We often talk about AI as if it can make decisions. But when AI enters real decisions—those that affect people’s rights, opportunities, and access to services—one thing becomes clear: responsibility does not transfer. It stays with us.
This truth was reinforced when I recently completed a course at NUS on Human Resource Management and Artificial Intelligence.
What We Are Seeing on the Ground
This became very real to me while helping craft our agency’s Responsible Use of AI Policy.
A few things stood out immediately.
First, our people are already using AI informally and often without fully understanding its limits.
Second, parts of the workforce have not yet fully leveraged IT—not just in operations but also in systems development. Yet, the conversation about AI is already here.
Third, we realized that if we were to take this seriously—especially from a privacy and data protection standpoint—access could not remain informal. It has to move toward enterprise-level subscriptions, with safeguards in place.
Perhaps, most telling: we still do not have clearly defined use cases, particularly in HR, where the implications of AI decisions are deeply human.
What AI Actually Does—and Does Not Do
All of this brings the same point back into focus.
AI has no intent, no sense of fairness, and no understanding of consequences. It processes patterns. So when a system denies a loan, flags a civil servant, or prioritizes a citizen’s request, the outcome may feel final—but the judgment is not its own.
Someone chose the data, defined the model, set the thresholds, and decided where it would be used. That person remains accountable.
What We Give AI, It Scales
Those choices are never neutral. AI inherits what we give it—our data, our assumptions, our priorities. If these are incomplete or biased, AI does not correct them. It scales them. What used to be a small human error can become a consistent system outcome.
This is where responsibility deepens. It is not just about control. It is about vigilance.
Why This Matters in Public Service
In public service, this matters even more.
Decisions must be explainable, defensible, and open to challenge. You cannot tell a citizen, “the algorithm decided.” That is not due process. There must always be a clear line back to a human decision-maker or institution.
Trust does not sit with systems. It sits with institutions—and the people behind them.
AI will only strengthen public trust if citizens believe that human judgment, oversight, and care are still present. Without that, even good technology can quietly erode confidence.
AI Reflects What We Value
There is also a deeper truth: AI reflects our values.
Efficiency? Fairness? Inclusion? Control? These are not technical decisions. They are human decisions. AI simply operationalizes them—at scale.
A Governance Principle, Not Just a Safety One
So human responsibility for AI is not just a safety principle. It is a governance principle.
It means we define the purpose and the limits. We validate and audit outcomes. We remain accountable for decisions. We intervene when systems fail. AI does not reduce responsibility. It intensifies it.
— APMG Noreen
Image Credit: Image generated using AI (DALL·E by OpenAI), based on the author’s concept on human accountability in AI.
