In early 2013, a colleague and I attended the inaugural meeting of the UK Robot Ethics association. There, we suggested that developments in robotics and computing technology meant that we needed to re-evaluate some of our economic thinking. Machines were now increasingly capable of replacing human cognitive power as well as physical power, as had primarily been the case in the past. There is an orthodox idea in economics according to which increases in productivity driven by technology will not create long-term or ‘structural’ unemployment. Conventional thinking has it that as technology-driven productivity increases expand the economy, new jobs will be created. And indeed, this has been the case historically.
Machines are now increasingly capable of replacing human cognitive power as well as physical power
We pointed out that while in the past the automation of traditional physical-labour-intensive jobs had led to the expansion in the labour market of jobs requiring cognitive rather than physical force, this time around the automation of human cognitive power would leave us struggling to figure out what capacities we had left to exploit in the labour market. There are three obvious answers: really high cognitive function roles, such as computer programming; emotional work like therapy and some care roles; and jobs requiring a great deal of physical human-to-human contact like physical or massage therapy. But these are relatively niche parts of the labour market. When machines replaced horses as the main suppliers of power for the transport industry, horses did not vanish altogether from the economy, they simply became confined to very niche areas of it, namely recreation. Could something similar happen to human beings in the age of intelligent machines?
At that meeting in 2013 we were politely ignored by the organisers and gently mocked by precocious engineering post-doctoral researchers, who endeavoured to teach us a bit of economics 101. However, when we queried these bright young minds about what roles humans might continue to play in an economy where machine power could increasingly replace human power both physically and intellectually, the answer also seemed to refer to niche parts of the economy and often to artisanal baked goods. Sadly, banana muffins do not an economy make.
Do androids dream of a 9-5?
Fast-forward to today and the story is completely different. Not only in academia, which is usually slow to adapt, but also in the mainstream media, articles about robots stealing our jobs appear at such a frequency that one could be forgiven for suspecting that they were written themselves by highly efficient intelligent machines.
The idea that automation could lead, in the not so distant future, to large-scale unemployment has been discussed by Paul Krugman, Paul Mason, and many people not named Paul (like Moshe Vardi). It is an academic paper written by two Oxford economists, Michael Osborne and Carl Frey, that has perhaps had the largest impact on the debate. Their paper, ‘The Future of Employment: How susceptible are jobs to computerisation?’ argued that 47% of American Jobs were at risk of automation in the near future. Even if their findings are off by half, it would still be enough to raise the prospect of a major social and political crisis.
Political resistance to automation has been weakened by the decimation of labour unions
Critics of Osborne and Frey’s findings have pointed out that their methodology does not take into account job growth as a result of increased economic productivity, the possible impact of falling labour costs, or political resistance to automation. These are good points, but may not carry the kind of weight needed to comfort us. As Erik Brynjolfsson, Andrew McAfee, Paul Krugman, and the global consultancy McKinsey have pointed out, US dept. of Labour statistics show us that the link between productivity increase and job growth in the US economy has been seemingly severed since about 2000.
Political resistance to automation has been weakened by the decimation of labour unions in the United States and Europe. And as technology grows cheaper, the amount of flexibility in labour costs will diminish (McKinsey predicts that ‘Advanced robotics, for example, has the potential to affect $6.3 trillion in labor costs globally’). Any major further reduction in labour costs will likely have a negative impact on the capacity of national governments to provide the basic services of the welfare state.
Some see a silver lining in all of this. A proposal that is increasingly gaining traction is that job losses can be offset by the introduction of a universal basic income (UBI). Each citizen would be paid a basic income by the government regardless of whether they worked or not. Any job that a citizen might have would be income above and beyond the basic income.
There are many versions of basic income theory. It’s an idea that for various reasons is embraced by many on the left and the libertarian right, though often in very different ways. Basic Income pilot studies have been conducted around the world, usually with successful results. The dreaded lack of incentive for work that many critics of UBI worry about does not materialise. Given the greater economic freedom offered by a basic income, people don’t tend to become lazy dossers, but instead work differently and in ways that could easily be considered better. There are currently pilot projects planned in the city of Utrecht (Netherlands) and in Finland, where there has been some criticism that rolling all existing welfare payments into a single basic income will leave some of the most vulnerable recipients worse off.
Testing the law of unintended consequences
I don’t have any objection in principle to the idea of a basic income. Indeed, I agree that it may be the best approach to tackling the social and political fallout of automation. There are however some issues that I think need to be investigated, alongside the big question of how to pay for it. I’ll briefly mention three of those here.
Firstly, we live on a planet of finite and increasingly dwindling resources. The replacement of human labour by machine power will still require energy resources, although there are likely savings to be made. A driverless car or lorry, for example, may have an overall lower energy requirement than a lorry or car with a human driver. However, increases in productivity will also likely require increases in energy requirements. A population freed from labour by basic income will presumably still want to do things, and these things will require energy. In a world where humans engage in leisure while machines labour, energy requirements will very probably place an increased strain on the environment. UBI as an answer to automation does not address the ecological challenges posed by the growth-model of the economy.
Secondly, people don’t work solely to fill their stomachs. In modern industrial and post-industrial societies, we very often construct our social and political identities in large part from the work that we do. The German political philosopher Hannah Arendt famously made a distinction between labour and work. Labour is what we do to survive, but work, which goes well beyond its narrow traditional meaning, is what gives our lives meaning. Work is how we build a meaningful world around us. The problem that Arendt pointed to was that in modern society work had largely been reduced to labour. This reduction means that many people lack identity and indeed public meaning-creating activity in their lives outside of their job and profession.
We very often construct our social and political identities in large part from the work that we do
Freed from the necessity of labour and jobs in a traditional sense, I imagine that many people will not so easily find work that will give their lives meaning. We need not simply accept this situation: it’s perfectly feasible that new social structures could be created that allow people to find public identity and meaning outside of the frame of traditional employment. Some of these already exist, but not nearly on the scale that would be needed to address the challenge of a jobless society.
A third reason to be concerned about job automation beyond the loss of income has to do with political power. The political power of the working classes, of the majority, has nearly always been based on the power that the workers had to slow or stop the economy. Where political and economic concessions that have improved the lives of working people have been won, it has happened not out of the kindness or principles of capitalists, but because working people had the power to bring the economy to a halt.
In breaking the power of the labour unions, political leaders in the west have already greatly degraded the political power of working people. One of the few areas where labour unions are still able to exercise some power is in transport. Rail workers, lorry drivers, and port-workers are still able to leverage their political power through strike action. This is, coincidence or not, one of the areas where automation looks to have a huge impact.
Were it not for the power wielded in the nineteenth century by organisations like the Birmingham Political Union, whose members were drawn largely from the working class, the franchise may never have been granted to non-landowners. This power was in large part based on a fear of political and economic unrest – fear of striking workers. The basic point is this: there is a fundamental relation between the capacity of the working classes to slow or stop economic activity and social or liberal democratic governance as we know it. The loss of this power by the majority has already had and will very probably have further negative impact on our democratic institutions. Universal basic income is an idea whose time has come in no small part due to the threat of job automation. However, it is an idea, which if it is to be implemented, will require a major rethink of major facets of our current societal arrangement.