World chess champion Garry Kasparov playing IBM’s Deep Blue computer in 1997 © AFP
Conversations about artificial intelligence incite a reaction that is equal parts excitement and dread. Excitement because of the potential cost-saving efficiency gains that may come with AI. Dread because any mention of automation induces a fear of redundancy and disposability among mid-skilled and low-skilled workers.
In public sector management the conversation about machine learning applications is no different. Early this year for instance, Deloitte and University of Oxford published research suggesting that up to 10% of UK public sector jobs would be automated by 2030. If investment in technology is low, this could save the government up to 4% in labour hours, and up to 31% if investment if investment is high. At the heels of this optimistic projection though, is the looming threat of mass retrenchment of civil servants, and in the long run, widening inequality.
These fears notwithstanding, artificial intelligence is here to stay.
Governments have to decide how to best manage the disruption of the civil service that is part and parcel of incorporating artificial intelligence into public sector management. If managed properly, AI could not only increase public service efficiency, but reimagine the very role of civil servants as it pushes them towards tasks that involve creative thinking and rigorous problem-solving. Artificial intelligence is potentially at its most powerful if used to augment human tasks instead of replacing humans.
We only need to look at computer-assisted chess (commonly referred to as Advanced Chess), to see a level of play that is unprecedented in both human chess and computer chess. Artificial intelligence could therefore serve to assist civil servants particularly when performing tasks that are of lower complexity – jobs like processing migration paperwork, or handling fraud detection and compliance checks on public projects. This in turn allows the civil service to engage in more creative tasks, improving the quality of public service delivery immensely.
A worry usually raised against the rosy picture of human-computer collaboration pertains to the exponential nature of developments in the capabilities of AI. If artificial intelligence is able to perform tasks requiring simple judgment, what is stopping these systems from eventually developing a sophistication enabling them to perform the very tasks that require creative thinking and problem solving skills? Imagine a Level 5 government that is completely autonomous – a reckoning for the public sector straight out of a J.G. Ballard novel.
Though a distant worry, this fear of the complete redundancy of humans in intelligent systems misunderstands AI-aided governance. Machine learning happens through supervised learning, unsupervised learning, and reinforced learning. All of these processes require a civil servant’s parental oversight to create training data sets for the technology, to set the agenda for policy making, to monitor decisions and importantly, to review them in cases when algorithmic decision-making fails.
It is ultimately up to governments to harness AI for the better in public service delivery and civil service development. This will necessarily require that governments innovate from a place of empowerment rather than fear; replacing the narrative of redundancy and replacement with that of augmentation and enhancement.
