Introduction
As we emerge from a global pandemic that has prompted a rethinking of how we work, with what kinds of technologies and how we value different jobs, many previously well-established ideas about the future of work have become less certain. Artificial intelligence (AI) and its impacts on work was a pressing topic before all the change wrought by COVID-19, and has been brought to the fore as much work rapidly digitised under lockdown.
As part of the British Academy and UCL project on AI and the Future of Work, we have been holding discussions and interviews to explore how AI may impact work experiences. We have discussed a variety of interdependencies that complicate simple narratives of “upskilling” people for the AI age or “levelling up” particular regions. Through our conversation, we developed some ideas for conceptualising the interconnected issues that could be influenced by AI developments in the world of work, and made some suggestions for how to break into the chain of interlinked problems we face.
What work does AI impact?
While it seems that many human manual and caring abilities are unlikely to be performed by AI any time soon, this technology has been predicted to impact a wide range of professions and services such as health, social care, education, and the criminal justice system. But in many of these cases, AI will more likely offer “augmented intelligence” for professionals to work with, rather than replacing them.
When AI is brought in to change how people work, there are often assumptions made about it taking away “boring” tasks and freeing up workers to do more “interesting” things – yet sometimes we see the opposite happening. What is in fact needed is an in-depth analysis of the lived experiences of workers, to establish who enjoys which aspects of which jobs most, versus what kinds of work are really and truly experienced as dull and unrewarding, so that AI development can focus on the latter. Instead of a focus on what should we automate, with respect to the human experience of work, there has been more emphasis on what can we automate, with respect to what AI is good at now.
In addition to augmenting work, AI does undoubtedly have the capacity to replace it in some cases. The impacts of automation in some sectors, and its absence in others, in combination with augmentation of professional work, can be summarised by the term “hollowing out”: there will be highly skilled and well-paid jobs, possibly in increasing numbers; there will be a continuation of typically lower-paid “hands-on” jobs. There is also an evident increase in mainly low-paid jobs generated by, for example, the automation of large parts of retail, producing a demand for delivery drivers. But in the centre of this spectrum, many jobs will disappear through automation – there are estimates of up to 50 per cent of current employment being vulnerable in this way.
This brings us to another very important issue – liveable wages and employment relationships that provide security. Not only do we need to consider AI’s impacts on the qualitative aspects of what makes working experiences enjoyable, we cannot ignore the financial fallout for those susceptible to the hollowing out process. There are evident challenges in this kind of future scenario for ensuring that there is an economy that delivers good work with good incomes that enables people to lead good lives. This is where we must examine the interdependencies.
Interdependencies
The challenge of providing good work and good incomes for a large proportion of a population is well known. How well established these issues are can be seen in the “A Century of Cities” study, which shows that those cities that were “poor” in 1911 were usually still poor in 2011. The “poorness” in these analyses is underpinned by poor work and low incomes. This is the basis of the “levelling up agenda” in current parlance – but it is an old problem. Work is central to any analysis of this challenge and in the future the impact of AI will be a key factor. But before we can consider the impacts of AI, we must understand the existing interconnected factors into which work fits now.
A simple example of this chain of interdependencies is the argument that if there is what appears to be a housing problem, it is because there is an income problem; there is an income problem because there is an employment problem; there is an employment problem because there is a skills problem; there is a skills problem because there is an education problem. When we look at the impact of AI on the future of work, therefore, we have to take this kind of chain into account. This can be further broadened; AI will have a major impact on the future of work but so will climate change, pandemics, changing trade patterns and other dimensions of socio-economic change. There is a network of interactions between drivers.
Future policy
Firstly, acknowledging that AI is developing within a context already fraught with inequalities is important. It is also important to think about the ways AI might exacerbate existing problems alongside other emerging issues. As such, working on AI policy cannot be done in isolation. One key recommendation arising for policymakers from our conversations is that there needs to be more cooperation between departments – central, regional and local – as while they may be working on different goals these are, in fact, connected.
Next, acknowledging the vast interconnection of forces at play does not make us helpless in addressing the challenges of work, technology and equity. It is not the case that technological, economic, climate, health and other changes are inevitable; there are political choices to be made that will impact how these play out. Therefore the emerging picture of chains of interconnected problems poses the question – where can we break in?
AI is likely to automate and augment work differently across different regions, with many areas in the former industrial heartlands now emphasised in the “levelling up agenda” predicted to be impacted the most. In these areas, the chains of related problems make the success of any attempt at “levelling up” dependent on its ability to effect change across the chain.
Let’s take as an example the chain we have identified: a housing problem; due to income problem; due to an employment problem; due to a skills problem; due to an education problem. By understanding these problems as interrelated, we can in fact identify where we may be best able to break in and effect change that can impact all the problems in the chain. In this case, we argue that education must be the starting point, not merely skills, as has been the framing in current policy. This means schools, but also further eduction colleges and universities. Collaboration across levels will be needed to provide education for people to work well in future careers alongside AI. Understanding education as the starting point for breaking into this chain of problems means that teachers are crucial. We need to start from valuing teachers, training teachers and will need to work together with teachers if we hope to solve the problems in this chain.
What do we mean by contrasting education with skills? Education implies a broader base of capabilities and we argue that critical thinking should be a central focus among these, as well as an emphasis on creativity. These have been identified in our conversations as needed for “future proofing” people for new ways of working that will emerge as technology continues to develop and the world continues to change. Education also implies an ongoing, lifelong developmental process, not a clearly delineated set of goals to be met and moved on from.
Critically, education that helps people into good work has to permit flexibility. As we are seeing in the current concerns about jobs being hollowed out, many careers traditionally perceived as stable and secure (e.g. finance and law) are likely to come under threat as AI builds ever stronger language-processing capabilities. This highlights how trying to predict which jobs will be secure in the future – and thus aiming to train skills for those types of work – is a poor strategy.
Indeed, coding skills have been put at the foreground of the government skills agenda. And yet, many coding jobs may in fact offer very mundane, repetitive work. Furthermore, developing AI that can produce code is also a goal for many big players in the field, with existing approaches already able to create overarching structures that merely need a human to fill in some blanks, and aim to let users develop programmes without ever needing to code themselves. This suggests that coding work may be just as susceptible to hollowing out and thus training basic coding skills is not necessarily an avenue to future good work, either in terms of the qualitative experience or financial security.
We are not arguing that people should not learn to code but rather that a broader educational approach encompassing both STEM and arts subjects is needed. For instance, numeracy skills are extremely important, but an education in maths is something very different, opening avenues for new ways of thinking and offering continuous challenge. Likewise, good work is challenging, interesting and creative. Education should reflect this. Furthermore, in order to be able to adapt and retain good work in unpredictable future circumstances, a high level of flexibility is needed. Therefore, it is up to policymakers, working with teachers, to ensure ongoing and flexible access to education.
A future of good work?
The COVID-19 pandemic has shaken up many assumptions about what work must be like and AI development continues to reshape the world of work – and its impact will only accelerate. This offers us a moment of potential, with heightened public awareness of different possibilities of the future of work and opportunity to steer AI development in work in better directions. AI can both replace and augment different kinds of work, and it is important to make political decisions that will mitigate the worst risks and ensure good work for people.
The UK government “levelling up” and skills agendas are narrowly focused on providing training and jobs in specific regions of the country, often regions where AI is predicted to negatively impact work. We argue that the interconnected chains of problems that have generated the inequities that AI could exacerbate can best be broken into through education. We argue for coproduction with teachers, across educational levels, building towards an educational approach that develops critical thinking and creative capacities. This is more likely to provide good work which is challenging and empowering, and also more likely to provide flexibility to adapt to technological change thus also yielding good work which is secure.
AI has tremendous potential to improve our working experiences, but AI isn’t developed in isolation – it intersects with all of the problems we already see in society. A future of good work alongside AI requires us to tackle these complex, interconnected, longstanding problems.
Helena Hollis is a UCL PhD researcher in Information Studies, and is also working on the UCL and British Academy project on AI and the Future of Work. Sir Alan Wilson FBA, FRS is Director, Special Projects at The Alan Turing Institute.