As a result of technological developments since the beginning of the 2000s, we have often discussed technology laws (e.g., cybercrime, information security, electronic commerce, cybersecurity, internet law, and personal data protection). However, law technologies, using technology in the legal world, is a relatively new topic.
We see that developments in this area parallel the widespread use of the internet. Significant milestones for lawyers started after 1993 when Turkey connected to the internet for the first time. The electronic issue of the official journal, the creation of legislation and case-law database programs, and the implementation of applications such as the electronic signature, the electronic justice system, and the electronic government portal were all realized in the early 2000s.
All these technological developments affect legal developments, and as our expectations from technology increase, we need new regulations. In this sense, the new Turkish Commercial Code accelerated the use of technology in law and affected other legislative studies. Many new applications entered our lives, including electronic notification systems, electronic trade registry records, electronic book-keeping, general meetings and board meetings, corporate websites, electronic money, electronic notary public, and data matrix embedded checks.
These are all the key technologies we have been using in our duties as lawyers, which have been essential for the Turkish legal world for the last 20 years. What about the technologies we are not yet using but will start using soon?
The most popular and debated of these technologies today is AI, i.e., artificial intelligence. The difference between artificial intelligence and computer programs we have used so far is its ability to think, learn, and improve. The first appearance of artificial intelligence was in the 1997 chess championship, in which Deep Blue defeated the world chess champion Kasparov; for the first time in this championship, a computer won. With this growing technology, after 20 years, we learned that in 2017, Facebook suspended an experiment as their two artificial intelligence programs developed a language that people did not understand. Undoubtedly, artificial intelligence researchers are also working on using this technology in the legal world.
Technology versus Human Lawyer
In February 2018, researchers conducted an interesting experiment in the US, where this time, artificial intelligence and humans competed on a legal matter. The competition compared the human lawyer’s performance and artificial intelligence called LawGeex to review standard commercial contracts. For this purpose, researchers decided to use five typical confidentiality agreements. Confidentiality agreements are agreements made at the beginning of negotiations in many commercial operations, and they are strategic contracts for the protection of trade secrets and intellectual property rights.
20 US lawyers from various law firms and companies with more than ten years of experience in corporate law and contract review and artificial intelligence developed for three years and trained by thousands of contracts competed in identifying legal issues in these five confidentiality agreements. Researchers conducted this experiment under the supervision of an independent consultant and lawyer with the contributions of academics, data scientists, legal education experts, and machine learning experts.
First, law professors from reputable universities and experienced lawyers identified 30 legal issues that may arise in a standard confidentiality agreement. These issues provided the basis for assessing lawyers’ and artificial intelligence’s accuracy in identifying issues while reviewing confidentiality agreements. They chose five confidentiality agreements among the ones used in real life by five large and different US companies. The artificial intelligence algorithm has not processed any of these five agreements before, and artificial intelligence reviewed these new agreements based on thousands of other contracts it has previously processed. The experiment was designed in line with the scenario as if customers uploaded a new transaction to artificial intelligence for the first time in real life. On the other hand, researchers gave each lawyer four hours to identify the legal issues in these five confidentiality agreements and paid them for their time, like real legal work.
After two months of extensive testing, artificial intelligence, with an accuracy of 94%, prevailed over lawyers who achieved an average accuracy of 85%. Lawyers spent 92 minutes on average for five agreements, while artificial intelligence spent only 26 seconds! Artificial intelligence reached 100% accuracy in one contract, while the highest accuracy was 97% for lawyers. Besides, throughout this assessment, artificial intelligence never needed coffee!
This result does not mean that artificial intelligence will do the lawyers’ work from now on. At least for now! However, we can conclude that artificial intelligence can be an essential tool to help us conduct our work and save time.
Artificial Intelligence in Law
Regarding the current and future use of artificial intelligence in the legal world, a detailed article published by Fordham Law School sheds light on us:
The request for pre-trial documents, one of the rules of procedure applied in some lawsuits in US courts, is indicated among the matters artificial intelligence is very competent. In the request for pre-trial documents, both sides of the case must submit all information and documents related to the other party’s dispute for review. In this way, the parties to the lawsuit have access to evidence supporting their claims and defenses. After submission, artificial intelligence enters the stage and helps examine all the information and documents presented. It can quickly identify the necessary documents thanks to predictive coding, a more advanced program than a keyword search. Not surprisingly, there are documents that this program does not notice when necessary or does mention when it is unnecessary, but human lawyers also make similar mistakes. Some courts, which find this software’s cost and performance comparable to the traditional document review method, allow its use in the electronic document review procedure. Some law firms in the US have also begun to use this technology for large commercial lawsuits. Thus, artificial intelligence is taking over the document review, on which less experienced litigation lawyers spend a great deal of time.
As for the search for high court precedents, another significant part of lawyers’ work, the research anticipates that artificial intelligence will entirely edge out humans in the future. Today, many law firms in the US have already begun to take advantage of artificial intelligence, searching for high court precedents on a specific issue, which was traditionally a task for a candidate or less experienced lawyer. Thanks to the advancing technology, there will be semantics-based software rather than a conventional keyword-based search program. So, artificial intelligence will respond to a specific word given to it and further consider other relevant information while responding to a search request. Likewise, researchers experiment with another program that focuses on legal matters. This software will have superior features over semantics-based software, enabling searching according to concepts instead of words.
Again, researchers are also working on another topic, enabling artificial intelligence to decide whether a precedent found in the search will be a strong enough ground for a relevant lawsuit. For now, a human lawyer makes this decision, but researchers expect improvements in the next ten years.
Template legal documents are another area that artificial intelligence can help. Artificial intelligence can personalize such form documents by adapting them to individual circumstances. For example, in the US, with the LegalZoom program, one can create a draft testament by entering information about their assets and other information depending on how they wish to use the estate. Experiments are in progress for creating many other legal documents with artificial intelligence. Developing the forms by linking certain contract forms to court precedents relating to these contracts is among the experiments. For now, a human lawyer must work on and make countless corrections to a draft legal document created by artificial intelligence. Though, given the time spent preparing the first draft by a human, it is no doubt that artificial intelligence will accelerate us.
So, researchers expect that artificial intelligence will prepare the first draft for many legal transactions in the next ten years. In the future, such as a human lawyer, artificial intelligence may design personalized and case-specific contracts according to a matter’s characteristics and relevant legal issues and follow these contracts’ outcomes through future lawsuits.
Similarly, artificial intelligence will also seem to play a role in creating court petitions and legal opinions. Like in contract drafting, artificial intelligence will initially form the first draft of a court petition or a legal statement, and a human lawyer will need to work extensively on this draft, but the draft would be equivalent to a draft of a candidate or a less experienced lawyer. A simple example of this was tried in journalism in 2014 through a computer program writing a news article for the Los Angeles Times about an earthquake. Researchers are currently experimenting with more sophisticated algorithms, which would shape their drafts according to the information gathered from legal search programs to be connected while drafting. So, artificial intelligence may also create the first draft of court petitions or legal opinions in the next ten years.
Additionally, artificial intelligence studies are interested in establishing connections between law and new disciplines. For example, predictive analytics is a new discipline in terms of the law. Researchers train several computer-generated models with data and use these models to predict the results of a hypothetical situation, like predicting the outcome of a lawsuit. An enormous amount of data, including case law, court petitions, and other documents, is quite suitable for data mining in this analysis method.
Unlike artificial intelligence, as human lawyers, we implicitly predict a lawsuit’s possible outcome through our legal advice, such as filing or not filing a lawsuit or settling, but we base this prediction on our experience, knowledge, and research. On the other hand, predictive analytics uses large amounts of data far beyond what an individual lawyer can access and go through. Some universities and companies in the US are currently using this method. For example, a program called Lex Machina dedicated to patent lawsuits was initiated with the collaboration of Stanford University’s law and computer science departments and formed into a company. However, it is worth noting that this method of analysis only generates a predictive result, not a definitive one, and for the time being, the opinion of a human lawyer is still crucial for the final assessment.
The Limits of Artificial Intelligence “for Now”…
There are matters where artificial intelligence cannot add value despite all its skills. For example, we do not expect artificial intelligence to be present at hearings soon. Furthermore, human lawyers specializing in new laws and regulations will likely prevail over artificial intelligence. For now, artificial intelligence is successful in conventional and stable areas. Therefore, we anticipate that its use in countries like Turkey, where the legislation changes rapidly and continuously, will probably not be satisfactory at the preliminary stage.
On the other hand, human lawyers do much more than legal analysis. A human lawyer establishes a long-term trustworthy bond with their client based on professional and ethical rules. Moreover, they often hear not only what their client says but also what their client does not say, question the incomplete or inaccurate information insistently, get the correct information, create solutions for issues that have never arisen before, and can take on the role of convincing their client, who may be reluctant to do what is in their favor. Artificial intelligence is unlikely to reach such a level of reasoning and emotional intelligence soon, but one day, it may also be possible. Even appointing artificial intelligence as a judge at that stage may be a matter of discussion!
So, looking at where artificial intelligence stands today, we need to transform two areas. The first is the need to adopt a different approach to legal education. If artificial intelligence starts to do the work of an inexperienced lawyer, how a human lawyer will get the expertise they need to acquire can be a problem. The second issue is the law of legal practice. Our bars and the Union of Turkish Bar Associations will need to work on lawyers’ use of artificial intelligence while providing legal services. We will need to discuss and resolve matters such as ethical rules, attorney-client confidentiality, and the use limits of such technology. Accordingly, as the use of artificial intelligence by non-lawyers may come to the agenda, specific measures and regulations will also need to be introduced to prevent any impairment to a layperson in using artificial intelligence without a lawyer’s supervision.
And last but not least, while lawyers are under civil, criminal, and disciplinary liability, will artificial intelligence, a software and computer program, have civil and criminal liability? If so, how? These issues are critical legal matters we need to resolve as artificial intelligence technology develops.
Nevertheless, we must follow all these developments and use the necessary technologies to the appropriate extent to provide better legal services to our clients.
Av. Müge Önal Başer, LL.M., LL.B.