Could robots do better than our current leaders?

The Fourth Industrial Revolution is expected to wreak havoc on labour markets, with AI and robots replacing various white-collar jobs. One job category largely excluded from scientific reports is that of government leaders, despite being one of the most critiqued, scrutinized and ridiculed jobs of all.

However, commentators from countries as diverse as India, the UK, New Zealand and Japan have started to suggest that robots as government leaders could drastically improve decision-making, by being much less irrational and erratic than their inherently flawed human counterparts.

Evidence tells us that emotions drive politics and that voters appreciate "likeable flaws", as experts stressed over and over again when explaining the unexpected election (and subsequent actions) of US President Donald Trump. Still, wouldn’t "robogov" be far superior when making complex, high-stakes policy decisions?

While being governed by elected or appointed robots, cyborgs or algorithms may sound like science fiction, the advantages of robogov should still be obvious. Being far less hampered by ideological extremism and tunnel vision, or by the egotism and narcissistic tendencies that seem to characterize many government leaders, robogov may be able to make rational, fair, well-rounded and evidenced-based decisions.

In doing so, robogov enjoys both higher speeds and wider reach than any human leader, due to its superior data analytics capability, as well as unlimited energy and "stamina" - one of the most debated factors in the last US Presidential election campaign.

Robogov will also be much less susceptible to corruption and unethical behaviour. For these reasons, we already entrust AI with deciding on complicated social policy issues, from qualifications for unemployment benefits to targeted services that suit differentiated citizens’ needs.

So why should we not also let robogov take charge of the highest political and administrative realms? We could. One day, we may. But we should pay serious attention to at least three potential shortcomings.

First, although well-constructed algorithms maximize the immense potential of AI for deep learning, its usefulness will be substantially curbed in contexts of imperfect, contested, morally ambiguous or partly hidden information.

These are exactly the kinds of contexts in which major policy decisions and matters of diplomacy must be made. This is also what makes them messy and rife with incompetence. Whether machine learning is able to work its way through this thicket of uncertainty remains an unanswered question. And just imagine one side of the table being "robotized" while the other is not. How will decisions be made in such circumstances?

Second, unlocking the potential of robots and software requires considerable infrastructural and educational capabilities. In most societies, such capabilities are unequally distributed amongst citizens, if they are within reach at all. The current tech boom may well worsen existing gaps. The rise of AI could also perpetuate inequality in areas such as gender, when its programmer or operator is inherently biased in the first place.

Third, making more extensive use of data and machines in sensitive decision-making also creates additional security risks such as data leaks, cyber-attacks or computational errors, the dangerous implications of which for global stability may outweigh the benefits of replacing hot-headed and ill-tempered humans. While human interaction can produce disastrous outcomes, it might also be more capable of quickly resolving disputes when needed, as evidenced by historical examples ranging from the Cuban Missile Crisis to the recent Trump-Kim Summit. (The latter had been preceded by nuclear threats just weeks before.)

For the time being, it seems neither possible nor optimal for robots to replace government leaders, despite the clear imperfections displayed by the latter group. Moreover, they are still the most justified actors for safeguarding human society against some of the potential detriments of automation itself, at least in democratic systems.

However, taking up that role effectively does require government leaders to examine, understand and master the pros and cons of data and robotics, to prevent embarrassing episodes such as the recent Senate hearing of Facebook’s Mark Zuckerberg. Leaders must also be much more active in debates about the regulation and legislation of AI ethics, transparency and accountability, rather than leaving this to their private sector counterparts.

Ultimately, a more realistic and desirable scenario is one in which AI and automation are neither competitors nor substitutes to humans, but tools that government leaders can engage effectively and sometimes defer to, in order to make better, fairer and more inclusive decisions. Cynics will point out that authoritarian regimes already misuse robogov to do the exact opposite. The rapidly expanded use of AI by governments for unprecedented surveillance, public naming and shaming, and social discrediting of citizens raises serious questions of this nature.

Perhaps there is reason for optimism to be found in another much scrutinized, high-stakes public decision-making arena: sports. Tech-based decision-making was met with resistance for years, with opponents using "likeable flaws" as well as ‘big brother’ type arguments. However, during the 2018 FIFA World Cup, we witnessed powerful decision-makers visibly sharing their power with technology by using the video assisted referee (VAR system), supported by Hawk Eye technology.

The response from fans has been broadly positive, although some Croatian supporters may beg to differ following a contentious decision in a World Cup match against France, exactly because the electronic referee supported the human referee to do a better job. The game itself, the decision-makers and the key stakeholders all benefited as a result. Hopefully, we’ll soon witness the same in the world of politics and policy, with higher trust in government and increased civic participation as a result.