generative artificial intelligence (GenAI) is the next big thing. Not sure what it is? You may be using it already without even knowing it. Any new technology may bring skepticism and, perhaps, a little anxiety. “Lawyers working with AI will replace lawyers who don’t work with AI,” said Erik Brynjolfsson, director at Stanford’s Digital Economy Lab. But new technology also means progress.
“AI can streamline routine tasks in pro bono cases, making legal services more accessible. Chatbots, for instance, can provide basic legal information to those who otherwise couldn’t afford it,” said Dennis Kennedy, director of the Michigan State University Center for Law, Technology & Innovation. What is critical is understanding new technology and how to use it.
And lawyers have an ethical duty to do so. SCR 20:1.1, comment [8]: “To maintain the requisite knowledge and skill, a lawyer should keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology, engage in continuing study and education and comply with all continuing legal education requirements to which the lawyer is subject.” That’s why Wisconsin Lawyer will be running a series of articles on “AI in the Law” over the next year – to help lawyers understand the benefits and risks of using AI in their law practices.
If you’re like me, knowing where to start on complex technology can be daunting. We hope the series of articles – combined with CLE programming – will give lawyers the information they need to help improve the quality and efficiency of their legal services while staying competitive.
Joe Forward, Saint Louis Univ. School of Law 2010, is State Bar of Wisconsin director of communications and editor of Wisconsin Lawyer magazine. jforward@wisbar.org
AI Presents Many Advantages Over Human Data Processing
Instead of automating manual tasks, AI performs frequent, high-volume, computerized tasks. AI adapts through progressive learning algorithms that allow the data to do the programming by pinpointing structure and regularities in data so that algorithms can acquire skills. AI analyzes more and deeper data using neural networks that have many hidden layers. Given these advantages, AI applications generate useful information in multiple settings that will necessarily involve attorneys. For example, financial services institutions have used algorithms to divide consumers along class lines for decades.11 The use of a person’s debt, credit history, and other characteristics that influence employment and financial decisions will continue as AI develops in the financial sector.12
In the area of health-care fraud, data has been examined for repetitive patterns of billing irregularities and analysis of relationships among 200 million electronic claims records.13 The unusual behaviors are identified through such tools as statistical models and mathematical algorithms, allowing an entire population of data to be analyzed. If employers initiate such a study for internal purposes, they may be required to produce the findings in response to a plaintiff’s request for the production of documents in an action for fraud or breach of fiduciary duty.14
AI also has multiple applications in the public sector that raise similar questions of reliability. In 2016, the Wisconsin Supreme Court in State v. Loomis upheld the use of AI analytics in the sentencing process.15 The case dealt with a defendant’s access to algorithm methodologies that were used to predict recidivism assessments. The defendant argued that this lack of access deprived him of due process to challenge whether the algorithm wrongfully considered his race and gender and other factors.
There are concerns that when algorithms rely on data from the criminal justice system, the data at issue then likely reflects a history of bias and discrimination, which is then reinforced by the computational processes of the algorithm.16 Because of this potential for learned discrimination, it is argued that the capacity to challenge the algorithm is a due-process right that requires a transparent explanation of the algorithm.17
The supreme court ultimately upheld the use of the program but with certain limits – an independent rationale must accompany a sentencing decision. This means that, as a matter of due process, the algorithm alone cannot be used to determine whether someone should be incarcerated or to calculate the length of a sentence. While limited in scope, Loomis illustrates a proper use of purportedly reliable algorithmic analytics in a situation in which the AI was not the sole source of a predictive assessment for sentencing purposes. For purposes of due process, corroboration and independent evaluation will fill the gap.
In most cases, the data upon which AI relies will be found in the creation and application of its underlying code. As noted in an article by Paul Grimm, very little is known about the data that AI applications are fed, how they are developed and trained, or whether they produce consistently accurate results.18 While some AI applications are “trained” using supervised machine learning, others are self-taught through unsupervised machine learning, and there are still others that use reinforcement learning.19 “Some [AI applications] can be differentiated by what they are programmed to do, such as classifying or ranking data by its value or relationship to other data, versus others, which do regression analysis by attaching specific values or weight to data in a large data set.”20
Disadvantages to ESI in the AI Context of Discovery
There are subtle disadvantages to ESI that provide pressure points for effective advocacy in discovery and at trial. As noted above, bias can affect AI systems, including how the outputs are interpreted and used by humans. Because machine-learning algorithms rely on historical data, they can replicate, perpetuate, and amplify the biases and discriminatory practices they are intended to prevent.21 For example, a chatbot that is meant to assist, inform, and resolve issues may unintentionally frustrate, misinform, and harm the end user if the technology does not consider the user’s cultural barriers and preferences.22
Bias in data can occur because the training data is not representative of a target population to which the AI system will later be applied.23 In addition, bias can be found in the code itself, which often reflects the subjective input of the person who is writing it, including potential biases.
As highlighted in the Grimm article, even though most AI tools place a great emphasis on achieving predictive accuracy and efficiency, they do not always consider “statistical or demographic parity, the distribution of false positives and false negatives, or other measures of fairness and bias, which arises as a result of the human input and interpretation of AI systems and their output.”24 Confirmation bias is the tendency for humans to search for, interpret, favor, and recall information that confirms their prior beliefs and values.25 Automation bias is the tendency for humans to favor results from automated decision-making systems and to ignore or discount contradictory evidence generated separately from such systems, even if it is correct, because they believe that the automated decision-making system is somehow more “trustworthy” or “objective.”26
A second serious concern with algorithms and their outputs is the lack of proper evaluation of many commonly used AI systems.27 Because of deficiencies in accuracy, validity, and reliability, many AI tools would not pass muster if they were subjected to the scientific method. In this context, validity is the quality of being correct or true, in other words, whether and how accurately an AI system measures (that is, classifies or predicts) what it is intended to measure.28 Reliability refers to the consistency of the output of an AI system.29 In other words, this is whether the same result is obtained under the same set of circumstances.30 Function creep refers to the gradual widening of the use of a technology or system beyond the use for which it was originally intended, often, but not always, without validation.31 These variables must be considered in the evaluation of AI data for purposes of discovery and admissibility.
Although there may never be an end to the philosophical debate comparing human intelligence to AI, there is no question that AI is here to stay, raising complicated questions concerning the preservation, production, and admissibility of AI data that present challenges and opportunities for litigation attorneys. To meet these challenges, attorneys must understand how AI fits into the discovery process and the best strategies for accessing and admitting relevant AI for motion practice and trial.
AI, ESI, and Discovery
AI is driven by its computer code and the underlying data it processes. Under normal circumstances, such information would be produced in response to a request for the production of documents under Rule 34 and Wis. Stat. section 804.09(1) (part of the Wisconsin Rules of Civil Procedure).32 The underlying code constitutes “electronically stored information,” and Rule 34 permits a requesting party to “inspect, copy, test, or sample … any designated … electronically stored information – including writings, drawings, graphs, charts, photographs, sound recordings, images, and other data or data compilations – stored in any medium from which information can be obtained either directly or, if necessary, after translation by the responding party into a reasonably usable form.”33 In this context, the traditional discovery rules pertaining to the production of documents and ESI will dictate the outcome of any dispute concerning document requests.
The analysis changes when AI is, legally speaking, a party.34 While a request for the production of ESI is not a mental examination, to require a party reliant upon a machine for thinking to submit the machine for discovery is in effect demanding that the party’s mental processes be examined.35 In this situation, having a party that is reliant on a machine for responses to fundamental questions raises proportionality concerns. Rule 26 and its counterpart, Wis. Stat. section 804.01(2), allow parties to obtain discovery “regarding any nonprivileged matter that is relevant to any party’s claim or defense and proportional to the needs of the case….”
A simple case of copyright infringement involving AI as the creator of the allegedly infringing work provides an excellent example because the core issue would be the AI’s decision-making process (as traced to the underlying code, which may reveal the state of mind of the person who created it) and whether the AI had ever encountered the infringed work in its mode of creation. Here, a request to produce ESI may be burdensome and disproportional given the needs of the case. Accordingly, the party opposing the production of AI must meet its initial burden of showing that a given request is disproportional to the needs of the case, which should include “an estimate of the number of documents that it would be required to provide …, the number of hours of work by lawyers and paralegals required, [or] the expense.”36
Arguably, a request for the production of ESI from an AI application also constitutes a mental examination. “While a request for the production of ESI would not be styled as a mental examination, to require a party reliant upon a machine for thinking to submit the machine for discovery is in effect demanding that the party’s mental processes be examined.”37 Under Rule 35 and Wis. Stat. section 804.09(1), a court may order a party whose mental condition is in controversy to submit to a mental examination by a certified examiner upon a showing of good cause.
However, a requirement for a mental examination is that the condition of the party be in controversy. “This then leaves a gap: A standard document production request for ESI could be objected to as an oppressive or unduly burdensome inquiry into the AI’s ‘mind’; yet characterizing the request as a request for mental examination under Rule 35 would fail when the mental state of the AI is not in controversy.”38 Lawyers should consider these nuances when planning discovery in cases involving AI.
If it is not possible for lawyers to obtain and examine ESI from an AI tool, it might be possible to “depose” the AI tool and permit the AI tool to answer direct questions that are relevant to the case at hand. Rule 30 and Wis. Stat. section 804.05 allow a party to, “by oral questions, depose any person, including a party, without leave of court….” Similarly, it is possible to discover information from AI through interrogatories or written depositions. Depositions provide the requesting parties with the opportunity to see or hear the AI’s information and draw appropriate conclusions.
A final question concerns the preservation of AI when litigation is reasonably foreseeable. In an appropriate case, AI must be identified and preserved.39 Therefore, proportionality concerns should be considered in the timely preservation of AI, which should be accomplished with an appropriately drawn and executed written litigation hold.40 Arguments related to AI should be developed early in the case and presented to the court in a timely fashion so that discovery continues to move forward without unnecessary obstacles.
Part 2 of this article will appear in the December Wisconsin Lawyer. That article will address the evidentiary issues that accompany the admissibility of AI, with tips for attorneys and judges who are confronted with such questions.
Also of Interest
The Source for Wisconsin Discovery Law
With Wisconsin Discovery Law and Practice, attorneys can tap into the knowledge and experience of some of Wisconsin’s most successful litigators and become exceptional in the realm of discovery. Learn more about the scope of discovery; the functions, advantages, and disadvantages of formal and informal discovery; how to modify discovery procedures; the effects of the general requirement of relevance on actions and cases; the role of privileges and immunities in limiting the nature and extent of discovery; and the application of the doctrine of waiver and forfeiture in a discovery setting.
Discovery Law and Practice is packed with hundreds of useful examples, practice tips, instructions, and sample questions addressing the key facets of discovery, beginning with initial discovery and including interrogatories; inspection of documents, places, and things; and judicial supervision and enforcement. The book’s 10 chapters, sample forms, and thorough index will help you clarify discovery law in Wisconsin.
Wisconsin Discovery Law and Practice is an efficient guide to the discovery process in Wisconsin. For those who spend only part of their time in litigation, the book will become a valuable reference as you work your way through your case. For those whose entire practice deals with litigation, the book will become your guiding authority for discovery in Wisconsin.
https://marketplace.wisbar.org/products?pid=AK0042
Endnotes
1 Kevin Dollear, GPT-4 Passes the Bar Exam, Ill. Inst. of Tech. (March 15, 2023), https://www.iit.edu/news/gpt-4-passes-bar-exam.
2 SCR 20:1.1; State Bar of Cal. Standing Comm. on Pro. Resp. & Conduct, Formal Op. No. 2015-193 (outlining competence requirement in ESI context), https://www.calbar.ca.gov/Portals/0/documents/ethics/Opinions/CAL 2015-193 [11-0004] (06-30-15) - FINAL.pdf.
3 The Sedona Conference et al., The Sedona Conference Commentary on ESI Evidence & Admissibility, Second Edition – A Project of The Sedona Conference Working Group on Electronic Document Retention and Production, 22 Sedona Conf. J. 83, 222 n.238 (2021).
4 Yavar Bathaee, The Artificial Intelligence Black Box and the Failure of Intent and Causation, 31 Harv. J.L. & Tech. 889, 899 n.37 (2018), https://jolt.law.harvard.edu/assets/articlePDFs/v31/The-Artificial-Intelligence-Black-Box-and-the-Failure-of-Intent-and-Causation-Yavar-Bathaee.pdf.
5 Joshua A. Kroll et al., Accountable Algorithms, 165 U. Pa. L. Rev. 633, 639-40 (2017), https://scholarship.law.upenn.edu/penn_law_review/vol165/iss3/3/.
6 See One to One Corporate Finance, Artificial Intelligence to Enhance M&A, https://www.onetoonecf.com/embracing-artificial-intelligence-to-enhance-mna/ (last visited Oct. 2, 2024).
7 See A.M. Turing, I. – Computing Machinery and Intelligence, 59 MIND 433, 460 (1950).
8 Paul W. Grimm et al., Artificial Intelligence as Evidence, 19 Nw. J. Tech. & Intell. Prop. 9, 10-12 (2021), https://scholarlycommons.law.northwestern.edu/cgi/viewcontent.cgi?article=1349&context=njtip.
9 Id. at 34-35.
10 Id. at 36.
11 Id. at 14.
12 Id. at 54-55.
13 Colin Caffrey, Can a Computer Read a Doctor’s Mind? Whether Using Data Mining as Proof in Healthcare Fraud Cases is Consistent with the Law of Evidence, 30 N. Ill. U. L. Rev. 509, 509-10 (2010) (explaining application to detect health-care fraud), https://huskiecommons.lib.niu.edu/niulr/vol30/iss3/3/.
14 Price Benowitz Accident Injury Lawyers LLP, Gathering Evidence in a DC Breach of Fiduciary Duty Claim, https://pricebenowitz.com/dc-civil-litigation-lawyer/breach-of-fiduciary-duty/gathering-evidence/ (last visited Oct. 2, 2024).
15 State v. Loomis, 2016 WI 68, ¶ 8, 371 Wis. 2d 235, 881 N.W.2d 749.
16 Carla L. Reyes & Jeff Ward, Digging into Algorithms: Legal Ethics and Legal Access, 21 Nev. L.J. 325, 334-35 (2020), https://scholars.law.unlv.edu/nlj/vol21/iss1/8/.
17 Id.
18 Grimm, supra note 8, at 12.
19 Id.
20 Id.
21 Jeannie M. Paterson, Misleading AI: Regulatory Strategies for Algorithmic Transparency in Technologies Augmenting Consumer Decision-Making, 34 Loy. Consumer L. Rev. 558, 559 (2022), https://lawecommons.luc.edu/cgi/viewcontent.cgi?article=2099&context=lclr.
22 Sherley E. Cruz, Coding for Cultural Competence: Expanding Access to Justice with Technology, 86 Tenn. L. Rev. 347, 350-51 (2019), https://www.ssrn.com/index.cfm/en/u-tennessee-leg/ (find article by using “Search” function).
23 John G. Browning, Real World Ethics in an Artificial Intelligence World, 49 N. Ky. L. Rev. 155, 168 (2022).
24 Grimm, supra note 8, at 45.
25 Id. at 46.
26 Id. at 47.
27 Id. at 48.
28 See Roberta Heale & Alison Twycross, Validity and Reliability in Quantitative Studies, 18 Evid.-Based Nurs. 66 (July 15, 2015), https://pubmed.ncbi.nlm.nih.gov/25979629/.
29 Id.
30 Id.
31 Grimm, supra note 8, at 51.
32 For purposes of this article, the “Rules of Civil Procedure” or the “Rules” refers to the Federal and Wisconsin Rules of Civil Procedure, which are similar in many respects. Federal decisions interpreting the Wisconsin Rules of Evidence constitute persuasive authority in connection with a court’s construction of Wisconsin’s correlating rule. See, e.g., State v. Schindler, 146 Wis. 2d 47, 54, 429 N.W.2d 110 (Ct. App. 1988).
33 See also Wis. Stat. § 804.09(1).
34 See Jason Chung & Amanda Zink, Hey Watson Can I Sue You for Malpractice? Examining the Liability of Artificial Intelligence in Medicine, 11 Asia Pac. J. Health L. & Ethics 51, 51, 67 (2018) (“Courts have traditionally deemed it impossible for machines to have legal liability as they are not legal persons….”), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3076576.
35 Yvette J. Liebesman & Julie C. Young, The AI Author in Litigation, 69 U. Kan. L. Rev. 103, 117 (2020), https://scholarship.law.slu.edu/cgi/viewcontent.cgi?article=1556&context=faculty.
36 Kleen Prods. LLC v. Packaging Corp. of Am., No. 10 C 5711, 2012 WL 4498465, at *1, *7, *9-10 (N.D. Ill. Sept. 28, 2012) (unpublished).
37 Liebesman & Young, supra note 35, at 117.
38 Id.
39 1 Arkfeld’s Best Practices Guide: IT § 1.3 (2024).
40 Michael Hamilton, Proportionality in Preservation, Bloomberg L., https://news.bloomberglaw.com/us-law-week/proportionality-in-preservation (Feb. 15, 2017).
» Cite this article: 97 Wis. Law. 10-16 (November 2024).
