Rodrigo Bertoccelli

What human values should AI be based on? Should it follow utilitarian or deontological ethics?

* Integrity and Development is a column by the Center for Studies on Integrity and Development (CEID), of Instituto Não Aceito Corrupção (INAC).

This article reflects solely and exclusively the opinion of the author and does not represent the views of CEID and INAC. The articles are published weekly

Artificial intelligence (AI) is taking on increasingly surprising proportions. One example is the initiative of Chinese company NetDragon, which appointed an AI named Tang Yu as its chief executive.

Tang Yu is the first AI to hold an executive position in the world and was recently named “China’s Best Virtual Employee of the Year” at the 2024 China Digital Human Industry Forum and Press Conference. But will its decisions always be ethical?

Has the time come for AI to also take on the role of Compliance Officer?

The central question is: what human values should AI be based on? Should it follow utilitarian ethics (like John Stuart Mill and Jeremy Bentham) or deontological ethics (like Immanuel Kant and John Rawls)? With so much capacity to process data, will the reasoning behind its decisions always be transparent? Should it be like us or better than us?

Alan Turing, a pioneer of computing, also raised ethical questions about AI by asking which human values the technology should align with.

These questions remain unanswered. Kate Crawford, co-founder of New York University’s AI Now research institute, argues that AI systems are embedded in social, political, cultural and economic worlds, shaped by humans and their institutions.

The debate on AI

Most of the current debate on AI and the law focuses on how the technology may challenge prevailing regulatory paradigms. According to Professor Antara Haldar, of the

University of Cambridge, one concern is the “red queen effect” (an allusion to Alice in Wonderland), which describes the difficulty of keeping regulation up to date amid rapidly evolving technology.

Another challenge is regulating a global technology at the national level. In addition, there is the problem of a new technology being developed by a group of private companies whose priorities (profits) differ from those of the public.

As tech giants push ahead in pursuit of general AI — models capable of outperforming humans at any cognitive task — the AI “black box” problem persists, challenging the decisions of CEOs and Compliance Officers. Not even the technology’s creators know exactly how it works.

Laws can be imposed to guide conduct. Through rules, we can require AI to obey our collective and corporate code of conduct. If AI’s neural networks imitate our brains, and law is widely considered a largely cognitive phenomenon, this should be possible.

Ethics, however, is different; it involves freedom and decisions in unprecedented situations. This is where efforts to regulate AI should begin, as should the ethical role of Compliance Officers.

AI and the law

It is always difficult to strike the right balance between promoting innovation and mitigating the potential risks associated with a new technology. With AI expected to increasingly change the practice of law, can the law still change the trajectory of AI? If “thinking machines” are capable of learning, can they learn to obey the law?

Law represented a radical new technology. Once internalized, it aligned individual action with community consensus. Several philosophers have focused on this unique characteristic. The 20th-century legal theorist H. L. A. Hart described law as a mechanism that allows rules to be shaped by changes in the underlying behavioral meta-norms.

More recently, Ronald Dworkin characterized law in terms of “integrity”, because it incorporates the norms of the entire community, rather than resembling a “checkerboard”.

If law were a patchwork quilt, it might better represent individual elements of belief and opinion, but at the expense of coherence.

Law therefore works as a substitute for individual human

behavior. It absorbs complex debates about morals and values and turns them into binding rules.

Transparency and accountability in the implementation of AI systems are crucial. Ethics in the collection and use of data in artificial intelligence is also fundamental. John Rawls is known for his theory of justice as fairness, presented in his work A Theory of Justice (1971).

He proposes principles of justice intended to ensure a fair and equitable social structure.

Rawls’s ethics focuses on how institutions distribute rights and duties and regulate the division of social and economic benefits. He introduced the idea of the “veil of ignorance” as a tool for determining the principles of justice of an ideal society, suggesting that ethical principles should be chosen without knowing the position one will occupy in society.

From these definitions, we can understand how ethics stands above laws and conventions as a sense of responsibility of a person or institution toward the world. This becomes a fine line when the subject is Artificial Intelligence replacing the decisions of a Compliance Officer.

Ethics in the use of Artificial Intelligence is an effort that accompanies the design, structuring and operation of the tool to prevent its misused power from corrupting its purpose.

AI’s critical point

The most critical point is the accuracy of information. Anyone using AI should bear in mind that the answers provided may be outdated. Review is therefore essential. A second important point is understanding that no tool replaces the creative element of the human being.

According to scientist Miguel Nicolelis, “if the future of AI is based on the database of its past, there is no future. If all your music is produced by an intelligent

system based on Beethoven, there will never be another Elton John”.

These observations complement the debate on ethics in artificial intelligence, highlighting the importance of an ethical and responsible approach to developing and using this technology.

AI may even learn to comply with laws and certain rules of conduct, but we cannot expect its decisions to always be ethical, since it is not as intelligent as it is artificial.

AI has no morals; it is merely a statistical probability model, capable of following a set of rules and making predictions based on historical data, and

therefore cannot replace a compliance professional.

Artificial intelligence (AI) is transforming corporate compliance by processing large volumes of data accurately and quickly. It identifies patterns and anomalies, increasing efficiency and reducing errors, as well as enabling continuous real-time monitoring.

However, human oversight is essential to avoid bias and unfairness. AI should be seen as a tool that complements human professionals, creating a more robust and efficient compliance environment.