Leadership and Moral Decisions in the Age of AI Algorithms

Rules alone will not make artificial intelligence (AI) trustworthy. What we need are leaders of character — who see themselves as responsible stewards rather than mere operators.

In 1966, the computer scientist Joseph Weizenbaum (1923-2008) built a simple program called ELIZA. Interestingly, it did little more than rephrase what users typed back at them as questions. Yet people developed emotional attachments to it. Weizenbaum was so unsettled by what he saw, that he spent a considerable portion of his career warning about the social consequences of the technology he had helped create.

Today, about sixty years later, complex machines are no longer merely reflecting our sentences back at us. AI agents — more or less autonomous systems that plan and carry out tasks toward a goal, largely on their own — are entering our organizations. Not as mere tools but as active participants capable of operating as team members. They draft, decide, sort, recommend, and, where we permit it, act. We are in the early stages of the development of what many refer to as hybrid teams: groups where some members are human and some are not.

However, this is not primarily a technological development. It is a social and moral one. And it is a force to be reckoned with for anyone who leads an organization.

Why Principles Are Not Enough

The standard response in the field of ethical AI has been to write ethical guidelines. There are now hundreds of them. Reviews of the field find no single agreed standard, but they do find a genuine convergence around a handful of themes: transparency, justice, non-maleficence, responsibility, and privacy. The OECD and the European Commission's expert group arrived at much the same list, adding explicability as a way of making the other principles real.

This is certainly a necessary achievement. However, it is not a sufficient one. Researchers who have examined these frameworks closely, such as Mittelstadt and Hagendorff, identify problems that should trouble any leader. The guidelines have no enforcement mechanism, so violating them rarely have any significant consequences. Moreover, developers seldom experience ethics as relevant to their actual work. And responsibility is distributed so widely that it often evaporates. When everyone is said to be a little accountable, no one actually is.

There is a deeper problem still. Rules are written for situations someone managed to anticipate. Autonomous agents, on the other hand, generate situations nobody anticipated. The more independently a system acts, the wider the gap grows between what the rulebook covers and what actually happens on a Tuesday afternoon in a real organization. Compliance was designed for a world of foreseeable cases. But that is not the world we are entering.

Character, Not Just Conduct

If rules cannot carry the whole weight, what else is available to us?

Here one of the oldest traditions in Western ethics turns out to be surprisingly current. Virtue ethics, running from Aristotle through Aquinas to Alasdair MacIntyre's recovery of the tradition in our own time, has a different perspective and thus, asks a different question. The question in this tradition is not, What act is permitted? but What kind of person should be acting? Its focus is character — the settled dispositions that shape how someone perceives a situation before any rule is consulted.

Central to this is what Aristotle called phronesis, practical wisdom: the capacity to judge well in a particular case where the general principles are difficult or impossible to apply. Several AI ethics researchers have argued that this is a firmer foundation for the field than a purely rule-based approach, precisely because it addresses the actor rather than only the act. And authors such as Kate Crawford and Ryan Calo have argued that such judgement has to be woven through every phase — idea, design, deployment, regulation — rather than bolted on as a risk assessment at the end.

For a leader, this translates into something concrete. The question is not only Are we permitted to do this? but Should we, and who will we become if we do?

Trust Is Built by People

Which brings us to trust. Trust is usually defined as accepting vulnerability to another's actions, in the confident expectation that they will act in your interest, even though you cannot monitor or control them. That is a fair description of an employee asked to rely on an AI agent's judgement — and, for that matter, of a citizen subject to an algorithmic decision.

Classic research on trust identifies its three main sources: ability, integrity, and benevolence. None of them arrives with the software. They must be built, and they are built by people.

This is where servant leadership — the tradition Robert K. Greenleaf named in 1970, drawing on a far older stream of thought about leadership as service — becomes newly relevant. Servant leadership is an approach where concern for others is the leader's first priority, expressed through making each employee's development and growth the leader's main purpose, and involves a reorientation from self-interest toward care for the employee, the organization, and the wider society. The research literature connects servant leadership both to virtue ethics and to trust.

Greenleaf's most underrated idea was perhaps the servant institution: the organization that understands its own role in society as one of service. That vision has rarely been more relevant. A leader formed in this way can act as a trust builder between employees and autonomous systems — not by pretending the technology is infallible, but by owning the judgements behind it. Why we are introducing this. What we will not permit it to do. Who answers when it fails.

Leaders as Stewards

There is an older word for what this asks of leaders: stewardship. Stewards do not own what has been entrusted to them, and they know they will be asked to give an account of it. That is precisely what a technology of unprecedented reach requires from those who deploy it. Not enthusiasm, and not refusal, but people willing to say: this was entrusted to us, and we are answerable for what we did with it.

AI agents represent a societal transformation, not merely a technical one. Whether we reap the benefits without paying an unreasonable moral price will depend on the quality of our guidelines, but even more so, on the character of those who lead.

This article draws on the author's research article "Tjenende ledelse og moralske maskiner: En etisk og tillitsbasert tilnærming til kunstig intelligens (KI) og KI-agenter", published in Magma forskning og viten, 29(3), 107–113. https://doi.org/10.23865/magma.v29.1567‍ ‍