Introduction

Work is a necessary part of human life. While historical, cultural, and social understandings of work change over time and place, human beings have always had to work. Whether it be the work of surviving against the elements, the work of raising a family or serving a community, or the work of paid employment as we understand it today, work has always been a necessary part of life. Technologies have been developed to ease and expand work, in almost all of its forms, but human actants have remained at the centre of the activity in all its forms to date. Many scholars now argue that the necessity of work may be coming to an end, with the invention of Artificial Intelligence (AI) and automated technologies offering the first moment in history in which all human beings might be freed from the necessity of work. Historically, freedom from work has been a luxury of the few, those able to offload their responsibilities to paid workers, willing family members, or subjugated slaves:

The “good life”, as Aristotle called the life of the citizen… was “good” to the extent that by having mastered the necessities of sheer life, by being freed from labour and work, and by overcoming the innate urge of all living creatures for their own survival, it was no longer bound to the biological life process. (Arendt 1998, p. 36–37)

AI and automation therefore promises a life free from necessity for all. Whether this is a good thing, leading to the end of capitalist exploitation (Mason 2016; Alexander 2020) and into a post-work utopia of endless free time and boundless creativity (Dinerstein and Pitts 2018; Bastani 2019; Danaher 2019), or a worrying development, resulting in the loss of meaningful engagement and economic stability (Alter 2017; Peters 2017; Kessler 2019), is hotly debated (Frase 2016). Regardless of the position held, it is undeniable that AI and automation are going to radically change the nature of work in the coming years.

However, this understanding of work as a necessary set of activities that can be automated away from human involvement is an imperfect and limited concept that doesn’t account for the entirety of work. Perhaps most clear and pressing are the arguments presented by feminist scholars, including Weeks (2011) and Ferguson (2019), regarding the unequal and distinctly gendered inclusion and exclusion of activities in the economic sphere: much of the necessary work of society is done in the home and outside of the workplace, and thereby remains unpaid and unappreciated. Furthermore, understanding work strictly in terms of economic necessity that can be wholly automated without issue does not adequately represent the entire nature of automation at work (Smith 2020), and I argue that the proposed futures envisioned through such concepts are likely to result in unintended consequences not captured in the understanding of work as a form of necessity that can be wholly automated: whether this be the erosion of meaningful work (Blanchflower 2019) and the creation of increasing normality of degrading and demeaning jobs (Jones 2021; Mueller 2021); or the more dystopian predictions of a world where human meaning itself is lost to inhuman automated technological systems (Carr 2016). I argue that the arguments made regarding the removal of work as a necessity through automated and autonomous technologies, whether the post-work account or the technological unemployment account rest on two fallacious assumptions:

  1. 1.

    That activities of work are categorically distinct from activities of non-work, and that activities of non-work can be achieved without the human engagement in work;

  2. 2.

    That the technologies that remove work from the purview of human engagement will leave activities of non-work unchanged, thereby allowing human actants to freely engage in non-work.

I will argue here that these assumptions rest on a specific and narrow understanding of work as predominately tied to economic necessity, supported by a very narrow set of examples of automated and autonomous technologies. As a result, I argue that they do not constitute a robust enough understanding of the automation of work, and will likely result in unintended futures not desired by their architects if adopted to fully remove the human worker.

What I offer in response is a conception of work that is not tied to a specific set of activities, nor to a specific contingent set of conditions, but rather a binary set of modes of activity (labour and work) that I argue are relevant and applicable for considering the development of AI in work now, as well as conceiving the possible futures of a world without paid employment.

Work as Necessity

Work appears as a necessity in human life in a number of different ways. Perhaps most clearly across the literature today is the way in which work functions as an economic necessity: under capitalism, a person must earn a wage in order to survive, and therefore paid employment becomes “a social value and organising principle of modern ‘material life’” (Just 2017, p. 446). Work then functions as a proxy for other forms of necessity: earning a wage through work can be used to buy food, provide shelter, facilitate health and social care, and participate in society more widely. In many ways, it is this form of work-as-necessity that AI and automation is intended to redress: automated technologies are deployed in workplaces to produce higher volumes of goods in a shorter and more regulated amount of time, and AI systems achieve informational-based work and management at levels that human beings simply can’t, with both ultimately seeking to remove (or at the very least significantly diminish the involvement of) the human worker. On this first point most scholars agree: the inclusion of AI and automation in the workplace is lessening the economic necessity of work, as the technologies being developed are intended to perform the roles that traditionally fall to human workers (McAfee and Brynjolfsson 2014; Ford 2009, 2021).

However, the society that results from such an inclusion of AI and automation is radically contested across the literature. For scholars who hold a ‘post-work’ position, including Bastani (2019) and Danaher (2019), this lessening of human engagement in work is an inherently positive move. Economically necessary work is argued to be meaningless in its own right, actively preventing the engagement in meaningful activities of self-actualisation, and therefore the move to displace these activities onto automated and autonomous technologies could result in a world “beyond both scarcity and work” (Bastani 2019, p. 12), in which we could “retreat into “virtual” worlds that are created and sustained by the technological infrastructure that we have built” (Danaher 2019, p. 3). Others contest that the offloading of work onto autonomous technologies will not result in a utopia of free time and creativity at all. Scholars, including Carr (2014) and Ford (2015), contend that these trends are occurring within a capitalist system, and so the removal of our capacity to earn a wage will result in technological unemployment, and a loss of the ability to operate and function in society (Attfield 2001). Rather than freeing us from necessity, we’ll become compounded by our necessities, now unable to fill them as we cede control to automated technologies, by “[trading] subtle, specialized talents for more routine, less distinctive ones” (Carr 2014, p. 67) and facing “the potentially devastating impact of long-term unemployment and underemployment” (Ford 2015, p. 17).

Economic necessity is not a comprehensive means for understanding work as a whole, because many activities that undoubtedly fall into the category of ‘work’ are removed from the economic sphere, happening privately and without pay. The economic system of work is wholly unequal, both in terms of the economic valuation of work contained within economic necessity, and the activities left out of the economic sphere. This is true prior to the development of AI, with many forms of meaningless, repetitive, and menial labour falling to “women, political outsiders, and alien insiders” because these are “chores that one would rather have someone else perform” (Veltman 2010, p. 57). As such, “a [false] dichotomy between nature and culture” is borne out in which “reproductive labour does not lend life significance” (ibid.), and those that perform it are withheld from engaging in more meaningful pursuits while allowing those they serve to do so. This is true of both those activities that are poorly paid within the economic sphere, and those that are socially necessary but economically unproductive, particularly the activities necessitated by the family (Mullin 2005; Lewis 2019). Moreover, this is further compounded by the inclusion of AI and automation in economically necessary work, because it furthers divisions between paid and unpaid labour, while also further extending the alienation and inequality between well-paid and poorly-paid jobs (Eubanks 2019): offices and factories are fully automated, while parenting and reproductive labour remain predominantly in the hands of unpaid women and poorly-paid workers. Wholly automating economically necessary work might therefore be possible, but would not result in a post-work utopia in its own right, given that necessary unpaid reproductive labour would endure into the imagined ‘post-work’ future. Without significant direct attention the gendered inequality would pervade, creating free time for some and continued inequality for others, compounding the gender pay gap (Abdel-Raouf and Buhler 2020) into a fully-fledged gender work gap (Weeks 2011). Post-work might therefore instead be understood as ‘post-employment’.

It might be argued here that economic necessity is not a sufficient means of understanding work, and that instead we might consider work as a form of social necessity, where meaningful engagement in work beyond simply paid employment is need in its won right (Breen 2019; Harding 2019). As Yeoman (2014) argues, “work occupies a peculiarly ambivalent position – simultaneously valued for providing the means for self-realisation and disvalued for being burdensome and compulsory” (Yeoman 2014, p. 235). Work is a social necessity insofar as it “identifies and satisfies what is of profound importance for living a human kind of life” (ibid. p. 241). In this sense, the inclusion of AI and automation must be approached tentatively, to allow human workers to retain the capacity to meaningfully engage in work, and to draw meaning from it: on this account alone, the dream of a post-work utopia is a hollow one, because it is through both individual and communal work that we gain a meaningful human life (MacIntyre 2016).

However, many scholars argue that often jobs today do not fulfil this social necessity for meaningfulness, and that they only fulfil economic necessity. Yeoman argues that “the harmfulness of non-meaningful work is derived from its inability to satisfy the inescapable human interests to be able to experience freedom, autonomy, and dignity” (Yeoman 2014 p. 240). Graeber (2018) argues that while we have a desire to engage in meaningful work culture that is tied to self-actualisation, many jobs today are wholly automatable, fulfilling no purpose beyond earning the workers a wage. It might then be argued that the inclusion of economic necessity in the world of work is the only issue, and that removing the economic component (through any means, including Universal Basic Income (UBI), political revolution, or full automation) resolves the issue of boring, meaningless, and mundane work, while allowing us to create a utopia of free time, in which we can spend our creative and self-actualising energies as we see fit.

On this account, the removal of economic necessity does not therefore eliminate work as a whole, and indeed under the correct (non-economic) conditions the activities currently undertaken as paid employment might well be the same as those that are hoped will fill the post-work utopia. The work of the professional artist, academic, and athlete now might well be the activities that populate all of our lives when the factories, offices, and workplaces are emptied of humans and filled with autonomous and automated systems: the goal of post-work might therefore be that “the current organisation of work should be changed, not that we should ‘abolish work’” (Deranty 2021, p. 436). Marx’s famous dream of being able to “hunt in the morning, fish in the afternoon, rear cattle in the evening, [and] criticise after dinner” (Marx 1846, p. 54) is therefore only an ideal now because we must continue to toil in economically necessary work that prevents us from achieving it. On this account, it is precisely because the economic necessity of work has been eliminated through full AI and automation that we are free to enjoy the social goods of work as we see fit in a post-work (or post-employment) future. The issue here is therefore not necessarily work itself, but the conditions under which it arises today.

Two Embedded Assumptions

I argue that approaching work in this way can be formulated as follows: due to global economic and social conditions work appears as a necessity now, but the economic component can be removed while still allowing for the social goods of work to be achieved. It therefore follows that autonomous and automated technologies can be developed to remove the necessity of work, allowing human actants to freely engage in meaningful activities of their choosing.

Within this claim, which I argue is implicit or explicit in both the post-work and technological unemployment literature, are two erroneous assumptions:

  1. 1.

    That activities of work are categorically distinct from activities of non-work, and that activities of non-work can be achieved independently of any work, or at least human engagement in work;

  2. 2.

    That the technologies that remove work from the purview of human engagement will leave activities of non-work both possible and unchanged, thereby allowing human actants to freely engage in non-work.

Let’s consider each assumption in turn.

The first assumption has two components: (1) that activities of work are categorically distinct from activities of non-work, and (2) that one can be achieved without the other. I argue that a distinction between work and non-work is endemic in the literature of AI and work. Already discussed in this paper are distinctions between meaningful and meaningless work (Yeoman 2014; Graeber 2018), but other distinctions persist across the literature: Bernard Stiegler (2016) draws a distinction between meaningless employment and meaningful work; Autor (2015) distinguishes between work that can be displaced onto automated and autonomous machines, and work that human workers fulfil with the assistance and enhancement of AI and automation; Susskind and Susskind (2017) differentiate between those professions that will disappear with AI and those that will endure; and distinctions between work and leisure are common across the literature on both automation and work (Sayers 1998; Bastani 2019; Danaher 2019). However it is explained by any given scholar or account, a distinction is often drawn between work and non-work, either in the conditions that they arise in or the activities themselves.

This makes sense now: economically necessary activities fall into the category of ‘work’, and meaningful activities that can be unpaid (but can also be paid in certain circumstances) fall into the category of ‘non-work’. It is hoped that the latter category will endure after automation, and that the former will dissipate into machine functioning. But in the post-work or post-employment future, this distinction falls away completely, leaving only non-work. This might not seem problematic, but on closer inspection I argue that this is highly problematic: precisely because economic remuneration is distributed so inconsistently across activities of work, and because it is so embedded in the economic structures of today, the types of activities left after full automation might be very different to those envisioned. The biological and social necessities of childcare, parenting, and social care will endure, regardless of what jobs are left over (Strober and Donahoe 2017), and the gendered and social inequality pervasive across many activities excluded from economic remuneration may only worsen. John Danaher’s stark claim that “HUMAN OBSOLESCENCE IS IMMINENT (sic.)” (Danaher 2019, p. 1) and that we can build “virtual worlds” sustained entirely by technology (ibid. p. 3) neglects that houses need to be managed, families and children need to be cared for, and communities require attention: either the resulting future is massively unequal, and compounds contemporary inequality rather than redressing it; or it is wholly post-human, in which Danaher’s own distinction between “work… and life more generally (outside of work)” (ibid. p.2) disappears as well, leaving no space for humans at all.

The issue rests not only with those activities that fall outside of the economic sphere. Many forms of meaningful work today arise from meaningless activities, even if the conditions of each are different. The meaningless labour of the factory gives rise to political organisation (Hutchins and Harrison 1911; Cant 2019), and shapes the very political landscape that is hoped will survive the emptying of those factories (Srnicek and Williams 2015). Beyond politics, it is hoped that our creativity will be let loose with the ending of economic necessity, but this “rests on the fallacious reasoning that labour power, if not consumed in the maintenance of life, will nourish [non-labour] activities” (Veltman 2010, p. 71). By wholly automating certain sections of work, there will be an inevitable and unavoidable knock-on effect in the binaries of non-work. The creation of AI systems that can create beautiful artwork that passes for human-made might deter people from learning artistic skills themselves, rather than encouraging them (Roose 2022).

This directly leads to the second component of this first assumption, because engagement in meaningful non-work can be contingent on the continued human involvement in work. It is certainly the case that meaningless work can be achieved without meaningful non-work, and the opposite can be true. The most obvious example is in the culinary arts: machines can certainly perform all the meaningless washing, cleaning, and prepping work of a kitchen, leaving human actants to engage with the other related activities in a meaningful way. But if the sophistication of those same machines grows, human involvement in the activity as a whole may become redundant. The case of academic research is perhaps more relevant for the purposes of this paper: if AI systems remove the academic work that appears as economic necessity, there might be no room left for academic research as non-work in the post-work future. If AI systems are created to produce research papers to a higher and more regulated level, in a fraction of the time it takes a human academic, and can enter them into a field of other AI systems for peer-reviewing with a higher level of regularity, speed, and accuracy than any human could ever hope to, then the newly-unemployed human academic might be actively dissuaded and prevented from undertaking the same activity as meaningful non-work. While the hope might be to limit the burden of marking, paperwork, resource-creation, the same technologies could easily overtake the meaningful aspects of the job as well, invalidating human engagement in academic work and negating the need, quality, and opportunity for academic non-work. The goods of such systems might be advancements in a range of scientific and healthcare fields, but would inevitably reach the arts and social sciences, and perhaps eventually academia as a whole.

The issues with this first assumption are further compounded by the second assumption highlighted here: that even if the activities of work and non-work are held to be categorically distinct, it is further assumed that the technologies used to automate work will not significantly impact the non-work ‘left over’. This has already been shown somewhat with the ongoing debates about AI in academia (Bearman et al. 2023; Crompton and Burke 2023), but perhaps a clearer example is the self-driving car. Such artefacts are commonly discussed across the literature as being imminent technologies, which scholars positing that “Autonomous Vehicles will be on our roads soon” (Evans 2021, p. 7). In addition to their private usage, autonomous vehicles are being designed for commercial use in delivery and logistics (Weber and Kröger 2018). This will undoubtedly reduce the economic necessity of work in these industries: but when the roads are filled with self-driving cars for private, public, and commercial use, the opportunities for humans to engage in the activity diminish massively, because the technologies will have to be, by design and definition, safer and more effective than human drivers (Nees 2019). This will disincentivise human drivers, reducing the likelihood of human actants taking up related activities, such as sport racing or the artistic expressions that arise from driving (such as David Hockney’s 1980 painting, and subsequent guided drive with music, Mulholland Drive: The Road to the Studio). This may even result in a situation where human driving becomes prohibited, precisely because it is less safe than leaving it to AI systems (Kranak 2020).

A further problem arises from the case of the self-driving car: that of the cultural and geographical conditionality of automated technologies within work. The design and marketing of self-driving cars, and indeed much of the research surrounding it, focuses on the uses of the technologies in the cities and highways of developed, industrialised nations (such as the UK, US, and China). But the utility of these technologies can be questioned when the logistical conditions of transportation changes: in Vietnam, for example, motorbikes are the dominant form of transport, traffic is often congested, and pedestrians move in the road (Hansen 2015). Automating the economically necessary work of delivering goods might then be possible in a Western nation, where lorries transport goods overnight on pedestrian-free highways, but is much more difficult where deliveries are performed either on or alongside swathes of mopeds, winding through busy streets full of pedestrians. Not only is this a logistical issue, of deploying technologies in radically different contexts, but speaks to the limitations of homogenising technologies that ignoring the specific cultural instances of work, and indeed life more generally (Geertz 1973).

Again, with the example of driving this might be seen as a perfectly acceptable trade-off: if the technologies can fulfil the needs of the activity in a safe, reliable, and affordable way, while also removing both economic necessity and inequality in the related industry, then those affected will be left with free time and economic security. Moreover, the artistic expressions that arise from driving will be replaced with artistic expressions that arise from self-driving cars. However, I contend that this trade-off occurs in all instances of human-replacement through automation, with very different levels of acceptability. Beyond driving and academic research, the automation of activities relating to parenting, healthcare, artistic creation as economically necessary will inevitably involve trade-offs for the non-work counterparts that are much less agreeable. Indeed, the very activities of leisure and artistic expression that are hoped will populate a post-work utopia are fundamentally endangered when their work-counterparts are wholly automated, precisely because these artefacts will radically change the shape of the activity, to the detriment of the humans involved.

The Limitations of Necessary Activities of Work

The issues surrounding the displacement of work onto autonomous and automated technologies that I am highlighting here can be summarised as follows: the economic necessity of work can be removed by many means, including UBI and radical socio-political change, but is also being explored as a technological future made possible through AI and full automation. In replacing or displacing work onto technological artefacts, it is hoped that activities of non-work will be left both possible and available. I contend that removing the human worker through the technologies proposed will change the shape of the activities that they are interjected in, limiting the availability of non-work after the fact. I argue that this can be captured in three reasons.

Firstly, the principles by which fully automated and autonomous technologies are developed and deployed in work, as opposed to non-work, and imprecise and totalising, perhaps more so than is often acknowledged. If technologies are developed to ease hardship, alleviate mundanity and repetition, remove menial activities, and generally remove any meaningless or boring activities, then there will be inevitable ramifications for activities of non-work. Many meaningful activities, of both work and non-work, arise from some combination of boring, repetitive, menial, and arduous labour: artistic skills take time, repetition, and failure to perfect; academic writing takes editing, revision, and lengthy research; and excelling at sport takes difficult and repetitive practice and training. The automation of the tedious and meaningless office work through AI systems that write emails and process data might be the same kind of technology that make human academic research or creative writing redundant; in the same way that the technologies that empty the factory, building site, and logistics centre might also be of the same kind as those that make athletic and artistic endeavour precarious. Even if unintended, the principles of automation intended to remove economic necessity are an imprecise measure that will have unintended consequences without significant definitional work now.

Secondly, some activities that are undoubtedly meaningless when they appear as economic necessities have meaning when they appear under other conditions, and so wholly automating them in order to remove the economic necessity inadvertently forecloses the possibility of engaging with them in other ways. The above examples of automating boring office work and inadvertently foreclosing academic research holds here, but a more prudent example might also be that of gardening. The technologies that would wholly automate gardening and agriculture on a mass scale might be adopted for personal use, thereby maximising the efficiency and beauty of gardens across the world. However, in doing so, the meaningful activity of individual and communal gardening enjoyed by many might also be lost. Garden centres and the specialist skills that are developed through the activity would become more difficult to procure if there are machines that do the activity much better, quicker, and easier: and even if gardening remains a human-led activity in private, it is now being performed in opposition to the automated form that far outstrips the abilities of any given human gardener. This is particularly pertinent and conflicting when we consider housework. Housework can appear as an economic necessity for those employed to perform it, but often does not: as such, there might still be automated technologies intended towards this activity, despite it not necessarily occurring as an economic necessity. In doing so, some of the gendered imbalance of this activity might be redressed, but other engagements with this activity might become precarious, such as the associated monastic practices (Wijayaratna 1990) or communal, cultural, and personal practices (Borgmann 2010). This is an embedded risk with the development of wholly automated technologies, and the positive development in one direction will inevitably have implications in another.

Thirdly, as already mentioned, the inclusion of economic necessity in some activities does not wholly define their meaningfulness or desirability. That some artists, academics, lawyers, politicians, and athletes must perform these activities in order to economically survive does not wholly reduce the activity in question to meaninglessness, and yet by the principle that the technologies are being developed to remove their economic necessity, the professions themselves will risk being foreclosed in the post-employment world. The full automation of any meaningless component of these activities risks rendering the continued meaningful involvement of human actants wholly redundant.

In short, if activities are entirely done by autonomous and automated technologies, then any related or emergent activities might also be removed from the purview of human capacity. As a result, the post-work utopia may not be one of endless freedom and autonomy, but one of abject boredom, meaninglessness, and loss (Postman 1987; Carr 2016; Alter 2017), where human capacity is always inferior to the automated counterparts, and the very activities we hoped to fill our lives with become lost to us. By fully automating work, in a broad and impressive manner based on the assumption-laden notion of economic necessity, we risk untangling the web of activities that arise from, alongside, and in relation to work as it is commonly understood.

What I offer in response is therefore a more nuanced and dynamic understanding of work that does not rest on either economic necessity or a distinction between two distinct sets of activities.

Towards Modes of Activities

Thus far in the paper I have argued that understandings of work primarily relating to economic necessity are insufficient for flexible and open considerations of the future of work, particularly in light of automation and AI. I have argued that conceptions of work as a form of economic necessity that can be removed by autonomous and automated technologies fundamentally rest on the assumption that (1) work and non-work are categorically different, and that (2) the technologies to automate work will successfully remove economic necessity without significantly altering or affecting non-work. I have argued that these assumptions are fallacious, and that (1) work and non-work are not so easily separated in practice, and that (2) the technologies intended to remove work will necessarily have ramifications for the activities that remain after the fact.

Immediately, then, it is clear that this issue cannot be avoided by simply recategorising the activities of ‘work’ and ‘non-work’ into something else. The issue lies with drawing hard categorical distinctions between activities that are intrinsically tied to specific conditions. Certain activities, jobs, and industries might well be detrimental, meaningless, and menial as they appear under capitalist economic and social conditions: but to wholly automate these activities as they might appear under any and all possible conditions both now and in the future is a misstep. While this might well remove economic necessity, so too does it risk foreclosing engagement in these activities in a more meaningful manner. The complete automation of the office makes academic research and teaching precarious; in the same way that wholly automating the factory might render personal artistic or artisanal creation obsolete. One solution here might be to argue against full automation, and instead pursue automated technologies that augment and assist human workers, rather than replace them (Daugherty and Wilson 2018; Topol 2019). This is argued to allow workers to still exercise their autonomy and creativity, and to self-actualise through their work, while utilising technological artefacts (Roessler 2012; Breen 2019; Yeoman 2021). While this is certainly one possible solution, I argue that this will only address one of the two assumptions raised, and avoids the issue rather than solving it. If the kind of technological interventions developed still ultimately remove human involvement in a process, and fundamentally change the shape of an activity in order to eliminate economic necessity, but allow the human actants to do parts of the activity rather than fully eliminating them, we still risk radically changing the shape of the activities themselves and making human engagement redundant. If the principles guiding the development of the technologies are to remove economic necessity in some activities and enhance creative or meaningful engagement in others, the same resultant future will be reached over a longer period, because the foundational principles employed still rest on erroneous categorical distinctions between activities that can’t be borne out in practice.

I therefore argue that full automation is problematic, and is not the clear path to a post-work utopia as many scholars, including Aaron Bastani and John Danaher, might predict. I argue that examining the problem in this way reveals that there is a need to develop an understanding of work that is not categorically tied to activities in their own right. What is required, I argue, is an understanding of work not in terms of activities, but in terms of the modes in which a given activity takes place.

Under different conditions the same act can appear as meaningless, menial, repetitive, and detrimental paid employment; and meaningful, engaging, life-affirming, and self-actualising non-workFootnote 1. Cooking, for example, can be the an incredibly mundane act when the actant garners no meaning or actualisation from it, and is simply trying to fill their calorific needs at the end of a long day in preparation for the next long day; but equally, time in the kitchen can be incredibly meaningful if the conditions are correct, and the actant in question enters into the act with a desire for meaningful articulation and engagement. As such, it is not the act of cooking itself that is inherently meaningful or meaningless, but rather the conditions in which it appears and the mode in which it is undertaken. Even the most seemingly inherently meaningless and menial activity, such as cleaning toilets or working on an assembly line, can be undertaken in a meaningful manner should the conditions be right. This is not to say that such acts will always be meaningful, but that the right conditions can make them so. If the activity is part of an art installation, religious festival, political organisation, social interaction or cultural practice, then meaning, non-economic social good, and self-actualisation can be drawn from it. This is particularly pertinent when considering the difficult labour of the home, which have historically been combined with leisure and social activities for women (Langhamer 2000) and are instrumental as sites of political and social organisation. Again, this is not to say that these activities should not be affected by technological developments in order to protect the other ways in which they can be undertaken, but precisely to make the case that an activity in isolation cannot give a full indication of the ways in which it can be undertaken. As such, wholly automating any one activity, regardless of how clear-cut it might appear, can have unintended consequences.

What I propose, then, is two distinct modes of activity in which any given activity can be undertaken. The same activity can be undertaken in both modes, and any resultant automated and autonomous technology must be considered in relation to both modes. The two modes of activity offered here are by no means comprehensive, and I certainly leave additional space for other modes, including leisure, toil, duty, and so on: but for the purposes of considering the automation of work, I believe two modes are sufficient: labour and workFootnote 2. Let’s consider each in turn.

Labour, and the Enduring Relevance of Necessity

The first mode of activity I offer is labour. This is the mode of activity in which the human actant engages with their biological necessity, and engages in an act in order to secure their means of survival, wellbeing, and continued existence.

But this ‘biological’ necessity is very broadly intended, and captures all aspects of wellbeing and survival as they appear under different conditions and contexts. As such, in addition to any physiological wellbeing, labour also accounts for emotional and mental wellbeing, social necessities, and economic endeavours under capitalism. Labour is the mode of activity that relates to the fulfilling of such necessities, and is chiefly characterised by the intention to survive, maintain, or enhance an actant’s welfare: when an actant engages in act because not doing so would actively harm their wellbeing, then I argue that this is occurring in the mode of labour. The necessities (for sustenance, companionship, shelter, warmth etc.) are not strictly acts of labour in their own right, but rather necessitate engagement with the relevant tasks (cooking, building, gathering etc.), and when a human actant engages in an activity directly and primarily for one of these motivations, they are doing so in the mode of labour. This is not to make a deterministic claim: the necessities of labour can be ignored, through activities such as a hunger strike (Scanlan et al. 2008), but doing so might be motivated by the mode of work, which will be discussed next. Importantly, because human beings are not inherently solitary creatures, and these necessities are universal to all humans, labour does not only relate to individual actions of the actant in question, but also to those surrounding the actant, so acting in the mode of labour will also extend to caring for, and acting in relation to, the biological metabolism of those around us.

This is particularly pertinent because it allows us to understand both paid employment and unpaid reproductive labour in the same category. Under contemporary economic conditions, stable employment provides economically necessary wages that are used to secure other physical and mental necessities. When a person engages in their paid employment primarily to extract a wage to fulfil this necessity, then they are engaging in the mode of labour. But so too is the person that spends their time performing the necessary activities of the home or family engaging in the mode of labour, if their involvement is primarily motivated by the relevant necessities and motivations. Economic and non-economic necessities therefore occur in the same mode, I argue, although under different conditions. This is important to note because removing economic necessity will not remove the mode of labour. If capitalism is replaced with a new socio-political system, through technology or political revolution, then there will still be a need to engage in acts of labour, albeit of different types. We might not have to go out to work 9 − 5 Monday-Friday any longer, but we’ll still have to cook and clean, communally engage with one another and with our families, and undertake the necessary tasks to maintain our wellbeing in the post-work world. The drive to remove necessity beyond only economic necessity through the automation of work, I argue, is a reification of necessity itself with economic necessity under capitalism as it appears today.

It is clear from this understanding of labour that this is a cyclical process, one that does not end with the completion of any singular activity: each biologically necessary task only fulfils the temporary and ephemeral need at hand, and never brings about an end to biological necessity as a whole. The eating of a single meal does not end hunger indefinitely, nor hunger for all; the earning of a single wage does not cure all economic needs, nor eliminate economic need in general; and the birth of one child does not negate the need for its further care, or the general need to have more children. The entire process of labour cannot therefore be encapsulated, explained, or comprehensively defined by one single task, but only by the necessities and metabolisms driving the process as a whole. Even when a single actant dies and can no longer maintain their biological metabolism, the life process of the human race continues. The cycles of necessity and fulfilment, of survival and sustenance, will continue for as long as there are humans alive who remain biologically conditioned, and not, for example, transhuman or posthuman (Bostrom 2005; More 2013; Fuller and Lipinska 2014).

AI systems developed in relation to labour will also be subject to this cyclical temporality: there will never be a final AI system that can fulfil all needs, because these needs are evolving and changing over time and as culture and society shifts. Moreover, with machine-learning and the advancement of hardware specifications, the capability of the systems themselves will improve, and each system will be updated and replaced over time. Such tasks might even become acts of labour in their own right: the developers altering and improving the systems that support life in a fully automated and AI-based society might take on the role of powerplant engineers, transportation workers, and internet technicians today, all of whom keep the necessary facets of today’s technologically enhanced life running. When a system becomes integral to the continued survival of the culture or society that it appears in then its maintenance can be understood as an act of labour. An interesting further implication here is that these AI systems that replace and remove human necessity will still require maintenance and improvement (Russell 2019). As such, while, economic necessity can be removed through these automated and autonomous technologies, the technologies themselves then become a source of technological necessity in the post-work world. We would rely on them to fulfil our necessity, and their maintenance and upkeep would become our primary labour, as paid employment is today. We are not therefore removing necessity itself from the human condition in the post-work world, only economic necessity as it appears today.

This is particularly important to highlight, because it clarifies the kind of future that will be realised through a wholly automated post-work project: we remain creatures of necessity, but become bound to that necessity being filled by the technologies that replaced us in work. Our necessities, which will remain in every future short of a complete transhuman or posthuman vision (de Grey and Rae 2007), then become wholly focused on the technologies that are supposed to remove our concern for necessity.

As such, approaching the issue in this way demonstrates a clear path for the development of automated tools that operate in the mode of labour. Those activities that clearly relate to the biological necessities of an actant can be (and already are) mediated by technologies. In developing increasingly autonomous and automated technologies to intercede in these activities, the mode of labour can be a guiding principle: identifying the biological (broadly construed) necessities that are the focus of a given task can allow us to better target technological interventions. Increasing the production and accessibility of foodstuffs is a clear tenet of labour, but better appreciating how the necessity of hunger and sustenance fits into a wider condition of necessities might also prompt the larger-scale and more accessible production of healthy food, to better respect the biological condition of all human beings over time. While we can use this framework to interject on wholly economic examples (by identifying the ways that they negatively impact social and mental wellbeing, even if they provide economic wellbeing), so too can we easily include non-economic examples in the same discussion. If a person working externally from the home in order to earn a wage is intending towards the same type of familial and biological necessity as the person that stays (or has to stay) at home in order to fulfil the same necessities directly, without an economic mediator, then we can readily consider technologies that better fulfil both needs.

However, considering labour in this way also highlights some of the unintended consequences brought about by full automation. Because necessity will remain a facet of human life after automation, and because the same activities can be undertaken in different modes, fully removing the human actant from activities of labour might well result in a future society in which there is simply nothing of meaning left for us to do. When automated machines outperform humans in the activities previously performed under economic necessity, but our biological necessities endure, we might become reliant on these automated systems to provide for us, but lose the ability or capacity to engage in activities beyond this dependence. The adoption of full automation might therefore have more significant transhumanist and posthumanist commitments than is initially allowed. Perhaps more pressingly, approaching the issue through the lens of labour further demonstrates the unequal future that will be achieved through the development of automation entirely relating to economic necessity. If we automate all economic instances of labour but ignore all non-economic instances, then the same inequalities that plague the world endure, and the free time achieved by emptying the workplace does not translate to freedom in the home, the family, or the community (Hofmeister 2019). Technologies developed in the mode of labour must respect the openness of activities to be undertaken in different modes, and not completely pursue one avenue at the cost of all others: while also acknowledging the role of labour in human existence both now and in the future, so as to not inadvertently create the unequal and unjust society that is trying to be avoided or overcome.

Importantly, I argue that the need or desire to engage in meaningful work (Yeoman 2014) is of a different type to that of labour, and occurs in the second mode, work. Engaging in labour is primarily motivated by a concern for survival and wellbeing, and so we can understand an actant who avoids certain activities or leaves certain jobs because they are emotionally, socially, mentally, or economically demeaning: where engaging in them actively undermines the actant’s wellbeing, then not engaging in them is an act intended towards their wellbeing. A basic need for all humans, particularly under capitalism, is to engage in employment or activities that do not demean and harm the actant: but leaving a job because it is doing you harm and actively seeking one that is meaningful are distinct drives, I argue. A person might leave an office job because it is actively harming their mental and emotion wellbeing: this we can understand as labour. That person could just as easily find another mundane form of paid employment, in a factory or retail store, that does not cause the same harm but equally is not meaningful to them, just as they could pursue a lifelong dream of working in an industry that they draw meaning and self-actualisation from. Meaningful work is certainly a need, but one that does not necessarily appear in the mode of labour.

What follows from this, however, is a demonstration of the interrelated nature of labour and work. Because I am providing modes of activity, rather than sets of activities, the same activity that is undertaken without meaning or significance in the mode of labour can be undertaken meaningfully in the mode of work: but if that activity is wholly automated, such that it functions to achieve the goals of labour in an autonomous and automated fashion without any significant human involvement, then the ability to engage in the activity in the mode of work may dissipate. Automating the office, the factory, the shop, and the restaurant might very well remove the problematic instances of labour that arise there (which are undoubtedly in need of redress and attention), but in so doing might also inadvertently eliminate the very activities we may wish to spend our free time participating in.

Having established an understanding of labour, lets now consider the mode of work.

Work, and the Emergence of Meaning from Necessity

Embedded in the binary definition of labour offered thus far is the correlated mode of activity: work. I define work as the mode of activity primarily concerned with meaningful engagement and self-actualisation, one in which the actant can meaningfully articulate their selfhood, beyond their biological necessity. Work is therefore not concerned with the cyclical biological necessity of labour, instead taking place within definitive acts that stand independently of one another.

The desire to engage in work for reasons beyond biological necessity is a common facet of life across world (Bowie 2019; Mei 2019; Steger 2019). In light of Covid-19 and its resultant lockdowns, the question of meaningful work is more present in the public imaginary than perhaps ever before in history (Savage 2020; Soojung-Kim Pang 2020; Lipman 2021). However, under today’s economic and social conditions, many people must settle for paid employment that only satisfies their biological necessities, and pursue meaningful activities outside their jobs (Keohane 2015). This desired meaningfulness can take multiple forms: it can be the expression of a political or moral belief; the demonstration of a virtuous quality of the actant themselves; an inclination or value that the actant holds towards the world; a fulfilling pursuit beyond their own survival, and so on. In work, the actant is engaging in the act to achieve this meaningfulness, because they hold the value, idea, or inclination to be meaningful, worthy, and desirable. Through such an activity, the actant engages in self-articulation, and identifies themselves as the one who is acting. As already noted in the previous section, the desire to obtain employment or engaging activity that does not detrimentally affect a person’s wellbeing is an act of labour, because the actant is operating in line with their biological metabolism: meaningful work, on the other hand, occurs without primary regard for the actant’s wellbeing. This can therefore account for situations wherein an actant continues to perform a job role or activity that actively harms them, because doing so provides them with meaning beyond their biological necessity.

Work can occur under the conditions of paid employment, but does not always: economic remuneration is not mutually exclusive of meaningful work, but where the economic incentive of a given activity overrides the meaningful engagement, and the economic necessity becomes the primary focus of the activity, then we can say that the activity is now being undertaken more clearly in the mode of labour. Indeed, philosophical and sociological notions of ‘gift’ are relevant here, when an exchange of goods is undertaken through respect and reciprocity, arguably more in line with work than with labour (Mauss 2001). This is one such avenue for radically changing social and economic systems, to better reflect the dignity of the human beings involved in the exchange; and, in the language of this paper, reflect and protect the mode of work (Hénaff 2010). However, it is important to also note that contemporary economics does not function as a gift economy, and so in the majority of instances today where work is undertaken for economic remuneration, it will not be in the form of a gift exchange, and will instead be undertaken as a capitalist economic exchange.

However, regardless of whether work appears under economic conditions or not, it is distinct from labour both for its intention towards meaningful engagement and a lack of cyclicality. Where acts of labour always lead into further acts, bound by necessity, work takes place in definitive acts that stand independently. Further action might be required once the meaningfulness of a task has been achieved, but this does not detract from the meaningfulness of that given moment. A poignant political speech, a piece of art, or an important moment of personal identification doesn’t become subsumed in an overarching process in the same way as, for example, a meal eaten out of sheer hunger. Even where such instances belong to longer processes, such as a political speech or an art exhibition, the individual act can be taken as meaningful in its own right. Such acts might require repetitive, menial, and strenuous preparation, which might suggest that the act belongs to the mode of labour, but preparatory sub-tasks which are intended towards the meaningful self-articulation of the actant occur in a different mode to those of labour. If the repetitive and menial sub-tasks are pursued with the intention of achieving something meaningful and self-articulating, they are occurring in a very different manner to those pursued for the continued survival of the actant. Learning an instrument or skill, working on a piece of writing or art, or building an object can all occur in the mode of work, precisely because standout and definitive moments of meaningfulness will emerge out of the repetition that are not re-subsumed into the process.

Automating work is therefore more difficult than labour, because work gains its significance from the human actant engaging in it: removing the human actant also removes the work, in some sense. This is further compounded by labour and work being interlinked, and appearing as possible modes in the same activity: automating wholly in the direction of one mode may foreclose and jeopardising engaging with it in another. It is perhaps in this mode that ‘full’ automation is clearly the most problematic. It is the activities of work that scholars laud as populating the post-work future, but engagement in this mode might be muted if the technologies bringing about the post-work future negate the possibility of engaging in the mode itself.

Approaching the automation of work through AI and automated technologies benefits from this understanding of work as a mode of activity, because it shows the clear limits of technological involvement in activities if we are to respect and acknowledge the multiplicity of modes in which they can be undertaken. Rather than totalising a given activity in a specific mode, and foreclosing the possibility of undertaking it in any alternate mode, the technologies in question must be much less complete in their enclosure. The same goals, of ending economic necessity and making work more fairly distributed for example, can be achieved through social and political ends that do not nullify human engagement in these tasks, in either mode. The wrongs of the world today can be resolved without committing to a world of tomorrow that is inevitably less human, more sterile, and potentially more unequal than is currently predicted.

Conclusion

Work currently plays a pivotal role in contemporary economic necessity, and developments in automated and autonomous technologies are seeking to resolve this. However, I have argued that the total automation of work rests on two fallacious assumptions: firstly, that work and non-work are categorically distinct activities, and that work can be removed from human involvement without any detrimental effects on non-work; and secondly, that the technologies that remove work will leave non-work possible and accessible. I have argued that these assumptions are incorrect: firstly, the distinctions between work and non-work are much less clear than is often assumed; and secondly, the technologies proposed for removing work are not attuned to the subtle distinctions between work and non-work. As a result, the proposed futures of endless free time and creativity are misplaced, and there is a high likelihood that wholly automating work will perpetuate current forms of inequality, or will result in a post-work future in which no meaningful activities are left for human engagement.

In response to this issue, I have proposed two alternate understandings of ‘work’: the modes of activity that I have described as labour and work. Labour accounts for the continued necessity that occurs both now and in post-work futures, and work accounts for the meaningful activities that we hope will fill all futures, while acknowledging how these two modes of activity are interlinked. As such, these two modes of activity can account for the current issues endemic in economically necessary paid employment while also offering useful and practical tools for designing the future of post-employment.

By developing these two modes, I have argued that the desire to remove economic necessity can be achieved through social and political means, rather than technological. The full automation of work is imprecise, and fundamentally misses the nature of necessity in the human condition. The future of work will certainly be an automated one, but it must remain a human one. There are a number of prevalent approaches advocating for a more human future, including Humanistic Management (Pirson 2017), the politics of care (The Care Collective 2020), and the reclamation of Workplace Democracy (Ferreras et al. 2022). For both researchers and managers, I argue that in the first instance we must understand the distinctions between labour and work, and the ways in which autonomous and automated technologies ignore, and indeed negate, these distinctions. From this foundation, a future of work that can maximise the utility of autonomous technologies while continuing to nurture and enshrine human capabilities is possible.