Hong Yu Liu is a PhD candidate at University of Cambridge in the United Kingdom. In this article he introduces the debate related to the impact of artificial intelligence on the labour market, and highlights the importance of understanding the labour process and the socio-technical context in studying the technological impacts of work. He argues that more empirically-based studies are needed for this intellectual debate and policy development concerning AI in the future.

Work in the time of the “Fourth Industrial Revolution”

Over the last decade, governments, companies, and other organisations around the world have been pouring significant sums of money into developing artificial intelligence (AI), with increasing discussion and applications of AI across different business sectors and in everyday life. According to a report produced by The Organisation for Economic Co-operation (OECD, 2018), since 2011, more than $50 billion USD has been invested in AI start-ups worldwide, with the amount of private equity invested doubling over the period 2016 to 2017. Another study by international auditor PwC (2017, p.12) predicts that the global GDP will increase by up to 14 per cent in 2030 due to the new economic opportunities that AI brings. In 2018, the UK government announced a £1 billion AI investment project to build a new research centre for data ethics and innovation and to provide specialised training for computer scientists.

For some analysts, the development of AI in recent years have begun a revolution that is fundamentally changing our way of life, work and, indeed, the basic structures of human relationships in societies. The World Economic Forum (WEF) shares this belief when it posits that the convergence of AI, robotics, Blockchain, Big Data and other technologies is bringing about a “Fourth Industrial Revolution” to the globe, a revolution which may have a profound impact on economies, societies and the lives of every individual on the planet (Schwab, 2016). According to Klaus Schwab (2016), the Chairman of WEF, the “Fourth Industrial Revolution” is building on the use of digital and information technology (what he refers to as the “Third Industrial Revolution”, see table 1 below), and it is characterized by a fusion of technologies that is blurring the lines between the physical, digital, and biological spheres of life. Schwab is confident about the “Forth Industrial Revolution” that it will “craft new models that create opportunities for all”. (SCMP, 2018)

Table 1. Four Industrial Revolutions as posited by the World Economic Forum

While some writers have questioned the “revolutionary” nature of this transformation (Boyd and Holton, 2017; Atkinson, 2019), the notion of a “Fourth Industrial Revolution” has become something of a buzzword in policy documents today, as we can see, for example, in the German policy document “Industrie 4.0”, in its focus on the increasing digitisation of manufacturing processes (European Commission, 2018).

In recent years, social scientific research on AI has often revolved around the use of algorithms (a set of instructions that a computer or smart device can execute), and many of the discussions concerning AI and algorithms are concerned with whether or not these technologies will worsen inequality by replacing human beings in the labour force, which grows the divide between the “precariat” and the privileged in labour market. Whilst the idea of machines replacing workers and taking away human jobs is nothing new, researchers have re-introduced this debate by suggesting whether or not the technological impact on work is different this time.

3d rendering of robotic arms with empty conveyor belt.

For instance, Ross Boyd and Robert Holton (2017) are sceptical towards various claims that the impact of robotics and AI on society is as transformative as previous industrial revolutions. In their analysis, they believe that robotics and AI are just the latest in a series of technological changes and significant, but not necessarily transformative. To them, the claim that our society is undergoing a “Fourth Industrial Revolution” is questionable because steam and electricity (the “First” and “Second” Industrial Revolutions, according to Schwab [2016]) had revolutionary power only after they were widely available. It is uncertain whether or not AI will be available to such an extent in both developed and developing economies in the near future.

Researching the Impact of AI on Work

While such debates continue among social theorists and technological determinists (Ford, 2015; Boyd & Holton, 2017; Wajcman, 2017), some writers are keen to predict the future of work by attempting to ascertain the potential impact of AI on the labour market, particularly when it comes to the role which new technologies will play in labour-capital relations (Brynjolfsson and McAfee, 2014; Spencer, 2017) and what is referred to as the “technological unemployment hypothesis” (Frey and Osborne, 2013; Arntz et al., 2016). However, their work is mostly theoretical or purely based on economic calculations, and does not consider the socio-technical context of work or the political economy of labour production. These theorists also fail to demonstrate any understanding of the labour process and provide industry-specific accounts on the technological impacts of AI and other new technologies concerning, for example, the cost of their implementation or their social acceptability in the workplace.

Up till now, there has been much academic discourse on the use of AI and robotics at work. While the focus has been on the relationship between automation and manual labour (Brynjolfsson and McAfee, 2014; Graetz & Michaels, 2015; Lüthje, 2019), Lene Pettersen (2018) provides a theoretical discussion on the power and limitations of AI as it relates to knowledge workers. According to Pettersen, AI systems are unlikely to replace knowledge workers because these jobs (such as consultants) depend on “knowing” and intuiting the situation (rather than the regurgitation of simple items of knowledge), and this “knowing” and intuiting is difficult to code and programme because it is “processual and relational aspects that are situated in historical, social and cultural contexts”. (p.6) “Knowing” does not, therefore, involve generic or universal answers. Rather, it involves and depends on a given context and meaning related to people and values.

I agree with Pettersen in theory, but it is uncritical to expect that all jobs in the service sector will experience the same vulnerability in the future due to technological encroachment. At the same time, it is noticeable that AI is increasingly being used in the service sector for labour control and management, especially when it comes to jobs in the gig economy. Tech companies have redesigned and routinised service jobs, and use algorithms to replace the role of managers to increase efficiency, calculability, predictability and control (Lee, Metsky and Dabbish, 2015). In this connection, many researches (Lehdonvirta, 2018; Sun, 2019; Veen, Barratt and Goods, 2019) have pointed out that the algorithmic control has caused information and power asymmetries between capital and labour at work and limited the agency of workers to make decisions for themselves. One case that powerfully underpins the growing criticisms of such stringent algorithmic control at work, as Sun (2019) elaborated, is that of deliverymen in Beijing who work for extremely long hours because algorithms distribute tasks unevenly and unpredictably.

There are also serious concerns in the AI research community regarding the loss of human agency and control of people’s lives in the future (see, for example, Ford, 2015), as intelligent machines are becoming more capable than human beings in ability and efficiency in both manual or cognitive tasks. As a result of such productive superiority on the part of AI, people may potentially become more dependent on it and its autonomous power (Anderson, Rainie & Luchsinger, 2018). According to a survey by the Pew Research Center (2018), 37% of AI researchers from academia and industry think that advancing AI technology will not “enhance human capacities and empower people”, and believe that most “people will not be better off” than they are today (p.4). They are also convinced that people may need to sacrifice control of their lives (such as independence and the right to privacy) because of the perceived advantages gained from digital tools, including efficiency, convenience and superior pattern recognition in the future.

The Gig Economy: a view of the we work co sharing office space on Bishopsgate, in the City of London. Image: Chrispictures / Shutterstock.com

On a more positive note, despite the tight control regimes exercised by the algorithms, researchers have discovered that such control can be resisted through individual and collective action and resilience. Studies on workers in the gig economy have some workers to be introducing self-imposed targets and ignoring the financial incentives calculated by the algorithm; others are developing their own network to share information outside the platform and redistributing tasks themselves (Veen et al, 2019; Sun, 2019). Whilst existing literature about AI and its potential impact on work is insufficient and relatively fragmented, social scientists should collect more evidence to understand how humans and intelligent machines collaborate in the workplace, and to provide an empirically-based foundation for intellectual debate and policy development concerning AI in the future.

Banner Image: Alex Kotliarsky, Unsplash