Can we Have Pro-Worker AI? Choosing a path of machines in service of minds

Daron Acemoglu, David Autor, and Simon Johnson1 Massachusetts Institute of Technology and CEPR

Summary

• Over the past 40 years, the diffusion of digital technologies significantly increased income inequality.

• Generative Artificial Intelligence (AI) will surely impact inequality, but the nature of that effect depends on exactly how this technology is developed and applied. Nothing about the path of this (or any) technology is inevitable.

• The private sector is pursuing a path for generative AI that emphasises automation and the displacement of labour, along with intrusive workplace surveillance.

• Simply displacing workers is never good for the labour market, even when the displaced are highly paid. Displaced formerly high-paid workers are forced to compete for jobs with lower-wage workers, leading to a downward cascade in wage levels.

• A better path is available, along which generative AI would be complementary to most humans - augmenting their capabilities - including people without a four-year college degree.

• Choosing the human-complementary path is feasible but will require changes in the direction of technological innovation, as well as in corporate norms and behaviour.

• The goal should be to deploy generative AI to create and support new occupational tasks and new capabilities for workers. If AI tools can enable teachers, nurse practitioners, nurses, medical technicians, electricians, plumbers, and other modern craft workers to do more expert work, this can reduce inequality, raise productivity, and boost pay by levelling workers up.

• Public policy has a central role in encouraging this positive path of technology to complement all workers, elevating the achievable level of skill and expertise for everyone.

群内每日免费分享5份+最新资料

300T网盘资源+40万份行业报告

为您的创业、职场、商业、投资、亲子、网赚、艺术、健身、心理、个人成长 …… 全面赋能！

关注公众号获取更多资料

添加微信，备
注“入群”
立刻免费领取
200套知识地图+最新研报

致终身学习者社群

行业报告/思维导图/电子书/资讯情报

收钱文案、增长黑客、产品运营、品牌企划、营销战略、办公软件、会计财务、广告设计、摄影修图、视频剪辑、直播带货、电商运营、投资理财、汽车房产、餐饮烹饪、职场经验、演讲口才、风水命理、心理思维、恋爱情趣、美妆护肤、健身瘦身、格斗搏击、漫画手绘、声乐训练、自媒体打造、效率软件工具、游戏影音……

扫码先加好友，以备不时之需

• At this time, the five most important federal policies should be:

Equalise tax rates on employing workers and on owning equipment/ algorithms to level the playing field between people and machines.
Update Occupational Safety and Health Administration rules to create safeguards (i.e. limitations) on the surveillance of workers. Finding ways to elevate worker voice on the direction of development could also be helpful.
Increase funding for human-complementary technology research, recognising that this is not currently a private sector priority.
Create an AI centre of expertise within the government, to help share knowledge among regulators and other officials.
Use that federal expertise to advise on whether purported humancomplementary technology is appropriate to adopt in publicly provided education and healthcare programmes, including at the state and local level.

Introduction

The world is about to experience transformative and disruptive advances in generative Artificial Intelligence. A major set of concerns centres around the labour market and economic inequality implications of these advances. Will AI eliminate jobs in net? Will it further inflame the decades-long phenomenon of rising economic inequality? Will it boost labour earnings or instead make machines more valuable and workers more expendable?

The consensus in the economic literature is that previous waves of digital technologies – including personal computers, numerically controlled machinery, robotics, and office automation – have increased inequality. This is both because some of these technologies, such as personal computers, have been highly complementary to moreeducated workers (Autor et al. 1998, Autor et al. 2003, Goldin and Katz 2008), and also because many of these tools have been used for automating work, with an unequal impact on different types of workers (Autor et al. 2003, Acemoglu and Restrepo 2022a, 2022b). While digital technologies have undoubtedly created new goods/services and boosted productivity in some activities (e.g. Brynjolfsson and McAfee 2015), there is also evidence that productivity gains from these technologies have sometimes fallen well below expectations (e.g. Acemoglu et al. 2016).

Generative AI will have a substantial impact on the future of work and the trajectory of inequality. The nature of that impact is not an inevitable consequence of the technology itself but instead depends on how society develops and shapes AI.

• The currently predominant direction for AI emphasises automation, displacement of skilled labour, and diminished worker voice due to steppedup monitoring and surveillance.

• An alternative, “human-complementary” path could contribute more to productivity growth and could help reduce economic inequality.

In the next section, we outline what the automation path looks like and what its implications would be for work, inequality, and productivity. We then describe the alternative human-complementary path, drawing on both general principles and specific examples. We also explain why, despite its advantages, the humancomplementary approach is not likely to prevail based on current investments and corporate attitudes. We suggest policies that could help steer AI development and implementation in the more constructive direction.

THE AUTOMATION PATH

Automation – the substitution of machines and, more recently, algorithms for tasks previously performed by humans – has been a constant since at least the beginning of the Industrial Revolution. Machines were developed to perform tasks that have a high degree of predictability and are carried out in stable environments. Examples include spinning and weaving in the eighteenth-century textile industry, harvesting in nineteenth-century agriculture, and many office and clerical tasks in the twentieth century, such as telephone switchboard operation and routine bookkeeping. Mass production, which vastly reduced the cost of everyday products, depends fundamentally – though not exclusively – on assembly lines made possible by automation.

Not all automation is highly productive, however. When machines are deployed to perform tasks in which they are not particularly effective, lacklustre productivity gains follow. Most people are familiar with the frustration of seeking customer service from an airline, credit card provider, or computer manufacturer, only to be diverted through mazes of unhelpful computer menus. Firms may find such automation to be cost-effective, but it is not a meaningful productivity advance.

Whether it has large or small effects on productivity, automation tends to have major distributional consequences. The reason is that automation displaces workers who were specialised in the tasks that automation reallocates to machines and algorithms. The automation of blue-collar and office jobs using digital technologies has been an important driver of the rise in inequality since 1980 (Acemoglu and Restrepo 2022a).

It is inevitable that AI systems will be used for some automation, both for technical and business strategy reasons. On the technical front, a major barrier to automation of many service and production tasks has been that they require flexibility, judgment, and common sense – things that are notably absent from pre-AI forms of automation. Artificial intelligence, especially generative AI, can potentially master such tasks (Susskind 2021). A broad swath of computer security tasks that used to be performed by skilled human operators can now be performed by AI bots. Similarly, generative AI systems can write advertising copy, parse legal documents, transcribe physicians’ medical notes, and perform language translation. It is unclear how much this type of automation will contribute to aggregate productivity growth while these technologies are immature, but they could contribute to sizeable productivity gains as costs fall and reliability improves.

Businesses may choose machines over workers for reasons other than productivity. Automation appeals to managers who are seeking greater consistency and less opposition from organised or unorganised labour (Acemoglu and Johnson 2023).

All too often, businesses prefer to focus on automation rather than creating new job tasks and enabling workers to build new skills. Automation is always an easy case for a manager to make because it appears to save costs. Investing to make workers more productive or more useful may be a harder sell, since it’s seen as messy, uncertain, and expensive. Some managers simply prefer to “hire machines”, rather than hiring workers, because machines don’t complain about pay or working conditions, and they certainly don’t join unions. But a country is not a business. We have a shared interest in ensuring that adults are productively employed. This promotes economic resilience, social cohesion, and a strong tax base. Policymakers care more about the quality and quantity of jobs than do employers, and policy should support institutions, incentives, and investments with this in mind.

Beyond economic incentives, the dominant intellectual paradigm in today’s digital tech sector – among both business leaders and academic researchers – favours the automation path. A major focus of AI research is to attain human parity in a vast range of cognitive tasks and, more generally, to achieve “artificial general intelligence” that fully mimics and then surpasses capabilities of the human mind. This intellectual focus encourages automation rather than the development of human-complementary technologies (Acemoglu and Johnson 2023).

There is a widely shared optimism that human-replacing automation will produce so much productivity gain that all types of workers will benefit. It is true that if automation is sufficiently productive (and thus reduces costs by a significant amount), this can generate demand for other goods and services and, as a result, workers may benefit as well.

However, while productivity gains are obviously welcome, there are two problems with this line of thought. First, the benefits may be highly unequally distributed across different skill groups. For example, AI-based productivity gains might increase the demand for so-called “prompt engineers”, but this would not help workers displaced from accounting or financial services jobs – assuming they do not have a comparative advantage in prompt engineering (Acemoglu and Autor 2011, Acemoglu and Restrepo 2022a). Second, automation tends to reduce the labour share of national income, so even if workers benefit, most of the gains flow to entrepreneurs and capital owners (Acemoglu and Restrepo 2018). There are limits to how much shared prosperity can be generated exclusively by automation.

Another common prediction is that, because generative AI may automate managerial or knowledge tasks typically performed by professional workers, it could have equalising effects. For instance, if accountants and financial analysts lose their jobs, this might reduce inequality between retail workers and highly paid financial sector workers. This logic is faulty. Studies of previous waves of digital automation show that workers directly displaced by new technologies not only experience lower pay growth but also start competing with other groups of lower-paid workers, whose pay then declines (e.g. Acemoglu and Restrepo 2022a). Simply put, displacing workers will never be good for workers or for the labour market. Instead, AI can reduce inequality if it enables lower-ranked workers to perform more valuable work – but not if it merely knocks rungs out of the existing job ladder.

The Human Complementary Path

New technologies need not merely replace workers in existing tasks. They may also complement workers by enabling them to work more efficiently, perform higherquality work, or accomplish new tasks that were previously infeasible (Acemoglu and Restrepo 2018, Autor et al. 2022, Acemoglu and Johnson 2023). For example, even as mechanisation gradually pushed more than half of the US labour force out of agriculture, a range of new blue-collar and clerical tasks in factories and newly emerging service industries generated significant demand for skilled labour. The expansion of employment in industry and services between (roughly) the years 1870 and 1970 led to work that was not only better paid but also less dangerous and less physically exhausting, and increasingly rewarded the formal literacy and numeracy skills created by the expansion of universal public high school education.

This virtuous combination - automation of traditional work alongside creation of new tasks – proceeded in relative balance for much of the twentieth century. But sometime after approximately 1970, this balance was lost. While automation has maintained its pace or even accelerated over the ensuing five decades, the offsetting force of new task creation has slowed, particularly for workers without four-year college degrees (Acemoglu and Restrepo 2019, Autor et al. 2022). Non-college workers have been displaced from factories and offices by computerisation and, for blue-collar workers, also by import competition (Autor et al. 2013), but no new equivalently well-paid opportunities have emerged to attract these workers. As a result, non-college educated workers are increasingly found in low-paid services such as cleaning, security, food service, recreation, and entertainment. These jobs are socially valuable, but they require little specialised education, training, or expertise, and hence pay poorly.

The critical question we face in the new era of generative AI is whether this technology will primarily accelerate the existing trend of automation without the offsetting force of good job creation – particularly for non-college workers – or whether it will instead enable the introduction of new labour-complementary tasks for workers with diverse skillsets and a wide range of educational backgrounds.

There is a case for qualified optimism: generative AI offers an opportunity to complement worker skill and expertise.

Because so many of the routine tasks that workers previously performed have already been automated, a large fraction of current jobs require non-routine problemsolving and decision-making tasks. Empowering workers to perform these tasks more effectively, and to accomplish even more sophisticated decision-making tasks, will require providing workers with better information and decision-support tools. Generative AI is particularly well-suited to this type of information provision. An irony of today’s information era is that people are overwhelmed by information but often lack the time and expertise to parse this information effectively. Generative AI is particularly well-suited to address this problem. With proper development, AI tools can help surface relevant information at the right time to enable better decisionmaking.

Additionally, and closely related, human productivity is often hampered by lack of specific knowledge or expertise, which could be readily supplemented by AI. For example, an electrician might be unable to diagnose a rare problem that she could readily address if given relevant tools or appropriate training. Or a highly trained immigrant to the US may be inhibited from fully using her abilities because of limited English language skills. Generative AI tools can assist in such cases by boosting human expertise, supporting workers in unfamiliar situations, providing on-thespot training, and improving all forms of information translation. Overall, AI holds great potential for training and retraining expert workers, such as educators, medical personnel, software developers, and other workers with modern “crafts” (such as electricians and plumbers).

Finally, while generative AI may take over more of the operational tasks in certain occupations, such as accounting, financial analysis, or computer programming, if developed in the right manner, it could create new demands for human expertise and judgment in overseeing these processes, communicating with customers, and enabling more sophisticated services that leverage these tools.

Several recent studies provide “proof-of-concept” examples that demonstrate how generative AI can supplement expertise rather than displace experts. Peng et al. (2023) demonstrate that GitHub Copilot, a generative AI-based programming aid, can significantly increase programmer productivity.2 In a controlled experiment, the treatment group that was given access to this generative AI tool completed the required programming task about 56% faster than the control group without access to Copilot.

Noy and Zhang (2023) performed a related online randomised controlled trial, focused on writing tasks. Among the set of white-collar workers recruited for the study, half were randomly given access to (and encouraged to use) ChatGPT for writing tasks. Noy and Zhang (2023) found significant improvements in the speed and quality of writing output. Most importantly, the biggest improvements were concentrated among the least-capable writers. Although generative AI did not make the least-skilled writers as effective as the most-skilled writers, it made all writers faster and substantially reduced the quality gap between the two groups.

Finally, Brynjolfsson et al. (2023) evaluated the use of generative AI tools used for providing background information to customer service agents. They also estimated a significant improvement (about 14%) in productivity, and like Noy and Zhang’s study, these gains were the most pronounced for novice workers. Using these AI tools, novice workers were able to reach a level of proficiency within three months that previously took workers ten months to attain.

In all three cases, generative AI tools automate and augment human work simultaneously. The automation stems from time-savings: AI writes the first draft of computer code, advertising copy, and customer support responses. Augmentation happens because workers are called upon to apply expertise and judgment to intermediate between the AI’s suggestions and the final product – whether it is software, text, or customer support.

Promising Applications

Looking forward, we see at least three major sectors where human-complementary AI could be transformative, both for productivity and for shared prosperity.

EDUCATION

Generative AI tools can enable major advancements in education, together with new productivity-enhancing roles for educators. Classroom instruction is hindered by the fact that the teacher must choose one pace at which an entire class proceeds, even if it is too fast for some students and too slow for others. Individualised education programmes and personalised teaching tools can be effective in enabling less prepared students to excel, but these tools are labour-intensive and hence expensive. AI-enabled tools have the potential to vastly improve tutoring and self-instruction. Khanmigo, an app built on ChatGPT-4, is an infinitely patient, highly adaptable tutor that can break complex problems into their constituent parts, walk students step-by-step through solving them, and provide hints and examples along the way without directly answering questions.3 Research shows that Large Language Models (LLMs) can anticipate which parts of a problem humans will find difficult and suggest simplifications to improve understanding (see Bubeck et al. 2023).

These technologies can support educators as well as students. Teachers could focus more of their time on instruction and less on remediation. They could develop richer lesson plans that harness new tools, such as visualisation, simulation, and even real-time “interaction” with fictitious or historical figures. The right AI investments could enable significant gains in student learning, especially among currently underperforming pupils. Such reorganisation would also create a range of new tasks for educators - and would almost certainly necessitate further productive training for existing teachers and hiring of additional AI-savvy teachers.

It is not clear, however, that the private sector (or any public school district) has the right incentives to develop such AI tools. Will AI be developed to reduce gaps between more-capable and less-capable learners – as ChatGPT does for professional writers – or will it instead be deployed to reduce teacher headcounts and substitute for personal attention? This is not a question of what the technology can do but how we collectively decide to develop and deploy it. Applications of AI that appear to reduce labour costs may be more appealing to many cash-strapped school districts in the United States. Thus, our concern is that AI will be used to automate teaching, testing, and grading, rather than for new personalised education-targeting tasks.

HEALTHCARE

Given that one in five US dollars is spent on healthcare, any technology that improves efficiency, lowers costs, or broadens access to the healthcare system has potentially enormous benefits. Generative AI tools can improve healthcare delivery and accessibility, enabling productivity gains and generating valuable new worker tasks. For example, generative AI tools could support expanding scope of practice boundaries, enabling medical professionals at all levels to accomplish a broader range of tasks. Just as nurse practitioners have proved effective at diagnosing, prescribing, and treating (Asubonteng et al. 1995, Li et al. 2012) – tasks formerly done exclusively by doctors – the decision-support capacities of AI can enable a larger set of trained medical professionals to accomplish expert tasks without exclusively relying on the most elite medical professionals. Using AI, qualified nurse practitioners, nurses, and health technicians could diagnose routine health problems, recommend courses of treatment, and more efficiently route patients to further care options.

MODERN CRAFT WORKERS (MCWS)

In line with the electrician example above, generative AI tools can be transformative for modern craft workers (MCWs) more broadly.4 The US is currently undertaking a major infrastructure investment agenda, with growing employment in manufacturing, green energy production, and chip production, among other sectors. Skilled MCWs are in short supply due to an aging population and decades of over-investment in higher education at the expense of valuable vocational training. AI can be used to support training and enable MCWs to carry out a wider range of tasks that require specialised expertise. AI can provide relevant information, real-time instruction, and decisionmaking support in electrical work, plumbing, expert repair, design, and construction, among other activities. The current generation of AI technologies cannot replace the work of MCWs – whose tasks require dexterity, flexibility, and judgment that are far beyond the grasp of current robotics (even AI-enhanced robotics). But AI can enable these workers to do more with their skills by solving a broader and deeper range of applied problems in the field.

What Can Be Done

Tax system: The current US tax code places a heavier burden on firms that hire labour than those that invest in algorithms to automate work (Acemoglu et al. 2020). We should aim to create a more symmetric tax structure, where marginal taxes for hiring (and training) labour and for investing in equipment/software are equated. This will shift incentives toward human-complementary technological choices by reducing the bias of the tax code toward physical capital over human capital.

Labour voice: The direction of AI will have profound consequences for all workers. Creating an institutional framework in which workers also have a voice would be a constructive step – and there is an important role for civil society in pressing for this to happen, including through articulating needs at the local and state level. At a minimum, federal government policy should restrict deployment of untested (or insufficiently tested) AI for applications that could put workers at risk – for example, in high-stakes personnel decision-making tasks (including hiring and termination) or in workplace monitoring and surveillance. Health and safety rules need to be updated accordingly.

Funding for more human-complementary research: Given that the current path of research has a bias toward automation, additional support for the research and development of human-complementary AI technologies could have significant impact. It is hard to target human-complementary work in the abstract. It is feasible, however, to focus on specific sectors and activities where opportunities are already abundant. These include education, healthcare, and MCW training. Just as DARPA orchestrated investments and competitions to foster the development of self-driving cars and dexterous robotics, the federal government should foster competition and investment that pairs AI tools with human expertise, aiming to improve work in vital social sectors.

AI expertise within the federal government: AI will touch every area of government investment, regulation, and oversight, including (but not limited to): transportation, energy production, labour conditions, healthcare, education, environmental protection, public safety, and military capabilities. Developing a consultative AI division within the federal government that can support the many agencies, including regulators, tackling these challenges will support more timely and more effective decision-making at every level.

Technology certification: The federal government can encourage appropriate investments by advising on whether purported human-complementary technology is of sufficient quality to be adopted in publicly funded education and healthcare programmes. For this advice to be meaningful, experts need to be engaged and independent – i.e. they should not be directly or indirectly working for the tech companies. It is hard to attract talent to government or universities when the private sector is paying top dollar for the relevant expertise. This further strengthens the case for building a high-prestige, cross-cutting federal AI service.

Conclusion

There is no guarantee that the transformative capabilities of generative AI will be used for the betterment of work or workers. The bias of the tax code, of the private sector generally, and of the technology sector specifically, leans towards automation over augmentation. But there are also potentially powerful AI-based tools that can be used to create new tasks, boosting expertise and productivity across a range of skills.

To redirect AI development onto the human-complementary path requires changes in the direction of technological innovation, as well as in corporate norms and behaviour. This needs to be backed up by the right priorities at the federal level and a broader public understanding of the stakes and the available choices. We know this is a tall order. But this makes focusing on what is needed even more important.

REFERENCES

Acemoglu, D, and D Autor (2011), “Skills, tasks, and technologies: Implications for employment and earnings”, Handbook of Labor Economics, Elsevier. Chapter 12. Vol. 4, Pt. B, pp. 1043–1171.

Acemoglu, D, D Autor, D Dorn, G Hanson, and B Price (2016), “Import competition and the great U.S. employment sag of the 2000s”, Journal of Labor Economics, Vol. 34, No. 1, Pt. 2, pp. 141–198.

Acemoglu, D, and S Johnson (2023), Power and progress: Our 1000-year struggle over technology and prosperity, PublicAffairs, Hachette.

Acemoglu, D, A Manera, and P Restrepo (2020), “Does the U.S. tax code favor automation?”, NBER Working Paper no. 27052.

Acemoglu, D, and P Restrepo (2018), “The race between man and machine: Implications of technology for growth, factor shares, and employment”, American Economic Review, Vol. 108, No. 6, pp. 1488–1542.

--- (2019), “Automation and new tasks: How technology displaces and reinstates labor”, Journal of Economic Perspectives, Vol. 33, No. 2, pp. 3–30.

--- (2022a), “Tasks, automation, and the rise in U.S. wage inequality.” Econometrica, Vol. 90, No. 5, pp. 1973–2016.

--- (2022b), “Demographics and Automation.” Review of Economic Studies, Vol. 89, No. 1, pp. 1–44.

Asubonteng, P, K McCleary, and G Munchus (1995), “Nurse practitioners in the USA - their past, present and future: Some implications for the health care management delivery system”, Health Manpower Management, Vol. 21, No. 3, pp. 3–10.

Autor, D, C Chin, A Salomons, and B Seegmiller (2022), “New frontiers: The origins and content of new work, 1940–2018”, NBER Working Paper no. 30389.

Autor, D, D Dorn, and G Hanson (2013), “The China syndrome: Local labor market effects of import competition in the United States”, American Economic Review, Vol. 103, No. 6, pp. 2121–2168.

Autor, D, L Katz, and A Krueger (1998), “Computing inequality: Have computers changed the labor market?”, Quarterly Journal of Economics, Vol. 113, No. 4, pp. 1169–1213.

Autor, D, F Levy, and R Murnane (2003), “The skill content of recent technological change: An empirical investigation”, Quarterly Journal of Economics, Vol. 118, No. 4, pp. 1279–1333.

Bruner, R (April 21, 2022), “American companies have always been more anti-union than international ones. Here's why”, TIME, accessed on September 13, 2023. (https://time.com/6168898/why-companies-fight-unions/).

Brynjolfsson, E, D Li, and L Raymond (2023), “Generative AI at work”, NBER Working Paper no. 31161.

Brynjolfsson, E, and A McAfee (2016), The second machine age: Work, progress, and prosperity in a time of brilliant technologies, W.W. Norton.

Bubeck, S, V Chandrasekaran, R Eldan, J Gehrke, E Horvitz, E Kamar, P Lee, Y Tat Lee, Y Li, S Lundberg, H Nori, H Palangi, M T Ribeiro, and Y Zhang, (2023), “Sparks of artificial general intelligence: Early experiments with GPT-4”, arXiv Working Paper no. 2303.12712.

Goldin, C, and L Katz, (2008 [2009]), “The evolution of U.S. educational wage differentials, 1890 to 2005”, The race between education and technology, Harvard University Press, chapter 8, accessed on September 13, 2023. (https://scholar. harvard.edu/files/lkatz/files/the_race_between_education_and_technology_ the_evolution_of_u.s._educational_wage_differentials_1890_to_2005_1.pdf)

Li, J, J Westbrook, J Callen, and A Georgiou, (2012), “The role of ICT in supporting disruptive innovation: A multi-site qualitative study of nurse practitioners in emergency departments”, BMC Medical Informatics and Decision Making, Vol. 12, No. 27, pp. 1–8.

Noy, S and W Zhang (2023), “Experimental evidence on the productivity effects of generative artificial intelligence”, Science, Vol. 381, No. 6654, pp. 187–192.

Peng, S, E Kalliamvakou, P Cihon, and M Demirer (2023), “The impact of AI on developer productivity: Evidence from GitHub Copilot”, arXiv Working Paper no. 2302.06590.

Susskind, D (2021), “Technological unemployment”, in Bullock, J et al. (eds), The Oxford handbook of AI governance, Oxford University Press (forthcoming). Online draft accessed on September 13, 2023. (https://www.danielsusskind.com/s/Susskind-Handbook-Updated-21-September-2021.pdf).

ABOUT THE AUTHORS

Daron Acemoglu is an Institute Professor at MIT and an elected fellow of the National Academy of Sciences, American Philosophical Society, the British Academy of Sciences, the Turkish Academy of Sciences, the American Academy of Arts and Sciences, the Econometric Society, the European Economic Association, and the Society of Labor Economists. He is also a member of the Group of Thirty. He is the author of six books, including New York Times bestseller, Why Nations Fail: Power, Prosperity, and Poverty (with James A. Robinson) and Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity (with Simon Johnson). His academic work covers a wide range of areas, including political economy, economic development, economic growth, technological change, inequality, labour economics and economics of networks. He is also co-director of the MIT Shaping the Future of Work Initiative.

Daron Acemoglu received the inaugural T. W. Shultz Prize from the University of Chicago in 2004, the inaugural Sherwin Rosen Award for outstanding contribution to labour economics in 2004, Distinguished Science Award from the Turkish Sciences Association in 2006, the John von Neumann Award, Rajk College, Budapest in 2007, the Carnegie Fellowship in 2017, the Jean-Jacques Laffont Prize in 2018, the Global Economy Prize in 2019, and the CME Mathematical and Statistical Research Institute prise in 2021. He was awarded the John Bates Clark Medal in 2005, the Erwin Plein Nemmers Prize in 2012, the 2016 BBVA Frontiers of Knowledge Award, and the A.SK Social Science Award in 2023. He holds Honorary Doctorates from the University of Utrecht, the Bosporus University, University of Athens, Bilkent University, the University of Bath, Ecole Normale Superieure, Saclay Paris, and the London Business School.

David Autor is Ford Professor in the MIT Department of Economics and codirector of the NBER Labor Studies Programme and the MIT Shaping the Future of Work Initiative. His scholarship explores the labour-market impacts of technological change and globalisation on job polarisation, skill demands, earnings levels and inequality, and electoral outcomes. Autor has received numerous awards for both his scholarship — the National Science Foundation CAREER Award, an Alfred P. Sloan Foundation Fellowship, the Sherwin Rosen Prize for outstanding contributions to the field of Labor Economics, the Andrew Carnegie Fellowship in 2019, and the Heinz 25th Special Recognition Award from the Heinz Family Foundation – and his teaching, including the MIT MacVicar Faculty Fellowship. In a 2019 article, the Economist magazine labelled him as “the academic voice of the American worker.” Later that same year, and with (at least) equal justification, he was christened “Twerpy MIT Economist” by John Oliver of Last Week Tonight in a segment on automation and employment. Most recently, Autor was recognised as one of two 2023 NOMIS Distinguished Scientist and Scholar awardees.

Simon Johnson is the Ronald A. Kurtz (1954) Professor of Entrepreneurship and chair of the Global Economics and Management group at MIT’s Sloan School of Management. He is also co-chair of the CFA Institute Systemic Risk Council, co-director of the MIT Shaping the Future of Work Initiative, and a board member at Fannie Mae. In 2023, Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity, co-authored with Daron Acemoglu, was long-listed for the FT-Schroders Business Book of the Year and the Baillie Gifford Prize for nonfiction. Professor Johnson is also the coauthor of three books with James Kwak: 13 Bankers: The Wall Street Takeover, The Next Financial Meltdown, and White House Burning: The Founding Fathers, Our National Debt and Why it Matters to You; and, with Jonathan Gruber, Jump-Starting America: How Breakthrough Science Can Revive Economic Growth and the American Dream. Previously, Johnson was the IMF’s chief economist (2007-08), co-director of the NBER Africa Project, a member of the Congressional Budget Office's Panel of Economic Advisers, a member of the US Treasury’s Office of Financial Research advisory group (chair of the subgroup on global risks), and a member of the FDIC’s Systemic Resolution Advisory Committee..

THE CENTRE FOR ECONOMIC POLICY RESEARCH

The Centre for Economic Policy Research (CEPR) is a network of over 1,700 research economists based mostly in European universities. The Centre’s goal is twofold: to promote world-class research, and to get the policy-relevant results into the hands of key decision-makers. CEPR’s guiding principle is ‘Research excellence with policy relevance’. It was founded in the UK in 1983, where it is a Charity, and in November 2019 CEPR initiated the creation of an Association under French law, in order to provide a vehicle for an expansion in France. The members of the Conseil d’Administration of the Association are identical to the UK Board of Trustees.

CEPR is independent of all public and private interest groups. It takes no institutional stand on economic policy matters and its core funding comes from its Institutional Members, projects that it runs and sales of publications. Because it draws on such a large network of researchers, its output reflects a broad spectrum of individual viewpoints as well as perspectives drawn from civil society. CEPR research may include views on policy, but the Trustees/ members of the Conseil d’Administration of the Association do not give prior review to its publications. The opinions expressed in this report are those of the authors and not those of CEPR.

Chair of the Board

Sir Charlie Bean

Founder and Honorary President

Vice Presidents

Richard Portes

President

Beatrice Weder di Mauro

Maristella Botticini

Philippe Martin

Ugo Panizza

Mar Reguant

Hélène Rey

Chief Executive Officer

Tessa Ogden

Policy Insight 123: Can we Have Pro-Worker AI?