From hiring to contracts, AI’s use in legal departments is increasing. But that also means planning for new types of risks.
In our previous article, we explored several legal implications that artificial intelligence will have on patent law, and the availability of patent protection for AI inventions. In this article, we explore the impact of AI in the legal industry, including new AI tools for legal departments, and how to plan for risk when using these AI tools.
AI in the Legal Sector
Machine learning is an application of AI in which AI’s algorithms learn from past experiences and then apply this knowledge to predict future outcomes.
Because there are many similarities between the law and machine learning, the law is conducive to AI and its machine learning applications. For example, both the law and AI machine learning infer rules from historical examples to apply to new situations. Legal rulings involve applying propositions based on prior precedent to the facts at issue and deriving an appropriate conclusion. AI machine learning uses the same process. The law and AI machine learning are both logic-oriented methodologies (e.g., if X happens, then the result should be Y).
Natural language processing (NLP) is another application of AI in which the AI’s algorithms automatically process and interpret words based on the context in which the words are used. For example, rather than processing a word in isolation, NLP processes the word based on the other words used in the same phrase or sentence in which the word appears, and the topic or application in which the word is used. This is similar to law that requires attorneys to analyze terms in a contract or identify facts of a case that is similar to a case at issue.
Common Uses for AI Tools
Machine learning and NLP have enabled a number of AI tools to be developed to help legal departments reduce costs, develop data-driven strategies, assess risk, and become more productive. Below, we identify some of the AI tools that are available to legal departments.
Contract Review and Negotiation: Many legal departments spend a significant amount of time on contract review and negotiation. Contracts often have standard terms. These contract terms are often the focus in contract review and negotiation. AI tools using NLP have been developed for legal departments to perform textual analysis of proposed contract terms based on a legal department’s objectives. These AI tools determine which proposed terms of a contract are acceptable, and which are not. At present, contract review AI tools have not replaced review by attorneys. Instead, the AI tools serve as a check that allows for a more efficient review and identification of potential errors before contracts are finalized.
Contract Performance and Analytics: Once parties have a contract in place, it can often be difficult to monitor contract performance to ensure that agreed-upon terms and obligations are being met. Companies with many contracts between many different parties across different divisions of the company often grapple with this challenge. Like contract review AI tools, NLP-powered AI tools extract and conceptualize key terms from contracts and compare those terms with a company’s data metrics to determine whether contract terms and obligations are being met. These AI tools allow legal departments to harness the ever-increasing collection of data by companies to assess contract performance and compile analytics.
Litigation Prediction and Analytics: Machine-learning AI tools have also been developed to predict the outcome of cases based on relevant precedent, facts of the case, and prior outcomes in particular jurisdictions. Likewise, AI tools predict the likelihood of success for motions or other pleadings based on data-driven assessments. These litigation prediction models assist legal departments in making decisions on litigation strategies. In addition, litigation prediction models are also supercharging the litigation finance industry, where third party investors fund a plaintiff’s litigation case in return for a share of the award if the plaintiff is successful. Litigation prediction AI tools enable investors to develop assessments of which cases to finance, based on the likelihood of success from the prediction models.
Legal Research: Like litigation funding, other NLP-based AI tools build research platforms that have more sophisticated understandings of legal opinions. These platforms use NLP to uncover relevant law based on the fact pattern of a case, rather than keyword searching. AI helps legal departments to review past matters to assess risk, potential liability, and evaluate legal fee estimates based on analytics. These AI tools take advantage of a knowledge database containing information of interest to the company using the AI tool.
AI Assistance in Hiring
Another way in which companies have begun to use AI tools is to make the recruitment and hiring process more efficient. AI tools may be used for simple tasks, such as scheduling interviews and travel arrangements, or for sending targeted job listings. But companies are also beginning to reap the benefits of AI in other stages of the recruitment process, including using AI tools to screen candidates more efficiently. Not only have companies begun using AI tools to narrow thousands of resumes down to a reasonable number for further review, companies have also started using AI tools to conduct initial video interviews.
Despite the advantages of using AI in the recruiting processes, AI may also result in incidental bias and discrimination. One of the most commonly expressed concerns about using AI focuses on the underlying data the AI tools uses to analyze and make predictions about successful candidates. Because the AI algorithms are often trained using data pertaining to past successful and unsuccessful candidates, the AI may, for example, begin to favor candidates of a certain age, race, or gender. Similarly, to the extent the AI tool identifies characteristics shared by successful individuals (e.g., current employees), the AI tool may begin to favor candidates who also share those characteristics, which could include anything from educational background to membership in certain organizations.
These types of questions have prompted important discussions worldwide and have resulted in calls for transparency in how such AI tools are used in a hiring process. For instance, at least one proposed class action has been filed against T-Mobile US, Inc. alleging age discrimination regarding employment advertisements on online platforms. In another case, a consumer advocacy group, the Electronic Privacy Information Center (EPIC), filed a complaint with the Federal Trade Commission regarding AI hiring tools made by HireVue Inc. As described in the complaint, the technology conducts pre-employment assessments of job candidates, including video interviews in which thousands of data points are collected about a candidate and analyzed to predict the candidate’s employability. The allegations in the complaint highlight ways in which the hiring algorithms may result in bias, including the possibility that eye movement tracking captured in video interviews could discriminate against individuals with neurological differences.
As AI continues to develop in the coming years, and as its use becomes more prevalent in the hiring process, companies should keep these considerations in mind. Likewise, businesses should keep abreast of developing state and federal laws related to using AI. As an example of state law action, Illinois recently passed the Artificial Intelligence Video Interview Act, which establishes requirements for companies using AI to analyze video interviews. The law requires, for instance, that employers inform applicants before the interview regarding how the AI works and generally what characteristics it uses to evaluate candidates. At the federal level, various bills have been introduced regarding the use of AI, such as the Algorithmic Accountability Act of 2019, which would require businesses using “automated decision systems” (a term defined in the bill) to conduct impact assessments that evaluate the automated systems, including the design and training data for the system, to analyze its impact on “accuracy, fairness, bias, discrimination, privacy, and security.”
Conclusion
As AI tools become more commonplace, these sorts of requirements governing the use and testing of AI tools are likely to become more prevalent as well. For now, however, businesses may benefit from considering these issues in advance and taking initiative now to ensure AI tools are used fairly, accurately, and efficiently.
