Darren Dahl and Harish Krishnan: Leaders should know when to rely on AI, when to question it and when to override it. They should be able to justify the decisions they make and stand behind the consequences.
Article content
Many organizations, including the business school we teach at, are adopting artificial intelligence. But in the rush to build this capability, there may be a quieter risk emerging.
Article content
Because artificial intelligence can generate answers faster than we can fully assess them, we may be tempted to rely on its recommendations without remaining fully engaged in the judgment and responsibility that consequential decisions require.
Article content
Article content
Story continues below
Article content
The danger of AI isn’t simply that it may become highly capable. It’s that we may quietly cede human judgment and authority to machines because doing so is efficient or convenient.
Article content
Article content
Decisions that shape people’s lives are increasingly influenced or made by AI systems. Job applicants are screened by algorithms. Mortgage approvals are determined by automated risk models. Hospitals are using AI tools to help prioritize care and allocate scarce resources. In each of these cases, the promise is compelling: greater efficiency, faster processing and more consistent outcomes.
Article content
But beneath that promise lies the question that may be overlooked: When a decision has real consequences for a person’s life, who is accountable for it? If a job application is rejected, a loan is denied or a patient is deprioritized, where does responsibility sit? Is it with the organization that deployed the system? The individual who relied on it? Or does it somehow get diffused into the technology itself?
Article content
Instrumental decisions — such as a thermostat turning on a furnace when a particular temperature is reached — have long been automated. These decisions are made within the context of predetermined human needs and preferences. But agentic AI systems — consider a self-driving car — are being designed to make judgment calls over consequential, even life and death matters.
Article content
Story continues below
Article content
When the rules are clear and the situation facing the decision-maker falls within those rules, a machine can simply apply them. But decisions defined by deeper judgments that might involve trade-offs, moral ambiguity or that need critical thinking are different. Leadership begins when the decision situation isn’t one characterized by simple rules, and someone must exercise judgment and stand behind the result.
Article content
Read More
-
B.C. company goes out on the edge to stake ground in AI race
-
Pamela Anderson and Aerie take a stand against AI models
-
Advertisement 1
Story continues below
Article content
A person or institution can delegate instrumental choices to a machine and remain accountable, because the person or institution defines the objective, sets the decision rules and chooses to deploy the system. But when the decision rules run out, or no longer apply, responsibility remains with the person or institution authorized to decide. It can’t be shifted to a machine that can’t understand the broader implications of the decision or stand behind the result.
Article content
Why then do we see this type of decision-making now being delegated to machines? It’s often the result of optimizing efficiency or other objectives that are legitimate in isolation. Because modern AI systems execute instrumental decisions much faster and often more reliably than humans in certain contexts, they’re now being given increasing authority.
