To answer the research questions presented in the Introduction, we divided the presentation of the literature review into the following parts:
As regards consequences of technological unemployment, we can highlight the following:
Before discussing the causes of technological unemployment, a first division of the literature should be noted. The reviewed articles about technological unemployment can be divided into two broad categories. The first category is composed of those articles that take as a starting point the idea that technological unemployment is a possible negative consequence of automation that should be discussed and mitigated [1
]. Most of the reviewed literature falls into this first category. The second category includes the articles that somehow dispute the fact that automation will cause long-term technological unemployment [29
]. It is interesting to present these articles because they bring new perspectives to the discussion about automation and its possible consequences.
As shown in the Introduction, the concept of technological unemployment, as defined by Keynes, already gives an idea of the causes of this phenomenon: we, as a society, adopt technologies to replace human workers at a faster rate than our capacity of creating new work.
In previous industrial revolutions, automation already substituted human work for machine work. At first, machines would substitute for repetitive and manual tasks. Then, repetitive cognitive and manual tasks were within the reach of automation. Now, as AI gets ever smarter and robots more skilled, nonrepetitive cognitive and manual tasks seem to be increasingly automatable [2
]. As an example of the increasing capacity of technologies, we can look at the automation of case research performed by paralegals, the use of facial recognition coupled with cameras that automate part of security work [2
], or the development of self-driving cars, which may lead to the unemployment of truck drivers [4
]. The advance of the current wave of technology over skills that were previously exclusive to humans puts in check the Luddite Fallacy that held true in earlier industrial revolutions and implies that the jobs lost in a given economic area will be created in another one [1
]. For instance, 90% of the jobs in the USA economy are in the service sector where smart information and communication technologies can create huge waves of unemployment [38
It is also interesting to track the quality of the current job change by verifying if the jobs or activities being automated are more or less dangerous, boring, and exhausting than the ones being created in other areas [32
]. If the current wave of automation is really set out to cause a hollowing out of the jobs at the middle of the skill spectrum, then we might see an increase in the demand for both low-skilled and high-skilled jobs in the next decades [1
Something else that is particularly concerning about the current industrial revolution is the accelerated pace at which the advances in technology are taking, which could speed up the job market remodeling [2
]. Currently, many automation technologies rely heavily on software, which is something easily distributed across the globe, when compared with the main technologies of previous industrial revolutions such as steam machines [51
Moving forward from the causes that are part of the definition of technological unemployment—rapid pace of technological change and skills mismatch—there are other, less straightforward factors that can also be considered accelerators of automation.
National and international tax systems are two of these factors. At the national level, tax systems that are currently formulated to charge more labor than capital can act as a stimulus to the automation of work by helping to tip the cost balance in favor of machines instead of humans [4
]. At the international level, critics point at the fact that the international tax system is incapable of quickly adapting itself to solve issues of distribution of the tax base, particularly between developed and developing countries [28
]. Another issue with the international tax system would be its current basis on the distribution of rights for the taxation of income due to the cross-border transfer of capital and technology. That transfer happens between a “residence” country—usually, more developed countries with advanced technologies, and primary and often exclusive rights to tax such profits—and a “source” country whose rights to tax profits tend to be either not fully realized due to a lack of necessary technical competences among their local tax authorities or limited by the international agreements [28
One last important factor that can act as an accelerator of technological unemployment is the control of the technology development agenda by very few companies located in a small number of countries [3
]. The current development of digital solutions is dominated by companies from the world’s two largest economies: China and the USA [28
]. Digitalized goods and services allow replication at near-zero marginal cost; nondigital goods have near-global distribution networks, leading to the “winner-takes-all” problem: the income tends to flow to one dominant participant [3
]. For example, in the field of AI, a fundamental technology for the current automation wave, the main players are nine big tech companies: Baidu, Alibaba, and Tencent from China; and Google, Amazon, Microsoft, Apple, IBM, and Facebook from the USA [29
]. These companies have to face the challenge of technological unemployment in their own countries, and it is hard to believe that they will care about the impact of their technologies on jobs in less economically developed countries unless it brings some benefit to themselves [28
To some, the very fact that the growth of human knowledge, here in the form of technological change, is left to be determined by profit-maximizing companies is a problem in itself [4
]. From this perspective, technological unemployment is a consequence of the use of capital that is freed up by automation to investments that seek to create “superabundant capital” or “cash pools” that are separated from the real economy [31
]. Thus, changing the way society understands the very function of capital would avoid technological unemployment if it was used to fund human improvement as a whole instead of increasing the richness of very few people [31
]. Therefore, technological unemployment is a challenge to the field of business ethics [4
Working provides a sense of aspiration, meaning, and enjoyment and improves financial and social self-governance. Thus, unpredictability at work generates a significant sense of insecurity and discomfort [26
History shows that the technological revolutions greatly affected the environment, as previous economic growth implied more energy consumption—usually covered by oil and coal. Investments to make the transition to a green economy may improve employment, compensating partially the jobs lost due to technological unemployment [42
This section of our literature review is dedicated to presenting the different consequences of technology unemployment. As it can be expected, most of these consequences can be considered negative. Still, two of them—more free time and less consumption—have both positive and negatives sides, which will be explored.
Starting with the negative consequences of technological unemployment, two of the direst ones presented by the reviewed literature are increased economic inequality and lack of minimum living standards for a share of the population.
Economic history demonstrates that previous technological revolutions caused a disruption in the labor market in the short run, but in the long run, the situation has stabilized [24
]. Nevertheless, Keynes reminds us that “in the long-run, we are all dead” [52
]. That perspective is important since we do not know how long society will take to adapt to the current industrial revolution, and how badly can inequality grow during this period of adjustment [28
A couple of factors can make the current adaptation more challenging [28
]. First, disruptive technologies tend to demand increasing skills if the individual is to make a positive transition in the labor market. Second, in some regions of the planet, such as Latin America and Africa, there is a significant digital divide that could leave whole regions out of the industrial revolution. Third, workers’ rights are experiencing growing general insecurity, which can make automation adoption more abrupt and leave workers in hardship. Fourth, a general slowdown in the global economy has been furthered by the COVID-19 pandemic, which can make it harder for workers to find new jobs.
In this scenario, workers that are already part of the high-qualified job market can experience a surge in their demand and salaries, while the wages of untrained workers have been declining, as has been happening lately and could worsen in this adaptation period [32
]. Human enhancement using new technologies such as gene editing or chip implants could become yet another catalyst of this process of increasing inequality. This can happen since these technologies will have a high starting price, being accessible to only a small part of the workforce who might benefit from their enhancement while the majority of the population is relatively “disenhanced” [32
Thus far, we have discussed elements of the growing economic inequality caused by technological unemployment, but they could also result and be influenced by a lack of minimum living standards of part of the population. A possible consequence of technological unemployment is the creation of what some might call the “useless masses” that are left out of economic production [37
]. That can be a possibility if our society keeps on being organized around the idea of profit maximization and the use of capital [38
]. Figuring out the meaning of life can become the new “first-world problem”, which could be a concern of only a small and rather sadistic elite who will free the masses from this preoccupation by leaving them out to starve [38
Having more free time is a consequence of technological unemployment that has a positive and a negative side. Keynes long recognized that the technological advances would eventually solve the problem of meeting societies’ material needs and would leave humans with the challenge of finding purpose for their free time—what he called humanity’s real, permanent problem [23
]. Therefore, long-term unemployment could make workers’ lives feel meaningless [3
] and without financial and social self-governance [26
]. According to Danaher [3
], work is viewed as virtuous; thus, the lacking of paid employment may lead to idleness, boredom, and depression. The void left by work in our life could be filled by education, self-care, and care for others, improving our society by improving our institutions and dedicating our life to research, philosophy, and arts [31
]. To move to a leisure society, we must decouple working from income while changing the social role of the paid employment in personal dignity [38
]. This transition from economic-focused to leisure-occupied individuals requires a preparation that is entirely different from the one we have that reflects on how awful investors of free time our society can currently be [38
Even if mass and long-term unemployment is not reached, jobs in the future are likely to demand more education including higher technical and emotional skills, which means that part of the time freed by automation could be invested in better educating ourselves [40
As previously discussed, the fast pace of technological change is one of the causes of technological unemployment because it causes a skills mismatch in the labor market. Additionally, as a consequence of technological unemployment, societies might perceive that their workforces have outdated skills, when compared with the ones demanded by the new technologies, thus leaving part of the population behind in terms of economic possibilities [39
]. Therefore, there is an outpacing problem with retraining/reskilling for the new advanced technologies: the needed skills are more complex and harder to be learned; meanwhile, the newer skills are becoming obsolete increasingly faster [3
]. The acceleration in technology development is a trend that could lead to newer advanced technologies automating the new jobs created by the previous generation before the displaced workers are even ready to perform them [4
], making reskilling useless.
Many believe that higher education is the key to acquiring and keeping quality jobs; however, a university degree no longer ensures reliable and secure work [27
]. A large number of graduates from outmoded educational systems have mismatched or outdated skills for the labor market [1
]. Moreover, the costs of higher education are increasing rapidly, while wages are stagnated [49
As technological unemployment advances, governments will be increasingly put under pressure to expand their social welfare to include programs such as UBI [3
]. These programs are extremely expensive and may be seen as an economic burden, mostly due to the need to increase taxation to cover their costs since the traditional welfare state is unable to face these new costs without more funding [48
Furthering this challenge, technological unemployment brings a fiscal risk that could destabilize the existing social safety net, which is the decrease in tax revenues coming from labor if labor’s contribution to the economic production is reduced [28
One last consequence of technological unemployment that could be highlighted from the reviewed literature is a possible reduction in demand. This is another consequence that has positive and negative perspectives. The advance of automation means that less labor is required to produce the goods needed by people if the demand stays stable. Historically, automation increased the efficiency in the use of labor [42
], increasing production and lowering prices [47
]. Thus, until now, automation led to more new jobs than those that were lost [4
] and increased wages [4
], increasing the demand for products [4
]. Still, there is no guarantee that this process will repeat this time. Without redistributive policies, as the one previously discussed, the unemployed will lose purchasing power, and demand will shrink [31
]. Of course, this consequence has a limit because if no one can buy anything then nothing needs to be produced and the economy will collapse, but up to this limit, the disbalance between the demand and the production of goods can happen because of technological unemployment.
Having presented the causes that can bring about or accelerate technological unemployment, and the consequences of this phenomenon, we now turn our attention to the solutions proposed by the literature to this challenge. While reviewing the literature, we came across a considerable number of proposals from researchers. Here, we decided to categorize these solutions into two groups. The first group of solutions dealt with mitigating the causes of technological unemployment, thus minimizing or avoiding it. The second group of solutions aimed at helping society to deal with the consequences of technological unemployment.
3.3.1. Solutions to Causes
According to the reviewed literature, avoiding or reducing technological unemployment can involve several measures. One of these measures is augmenting workers instead of replacing them [30
]. This is not possible or desirable in every single role [39
]. In the cases where it can be applied, augmentation might bring the (sometimes temporary) benefit of allowing for a longer period of adjustment for the workers serving as an intermediary step from total automation [30
To be augmented, workers will need to have higher skills, mainly digital skills, which could result in a rise in the general level skill of the workforce increasing the offer of high-skill work and reducing the wages thus renewing the middle class [32
Another measure that could be adopted is sharing work. Instead of laying off the workers displaced by automation, companies could reduce the number of hours in the working week [35
]. A reduced workweek was one of the economic possibilities put forward by Keynes back in the 1930s when he believed that their grandchildren would have a 15 h workweek when they came of age [23
]. The shared work policy can also bring some relief to the social safety net, particularly to unemployment systems, since workers would not depend on it if allowed to remain working less time [35
]. Sedai suggests that this strategy should be limited to a reduction of 20–40% and that companies should provide the same employee benefits when reducing the working hours [35
Nostalgy or technology aversion might motivate the revival of certain occupations that could help to balance the unemployment caused by the increased use of technology. In the future, as automation replaces more and more humans, there could be a rejection of technology leading to an increase of jobs in traditional fields such as handcrafted products [2
]. Chomanski shows that this nostalgic work already exists, as in the example of the hiring of horse-drawn carriages and, as technology advances, humans could be hired to be nannies, painters, or chefs [30
]. Kim and Scheller-Wolf [4
] states that we must search for new business ideas to create a market for human labor. One idea is the creation of a “made-by-humans” campaign to create a market demand for human labor.
As it happened before, it might be the case that new economic sectors, some still unknown, might be created to replace the ones in which little or no labor is needed. Walden suggests some possibilities: an increase in the services provided for time-constrained households; development and implementation of new technology; data production, management, and analysis; a possible surge in world tourism; and expansion of health care workers as the world population continues to increase its life expectancy [2
]. Technological change sometimes renovates jobs instead of eliminating them. In this case, workers need to update their knowledge and skills but can remain in their jobs [48
Finally, the inequality between countries could be mitigated by international tax cooperation to address technological unemployment [28
]. In Berberov and Migolov’s view, developed economies could collect resources through an ephemeral global tax tool (e.g., a robot tax) to pay a universal basic income to the citizens of developing countries [28
]. Although, as discussed previously in the causes of technological unemployment, the international cooperative mechanisms might be limited to implement this, and the current limit of such international actions might be the incentive of research and knowledge sharing about strategies to combat technological unemployment, an example of which could be close to the United Nations recommendations [28
3.3.2. Solutions to Consequences
If technological unemployment happens, there are some measures indicated by the literature to mitigate or eliminate some of its consequences. Here, we made an effort to indicate the relationship between consequences and the proposed solutions, but it should be noted that some solutions might affect more than the one or two consequences of technological unemployment that are considered.
Related to the possibility of increased inequality and the lack of minimum living standards, the literature proposes basic income guarantees, charitable donations, the production of ones’ own goods, and a change in the social safety systems.
There are different versions of basic income guarantees, such as the UBI, in which every citizen receives a certain amount of cash regardless of their economic situation, mandatory profit-sharing by firms, and minimum employment schemes [35
]. Sedai defends that regardless of the chosen mechanism, providing a minimum living standard for its citizens is “exactly the kind of public good that government should create” [35
]. Such redistributive systems reduce the financial stress from most vulnerable citizens with minor impacts on the rich [44
If automation reaches a high level, and wealth is still or further concentrated in the hands of a few, future elites could become rich enough to help mitigate the technological unemployment with their charitable donations [30
Another possible solution to the lack of minimum living standards that could be caused by technological unemployment is the production of goods by the people who need them. As technology cost lowers and people have more free time, they could invest in learning how to code, for instance, to create their own machines to perform at least part of the work needed for subsistence [30
Changing the social safety net is also indicated as a possible course of action that could mitigate the consequences of technological unemployment [2
]. Particularly, the unemployment compensation system was created to support workers during temporary unemployment due to economic downturns while they waited to be once again employed in the same function [2
]. Technological unemployment is not a case of a temporary economic downturn but a more permanent transformation of the economy that requires workers to learn new skills to find new jobs [2
]. To counter this problem, Walden proposes that the current unemployment system could be changed to include an upfront aid to be used for education costs [2
]. The government-provided safety net could also be complemented by “unemployment insurance” sold by companies, and by workers’ personal savings that could be increased in a “post-automation” society that allows for a reduction of goods prices [30
These changes to the social safety net could also help with the update of the workforce’s outdated skills. Here, three other solutions might help: tracking occupational change, changing higher education, and increasing on-the-job corporate retraining.
Besides being an important measure by itself, tracking occupational change is an enabler of the other solutions to the workforce’s outdated skills and even to solutions to other consequences of technological unemployment [2
]. For the case of the USA, Walden proposes the creation of an early warning system of occupational change that would examine annual Bureau of Labor Statistics to detect how occupations are shifting, combined with the tracking of market trends in hiring with data from job postings and social media, and regular direct surveys of companies to access their hiring necessities in terms of skills and tasks [2
One of the uses of the tracking of occupational change would be to help higher education institutions in the analysis of their courses portfolio and the following (rapid) reallocation of resources to meet demand [2
]. Coupled with this rapid response to the market regarding their traditional undergraduate courses, institutions may also need to provide shorter courses focused on retraining workers [2
]. The standard curriculum should heavily focus on science, technology, engineering, and mathematics (STEM) to ensure that graduates can work with the machines at their workplaces [1
]. Another required change to the educational system would be the decoupling of the training in core competencies needed for any occupation in the future (e.g., computational competencies, and complex communication) and the specific skills required by a certain occupation [2
]. These changes will require more investment in higher education, which could be supplied by government funds and by private companies’ resources as the private sector gains a certain level of control over the courses curricula [2
]. Workers should be prepared to learn the advanced skill sets [49
]. Massive open online courses (MOOCs) promise to change higher education, providing flexible and affordable courses [45
Companies could help with this challenge by complementing higher-education institutions’ financing and by providing on-the-job retraining. As companies are the locus where automation takes place—they have a privileged position to analyze which new skills the workforce should have. This process could be incentivized by limited labor compensation deductions or noncompete contract clauses for a certain period to defend themselves in the case of the trained employee decides to leave the company [2
Finally, the negative consequence of fiscal risk could be mitigated by a fiscal reform. One of the proposals for fiscal reform is the robot tax, which requires that companies that replace humans with robots pay a tax that could then be used to provide a safety net for displaced workers; an idea that is already introduced in South Korea and is being considered in Canada, India, and China [28
]. Other proposals involve a tax rate on profit that is larger than the tax rate on wages [41
], progressive consumption taxes coupled with income taxation [38
], and the joint consideration of corporate and personal taxation [28
]. Automation cannot be used to reduce tax revenue [42
For the case of Russia, Berberov and Milogolov suggest that the tax reform should go in the direction of supporting Russian IT companies, reviewing personal income tax to reduce the burden on the people with relatively small incomes, incentivizing staff education and retraining, and reforming tax residence criteria of Russian digital specialists [28
For the case of Latin America, Aguilera and Ramos Barrera [33
] highlight three previously mentioned strategies to face technological unemployment: the use of income guarantees such as the universal basic income, retraining workers and stimulating them to live lifelong learning, and reducing the weekly working hours by sharing work.
Another proposed solution that is worth mentioning was made by Danaher [3
]: an increased integration with the machines. Merging our bodies and minds to machines would bring the benefits from the new technologies without the associated problems, such as technological unemployment.
3.4. Research Agenda
This section of our literature review is dedicated to presenting the different research roads that will lead to more knowledge about technological unemployment. We reviewed several papers with different backgrounds and goals and, although they all have some relation to the technological unemployment subject, they present different research agendas that could be taken as a starting point for future research.
Aguilera and Ramos Barrera [33
] investigated the technology impact on labor markets in the Latin American context. They found the investments in SandT have no significant impact on the unemployment rate. They claim that further research is needed for the region since their results differ from the ones obtained in developed countries. In addition, in the field of employment, they indicate the need to formulate tentative scenarios to identify unintended consequences of machines being the dominant form of production and value creation.
] empirically analyzed the impact of technological change on unemployment. His results indicate that faster technological change is likely to increase unemployment substantially. However, more research is needed about the transmission channels from technological change to unemployment and in the time pattern of the effect on unemployment. Finally, policy implications must be discussed as well as the government’s role in helping workers to fully benefit from the technological progress.
Carvalho and Di Guilmi [41
] consider that their model to analyze the relationship between technological unemployment and income inequality could be extended in several ways such as using a more sophisticated treatment of fiscal and monetary policy, and the inclusion of the possibility for households to buy shares during a stock-market boom.
] asked experts and nonexperts to give their opinion about the automatability of the occupations used by Frey and Osborne [5
] to train their model. The author considers that future work could be dedicated to extrapolating from jobs at risk to a percentage of the workforce unemployed by automation.
Fernández-de-Córdoba and Moreno-García [43
] modeled involuntary unemployment to find which conditions in the labor market prevent wages from falling due to the supply of unemployed workers. They focused on noncooperative solutions, and they claim that further research considering cooperative solutions is needed.
Peters et al. [27
] discussed the higher education policies, and why the simple “more education” solution has been failing to solve the problem of technological unemployment. They claim that we should give more agency to researchers, teachers, and students. Moreover, they stated that we need new visions of different social orders in which education, technology, and employment have radically different meanings.
] called for more research from normative political and moral philosophy about technological unemployment. According to the author, there is a wealth of relevant data about the issue, attention from the media, and research by economists, while philosophy is giving surprising little attention to the subject.