AI offers us an opportunity to rethink how we ought to work in the future. It has forced researchers and policymakers to confront facets of work beyond levels of employment, such as quality of jobs, precarity of work, or managers’ power over workers. But, there is often an important moral lens missing from these debates. AI is an opportunity to try for more egalitarian workplaces, labour markets, and societies, both domestically and globally.
Inequality is on the rise globally, and AI threatens to make inequality even worse, from increasing inequalities between capital and labour to more precarious and unsafe working conditions for many workers in the AI supply chain to worsening environmental impacts that harm future generations. But we, collectively, can do something about this: inequality is created and sustained through policy and organisational choices, and can be redressed through those same mechanisms. Governments need policies that put equality front and centre in work of the future.
Labour market policy, labour law, and businesses should prioritise:
1. AI that enables people to have an equal opportunity to develop core capabilities across their lives.
2. AI that increases egalitarian social relations among members of different social identity groups.
3. Policy interventions should decrease inequalities in power over the development and use of technology.
AI and equal capabilities
The first egalitarian principle that should guide policy is equal access to good work. What, however, is good work? One account of good work focuses on people’s preferences: good work is whatever people prefer, where they make choices free from coercion and manipulation and with roughly equal bargaining power (Parr 2024). Unequal bargaining power and adaptive preferences, however, pose a major problem to using this approach for current policy design: people often choose the work they can get, and form their preferences in circumstances where they know they don’t have many good options.
Instead, academic and policymakers usually identify core characteristics of good work: the opportunity to develop complex skills, make a social contribution, work on terms of social equality, and so on. While AI is often a threat to those goods of work, it also offers us an opportunity. For example, there has been recent excitement about the possibility that generative AI can level the playing field at work by increasing the performance of lower-skilled workers (Brynjolfsson et al 2023; Noy and Zhang 2023; Peng et al. 2023). If generative AI increases worker performance by helping them develop new skills, then a more widespread use of generative AI could help equalise workers’ opportunity to develop the core capabilities needed to flourish, which is important from the perspective of justice (Sen 1999).
However, AI will not lead to a more equal distribution of capabilities without policy intervention. Policy should target at least three areas. First, it should limit the prevalence of what Acemoglu and Johnson (2023) call so-so automation, or automation that reduces labour costs but does not increase productivity or provide workers new opportunities to gain important skills. Consider automated tills in the supermarket: they cut labour costs, but do not make groceries cheaper, nor provide a better shopping experience.
Second, governments should fund research on technology that complements human skills and increases the range of capabilities that people can exercise, as well as the necessary education to develop more human capital. For example, initial research suggests that common generative AI tools may harm, rather than enhance, learning (Bastani et al. 2024).
Third, governments should regulate the labour market so that individuals are able to develop a range of core capabilities across all the various activities of their lives. In modern economies, specialisation and differences in human capital mean that governments cannot perfectly equalise access to good work for all. Instead, governments should focus on regulating the worst work, and ensure that it does not impact people’s ability to develop and exercise core capabilities outside of work. And, governments should craft policy to minimise the amount of time that people spend in low skilled work, in order to free up time outside of work for them to access the range of core capabilities.
Inegalitarian social relations
The second egalitarian policy priority is the inclusion of marginalised workers. Policymakers can learn from philosophical research on egalitarianism to think beyond how AI will impact levels of employment to how it will impact the relations between different social groups. In other words, policymakers also need to draw on egalitarian principles for how individuals ought to relate to each other in crafting policy. For example, governments or employers should treat all members of the political community with equal respect and concern, which includes ensuring that members of different groups have access to the various goods enabled by work.
We have already seen that AI has contributed to discrimination in the labour market, and had negative impacts on marginalised workers. However, AI also has negative impacts on marginalised workers. One example are the impacts of AI on so-called hidden workers, or workers who are willing and qualified to work but are overlooked in hiring (Fuller et al 2021). They are drawn from heterogenous social groups, from caregivers, military veterans, those with disabilities, or recent immigrants. A major reason they are overlooked is that they are not recognized by automated recruiting systems as good candidates (Fuller et al 2021). Various features of how AI models are developed can lead to so-called algorithmic monocultures, which cause the same individual or group to experience negative outcomes across different AI recruitment systems (Bommasani et al 2022). Labour market regulation should aim to reduce algorithmic monocultures.
AI can also play a positive role in promoting gender equality in the workplace. Technology has the potential to reduce the gendered division of labour, or the differences in lifetime earnings, the level of status and power in the workplace, and the level of other goods of work between women and men. A significant cause of the gendered division of labour is the so-called wage penalty for flexible work or the motherhood penalty, which is a disproportionate decrease in earnings for flexible and part-time work that is preferred by women, especially by mothers (Bertrand et al. 2010). Technology has the potential to decrease the wage penalty for flexible or part-time working, by increasing the substitutability of one worker for another (Goldin and Katz 2016). The call for more flexible work, however, is not a call for stressful, precarious gig work subject to managerial surveillance and wage theft (Kaldolkar et al. 2024). A careful crafting of policy is needed in order to avoid a further increase in bad, precarious work enabled by AI.
Inequalities in power
The third egalitarian policy priority is a redistribution of power over the development and use of technology. Power over technology is an important form of social power. For example, it gives owners of capital power over working conditions and the distribution of profit and other rewards from work (Acemoglu and Johnson 2023). Policy ought to target background distributions of power over technology and innovation, rather than correcting downstream effects like the distribution of profit between capital and labour. In addition, policy should target other forms of unequal power in the labour market, such as an inequality in bargaining power, which is an important matter of justice (Parr 2024).
It can be tempting to think that AI will help to reduce troubling inequalities in power on its own. For example, if generative AI tools increase the productivity of lower-skilled workers, this in turn could decrease wage inequality or inequalities in bargaining power between high and low-skilled workers. However, a decrease in inequality is not inevitable. For example, if AI automation creates competition between high and low skilled workers over tasks, this competition could actually increase wage inequality (Acemoglu and Restrepo 2022; Acemoglu 2024). The important lesson here is that policy and innovation choices matter, and that policymakers should target the distribution of power between firms, governments, civil society, and individual consumers directly, rather than hope that technology will lead to more equality without accompanying institutional change.
Kate Vredenburgh’s research on the future of work is supported by a UKRI Future Leader’s Fellowship [grant number MR/Y015975/1]; and the LSE Research Impact and Support Fund 2024.
Dr Kate Vredenburgh