Abstract
This thesis explores philosophical, ethical, and policy-related issues regarding the impacts of artificial intelligence (AI) on work.
First, I survey various relevant historical, economic, and philosophical literature on work. This establishes the groundwork for a growing field called the philosophy of work.
I argue in favour of a pluralistic account of work, contending that a (non-exhaustive) list of characterisations of work includes energy work, agreement work, income work, service work, economic work, good work, volunteer work, and goal work. All of these, I argue, are distinct and useful work characterisations.
Following this, wellbeing issues regarding work are considered. I show that all well-regarded philosophical conceptions of wellbeing tend to endorse, in whole or in part, a hedonic view of wellbeing. This hedonic view aligns with prominent theories and measures of wellbeing in psychology and economics, and thus established measures can be used with some confidence in assessing (the hedonic part of) wellbeing. The degree of wellbeing provided by various job elements (such as income, working conditions, work relationships, and even if one has a job or not) is shown to vary considerably.
Next, I consider the likely impacts of AI on work. AI, through the use of large language models, is very likely to have significant impacts on work and wellbeing, as well as churn in the labour market as a whole. The combination of AI and robotics could also have substantial influence on jobs in the future, as technology for doing manual economic work is improving quickly due to modern machine learning techniques. In order to consider policy options to deal with the impacts of AI on work, I illustrate eight possible future worlds which we might encounter, depending on the pace of AI progress and its capabilities, and depending on the level of economic inequality produced by various circumstances.
Finally, I argue that different world circumstances demand different policy approaches in order to best serve wellbeing as AI impacts work. For example, Universal Adjustment Assistance to better assist people in work transitions caused by AI, and offering support through job losses and training, may be sensible in a world similar to today’s, but are far less sensible in a world where AI quickly advances to being able to do a large portion of economic work more cheaply than humans. Where AI is able to do a large amount of economic work, a universal basic income or an income guarantee are far more reasonable policy options, accompanied by significant taxation, which can be afforded by significant increases in productivity from automation. Although advanced AI in work could cause some significant societal problems, many policy tools are available to deal with the impacts of these. Even if AI advances are largely unpredictable, having a multitude of policy options ready and available allows us to best deal with the possible impacts of AI on work in different circumstances, so that policy is able to keep up with a fast-changing work landscape.