The Possibility and Necessity of the Human-Centered AI in Legal Theory and Practice

Introduction

In 1948, Norbert Wiener, a founder of cybernetics, wrote: “we are already in a position to construct artificial machines of almost any degree of elaborateness of performance. Long before Nagasaki and the public awareness of the atomic bomb, it had occurred to me that we were here in the presence of another social potentiality of unheard-of importance for good and evil” (Wiener, 1983).

Today, “artificial machines” are already solving (or will be able to solve soon) multiple problems humanity faces. However, these machines undoubtedly, created new problems, too¹. What Wiener called “artificial machines” is now, in this or that form, a part of the life of society, and we can hardly imagine our life without artificial intelligence technologies. Therefore, it is not a surprise that in recent years there has been a lot of information «noise» around artificial intelligence and its potential to radically change the world we live and work in.

The objective of our considerations here is to show that as artificial intelligence technologies are being developed and introduced into our daily life, the necessity is proportionally increasing for the software developers, designers, and owners of the companies participating in the race to introduce new AI tools, to have humans and their needs but not machines and their efficiency, as the primary value and goal of advancement. It is not the goals of one person or company, not the technologies or machines, but a human and the human attitude that serves as the measure of morals and humanness. Realizing that good-hearted calls for humanness may sound abstract in the logic of technologically oriented development, we would like to discuss more specifically how the need to work with artificial intelligence within the approach called Human-Centered AI (HCAI).

The problem of the artificial intelligence orientation towards the good of humans is acute in all spheres of life but especially sensitive in some of them. These include education, medicine, and jurisprudence, where the price of a mistake – of a human or an algorithm – is the highest. Juridical solutions regulate human life and their relations with others and sometimes refer to existential issues – life, death, and justice.

In this article, we consider the problems the juridical theory and practice face with the advancement of AI in everyday life and how these problems correlate with the human-centered approach to AI. We define artificial intelligence as “an ensemble of rational, logical, and formalized instrumental rules developed and coded by human beings that organize the processes and activities to emulate rational/intellectual structures and fabricate and reproduce goal-oriented practices as well as the mechanisms for constructing further coding and decision making.” (Rezaev & Tregubova, 2019).

Today, one of the factors determining the development of artificial intelligence technologies is online culture – “an ensemble of communication networks, devices, algorithms, formal and informal rules of interaction, patterns of behavior, cultural symbols, which allow and structure people’s activity in the internet and similar networks, providing remote access to creating, exchanging and obtaining information” (Rezaev & Tregubova, 2019).

The Internet provides vast data on which artificial intelligence algorithms are trained and the “platform” for these algorithms to act.

As a result of the simultaneous development of computation capacities of artificial intelligence online culture, artificial intelligence is increasingly involved in everyday life and human relationships. “Artificial sociality” appears: artificial intelligence becomes an active mediator and participant in social interactions (Rezaev & Tregubova, 2019).

From its inception, the AI project had an a-disciplinary character. The artificial intelligence developers strived to reproduce human intelligence, hence, boldly borrowed the necessary provisions from mathematics, psychology, cybernetics, etc. (Russell & Norvig, 2007). However, while developing the machines reproducing the functioning of the human mind required turning to the achievements from various fields of knowledge, this is all the more true to understand how these machines enter the everyday life of society and are built into social relations. In other words, researching the problems of artificial sociality has an interdisciplinary and potentially – “a-disciplinary” character. That is why, in this article, we rely on both the philosophical and sociological analysis of the problems of AI, and the results of research in jurisprudence and law.

Further reflections are organized as follows. First, we will pay attention to several vital aspects associated with the introduction of digital technologies into legal practices. Then we will consider the problems and prospects of the rapid penetration of AI into the everyday life of society, which changes the characteristics of juridical work and the structure of legal systems. Finally, we will turn to the human-centered approach to AI and its consequences for the legal sphere.

1. Digital technologies and law

Summarizing the influence of digital technologies on the legal system, one should emphasize the following.

First, digital technologies simplified access to legal information via online databases, legal search systems, and other online resources. This fact, accordingly, created opportunities for nearly every Internet user (that is, almost 90% of the Russian population)² to perform online research to obtain legal information. The Internet revolutionized search in all spheres of human life (Utekhin, 2019) and legal information is no exception.

Second, digital technologies improved communication between lawyers, clients, and other actors in the legal system. For example, videoconferencing allows lawyers to communicate with their clients distantly. Thus, access to legal services for residents of remote districts and regions is improved.

Third, digital technologies made it possible to submit and store legal documents electronically. In other words, organizing juridical practices with digital technologies significantly decreases the need for paper documents and simplifies information search and exchange (Rusakova, 2020; Stepanov et al., 2021).

Fourth, digital technologies have led to automating many legal processes, such as routine checks of documents and compiling contracts (including the so-called smart contracts (Efimova et al., 2020)). Accordingly, the demand for routine manual labor of lawyers and their assistants has significantly decreased.

Fifth, using digital technologies in legal practices gave rise to new branches of law, such as cyberlaw/law in cyberspace (Mazhorina, 2020), intellectual property law, and data protection law (Voinikanis, 2020).

Thus, digital technologies have already significantly influenced the development of law, making legal services and practices more accessible, efficient, and effective. At the same time, practicing lawyers, special literature, mass media, and everyday legal service consumers almost unanimously emphasize that digital technologies generate new problems for the legal system development. These are, first of all, the issues of confidentiality (Talapina, 2022) and accessibility (Panchenko, 2012) of legal databases and the problem of critical assessment of the information obtained from the Internet (Greger, 2017).

The current stage of digital technologies development in online culture suggests paying attention to how artificial intelligence transforms and shapes further development of legal practices. What are the advantages and disadvantages of using AI technologies in routine legal practice?

2. Artificial intelligence in legal practice and theory: pro et contra

Using artificial intelligence technologies in routine legal practice has both advantages and disadvantages. The main benefits are the following:

– Effective organization of the lawyers’ practices. The AI instruments automate and accelerate the performance of such tasks as document review, preliminary juridical examination of literature sources, and analysis of contracts (Talapina, 2021).

– Artificial intelligence may perform specific tasks more accurately than people, for example, find regularities in data or check documents for factual mistakes, and grammatical or stylistic inconsistencies (Andreev et al., 2020).

– The use of AI technologies, reducing the need for manual work, saves the funds of juridical companies and their clients.

– Artificial intelligence technologies provide lawyers with more complete, compre­hensive, and detailed information allowing them to make more grounded decisions.

The main disadvantages of using artificial intelligence are the following:

– The common disadvantage of using artificial intelligence for all professions is that some professions disappear while others will appear and come to the fore (Lee, 2019). A broad use of AI technologies in legal practice is still a potential, but very soon, it will inevitably lead to a review of jobs nomenclature within the juridical system structure; this will especially touch upon paralegals and other auxiliary staff (Lessig, 2019).

– artificial intelligence systems are, to a certain extent, translators of bias and prejudices characteristic of their creators (Gorokhova, 2021). The artificial intelligence algorithms may be biased or erroneous due to, at least, two circumstances: a) if they are based on and were developed with biased or erroneous data arrays; b) if they are misused. Hence, the introduction of AI technologies implies searching for ways to ensure just and bias-free artificial intelligence systems.

– artificial intelligence technologies, like any other technologies, bear safety risks. AI technologies cannot guarantee complete cybersecurity (O’Neil, 2018). Artificial intelligence may minimize but not eliminate data leakage or hacking. Accordingly, confidentiality – the cornerstone of legal practices – is threatened when using artificial intelligence technologies.

Thus, using AI technologies in everyday legal practices provides multiple advantages, but they should be weighed with potential risks and drawbacks. Lawyers must not only realize the capabilities of AI, but also thoroughly review their use and see their limitations and potential risks.

Besides the problems with using the algorithms and machines which are already manifested in everyday life, one should also keep in mind the actual problems generated by the ubiquitous penetration of artificial intelligence into legal practices:

– confidentiality problems. The effective performance of artificial intelligence systems often requires access to large amounts of personal data, which causes concerns about confidentiality and data protection. Regulators and legislators must find a balance between privacy protection and promoting innovations in the sphere of artificial intelligence (Gorokhova, 2021).

– legal liability for the actions performed by artificial intelligence. As artificial intelligence systems become more autonomous and make decisions without human interference, questions arise about who is responsible for their actions (Vavilin, 2021; Baturin & Polubinskaya, 2022). For example, if an AI driverless car caused an accident, should a developer, a user, or the artificial intelligence system per se be liable (Rudenko, 2020)? Who will bear responsibility if something goes wrong when AI instruments are used? Who will be responsible for accidents or mistakes caused by an artificial intelligence system: a programmer, an owner (of what?) or the AI designer? These already are the juridical questions of today.

– the issues related to intellectual property rights to the products created by artificial intelligence technologies (Lee et al., 2021). For example, who will be deemed an inventor or an artist if an artificial intelligence system creates a work of art or invents a new technology?

– a critical element is using artificial intelligence tools (for example, ChatGPT) for juridical interpretation of documents and applying legal norms, especially regarding the complex and nuanced character of juridical substantiation of a certain decision. There are grounds to fear that artificial intelligence will not be able to comprehensively grasp human considerations and judgments necessary for effective juridical decision-making (Tsvetkov, 2021).

– lack of communication and real-life human contacts. This is a significant disadvantage for legal practices, which may, by default, touch upon existential matters of life and death, restriction of freedom. Judges note: justice is impossible without a holistic view of the situation, including moral and emotional aspects, which is inaccessible for AI (Bykov & Narskaya, 2022).

Noteworthy, the very question of whether AI technologies should be regulated and how is the object of discussion (Etzioni & Etzioni, 2017). The frameworks are just starting to be elaborated in this sphere, the “pioneers” often being the legislators of the European Union (Hickman & Petrin, 2021; Fink & Finck, 2022; Ulnicane, 2022). In Russia, legal regulation of artificial intelligence technologies is also being developed. In 2019, the National strategy of artificial intelligence development up to 2030³ was adopted, specifying the basic definitions and general principles of using AI technologies.

Thus, the development of artificial sociality poses both a practical and conceptual problem for jurisprudence. The practical/functional problem is how artificial intelligence technologies will change legal practices, while the conceptual problem refers to their legal regulation.

We believe that a set of problems which have already emerged and which are bound to emerge in the future in legal practice can be solved more effectively with the approach called a human-centered AI.

3. Human-centered artificial intelligence

The approach called a Human-Centered AI in the scientific literature (Ford et al., 2015; Shneiderman, 2021)⁴ implies, first of all, understanding the straightforward fact that people and machines are not the same⁵. There is no need to aim at making an artificial intelligence tool similar to a human being. On the contrary, success will probably be achieved in the opposite direction when a human stays a human with their intellect, consciousness, subconsciousness, and emotional and spiritual world. At the same time, machines and algorithms will be developed by a human and, during “self-training,” follow their own logic of development, different from that of a human.

Unfortunately, this circumstance is being neglected, just like the human-centered approach to AI in general. Most technological leaders in the USA and other countries continue spending a lot on developing software that can do just what people can do. Developers very well realize that they can earn easy money by selling their products to corporations having no other orientations in their development except those set by the logic of the market and profit (Zuboff, 2022). Everyone is focused on using artificial intelligence to reduce costs for the working force while caring little about the essence of the social progress and development of a moral human being and a just society.

Human-centered AI requires immediate attention to collecting and analyzing high-quality data to organize the development of artificial intelligence tools. The artificial intelligence algorithms are as effective as the data on which they are trained are effective. In contrast, partial or incomplete data may lead to not only unjust/false results but to ones opposite to the initial goal. Collecting information for self-learning models must be diverse and representative; the data must reflect the real world we live in and the people we work with, regardless of their social and class differences.

Elsewhere, we have already emphasized the fact that, under the current stage of capitalism development, under the extreme orientation towards financial indicators, profit, and functional efficiency, it is practically impossible to solve these problems (Rezaev, 2021)⁶. Nevertheless, it would be wrong for the tactics of social sciences development not to consider them at all and not to attempt to propose variants of their solution.

The market has never been and cannot be (even under artificial sociality) the touch­stone of beauty, goodness, and truth. Strategically, social knowledge substantiated the impossibility of a harmonious, moral, and just world without exploitation of human by human and without social and cultural inequality within the framework of a capitalist economy⁷. However, what the problems for society are and what the variants of the social development trajectories are under the still uncontrolled spreading of AI tools – these topics are just starting to be considered belatedly.

Characterizing the features of artificial intelligence development, one should remember that AI technologies are not neutral. Humans create them, and algorithms reproduce their creators’ values, biases, and prejudices. Thus, AI designers and producers must adhere to ethical and human-oriented approaches. It means, among other things, accounting for various viewpoints and opinions in the design process, providing transparency and accountability, and paying priority attention to the human personality and wellbeing of society in general, not that of individual subjects or technological systems.

The key point is the understanding that AI instruments are already a powerful means of solving some of the most burning problems facing society, but they are not a panacea. While defining and formulating the directions of social development, one should not rely exclusively on artificial intelligence when solving social, economic, cultural, and political problems. Even under “artificial sociality,” people must remain within the reality of human experience and admit that social progress requires more than just technological solutions.

Conclusion

This essay began with a citation from Norbert Wiener, the founder of cybernetics. We want to conclude it with the judgments presented by one of the founders of artificial intelligence research – Joseph Weizenbaum. Weizenbaum argued that the use of computers should be banned or at least restricted in two cases (Weizenbaum, 1982). The first case is connected with attempts to replace a human with a machine in the areas related to interpersonal relationships, love, and understanding. The second case is using computers in a situation where it can lead to irreversible consequences. In our opinion, Weizenbaum correctly formulated the basic principles of human-centered artificial intelligence, which relate to the general spread of AI technologies and their use in the theory and practice of jurisprudence, in particular.

In conclusion, we will formulate three principles of including AI into the legal theory and practice according to the methodological principles of human-centered approach (HCAI).

First. A human being must always remain within the chain of making/executing legal decisions. Legal scientists have persistently formulated this thesis. Artificial intelligence technologies may take on many tasks in legal practice, but it is a human being that must control, check, conceptualize, and weigh the actions and decisions of the artificial intelligence.

Second. Today, we need to elaborate the laws determining a rational and understandable modus vivendi for the activity of artificial intelligence in social systems aimed at a human being, not at profit and market. This is almost impossible within one state, especially a capitalist one. That is why the world faces the need to create international law for AI evolvement in society. Like any rule, the law may be violated – by mistake or out of malice. But violation of the law does not repeal the law itself; it just reveals the malicious persons who distorted the law.

Third. The progress of AI in everyday life of people poses the need for prohibitions, including juridical ones, a taboo for using artificial intelligence in certain spheres of human life (Rezaev, 2021). These are, first of all, spheres associated with existential issues. For example, an important issue is whether one should use artificial intelligence to determine if a person is lying (Oravec, 2022), or whether artificial intelligence may serve as an autonomous weapon (International Committee, 2020). Defining such spheres at international, national, and local levels, formulating legal prohibitions, and creating law-enforcement mechanisms is one of the priority tasks for Human-Centered AI.