MONCTON, N.B.—Among all the important roles Parliament plays in making representative democracy work, holding the government to account stands out. However, the constitutional convention of ministerial responsibility, which plays a central role in the Westminster parliamentary system, is operating under severe stress. Among other factors, the size of government, the 24-hour news cycle, social media operating with little editorial control, the reality that policies and most programs are now the product of many hands, and the shift away from cabinet government to governing from the centre all weigh heavily on the constitutional convention.
There is a development on the horizon that will make it even more difficult for Parliament to hold the government to account: artificial intelligence (AI). In 1979, IBM, in its training manual, declared that “a computer can never be accountable, therefore a computer must never make a management decision.” AI is already making managing decisions in government, and this is just the beginning.
Albania, for example, recently appointed AI as a member of cabinet and responsible for procurement. It pointed to the public interest in appointing a non-human as minister, making the case that AI removes human bias and bribery from the government’s decision-making process while promoting more efficient administration
In Canada, the federal government is turning to AI to shape and, at times, strike decisions in immigration, border security, and tax administration.
AI holds untold potential to streamline decision-making in government, to identify weaknesses in administration, outline where savings can be realized, improve services to the public, and support more evidence-based decision-making. Public administration is essentially about due process, processing applications, ensuring fairness in administering programs, and dealing with a wide variety of issues and problems. AI can do this far more efficiently than traditional government bureaucracy and, in time, it will be able to do a great deal more. AI can also help MPs deal with their constituencies, given its capacity to deal with complex administrative issues, analyze a region’s socio-economic circumstances, and understand past policy and program failures.
AI is improving at dizzying speed. The day is near when AI super intelligence will surpass human intelligence in all fields, and outperform the most gifted individuals. At one point, it will make the shift from a passive adviser to an active agent, giving it autonomy and the potential to escape human control to strike decisions as it sees fit. AI’s impact on accountability will be more far more strongly felt in the public sector, where accountability requirements are more opaque than in the business community.
What would be the role of Parliament and its accountability requirements if AI can operate outside human control? How could backbench MPs possibly hold ministers or their departments to account if AI can undermine human influence in striking decisions? How will Parliament and government establish the difference between benefits AI can generate, and the problematic loss of human influence in shaping policies and delivering programs? The doctrine of ministerial responsibility looks to individuals—namely ministers and their departments—to accept responsibility for government decisions. Who would be accountable if they and their departments only have partial control over decisions? Who would be accountable if, one day, AI-driven government policies or programs were to go rogue or decide that it is much better to strike decisions by leaving ministers and public servants outside of the decision-making process? More to the point, what would be left of our Westminster parliamentary system if ministers and their departmental officials are not the ones making decisions?
Leaving AI and the government to their own devices is not without risks. It has the potential to marginalize further the role of MPs. AI can considerably strengthen the hand of the executive by leaving MPs unable to understand how AI and its algorithms generate policies and program decisions. It is not even clear how senior departmental officials would work with AI once it makes the transition from passive adviser to an active agent. It is conceivable that, in 50 years, Canadians will look to Parliament as little more than a functioning museum.
Parliament is at the centre of the fast-approaching storm, and it needs to explore all facets of AI and its transformative impact on both it and government. Parliament needs to develop guardrails to preserve human oversight as government goes about implementing AI. It also needs to define the role it should play in holding ministers and their departments to account as AI becomes further integrated in the government’s policy and decision-making processes.
This calls for a non-partisan review of how to equip Parliament with the tools to incorporate democratic values, transparency and public participation in the affairs of state under AI. Given that Senators do not need to deal with the political crisis of the day or partisan considerations, the Senate is in a position to provide a sober assessment of AI, its impact on traditional accountability requirements and to answer how Parliament can maintain a degree of human oversight and accountability, what role Parliament will play in the gradual disempowerment of humans in shaping public policies and delivering government programs and services and what Parliament needs to work with AI as it becomes an active agent.
