From the discussions of the preceding chapters, a number of key findings emerge.
1. Advances in information technology (IT) are far from over, and some of the biggest improvements in areas like artificial intelligence (AI) are likely still to come. Improvements are expected in some areas and entirely new capabilities may emerge in others.
The past decade has yielded significant scientific and technological advances in many areas, including ubiquitous networking and sensing, AI and its subdisciplines of machine learning, computer vision and speech recognition, robotics, the Internet of Things, and other areas.
At the same time, productivity growth has slowed, a fact often read as an indication of decreased technological progress. However, the committee’s judgment is that the productivity data are not necessarily inconsistent with a period of significant technical change, because they measure somewhat different concepts and because there are important lags associated with adoption, co-invention, changes in organization, and the updating of skills that are typically required to translate technical change into economic value.
As companies increasingly adopt AI-based technologies to produce new products, their increasing research and development (R&D) investments in these areas are likely to further technical progress over the
coming decade. Beyond invention of altogether new technologies, we can also expect to see strong influences over the coming decade from diffusion and maturation of technologies that already exist today in early forms in research laboratories and leading-edge technology companies (e.g., self-driving cars, conversational AI agents). Both will affect the workforce; it remains to be seen how such advances will impact productivity.
2. These advances in technology will result in automation of some jobs, augmentation of workers’ abilities to perform others, and the creation of still others. The ultimate effects of information technology are determined not just by technical capabilities, but also by how the technology is used and how individuals, organizations, and policy makers prepare for or respond to associated shifts in the economic or social landscape.
Technology has been augmenting and replacing human labor in many tasks for hundreds of years, and we can expect this trend to continue. Increasingly, we will see robots used to automate more complex physical tasks in manufacturing, transportation, retailing, and many other industries, and AI to automate knowledge-based tasks.
Because most jobs involve multiple subtasks, and because technology typically targets specific tasks, one common impact of technology is to shift the distribution of tasks the human worker performs in a job (e.g., authors today spend less time proofreading for incorrect spelling, enabling them to spend more time on the content of what they are writing). Technology also makes new tasks and new jobs possible, transforming the nature of work in many, and ultimately most, industries.
While technology may threaten the existence of some jobs via automation, it is also producing new modes of education and training that are more accessible. It has also led to new opportunities for freelance employment via Internet platforms that can match job seekers to jobs in real time and to opportunities to work from home. The net impact of technology, mediated by the decisions of many organizations (businesses, governments, and philanthropic entities) and individuals, is multifaceted. Its full impact is not predetermined, but will depend on the decisions of governments, companies, and individuals about how to use technology and how to prepare for or respond to associated shifts in the economic or social landscape. Technologists, policy makers (such as private-sector managers and public officials), and other leaders have the power to design IT and deploy it for the benefit of society, driven by a broad discussion of what impacts are desirable and a deeper understanding of how design, deployment, and policy decisions can achieve these impacts.
3. The recent increase in income inequality in the United States is due to multiple forces, including advances in IT and its diffusion, globalization, and economic policy.
While technologies enable the automation of some tasks, the demand for labor is not uniformly affected. Changes in demand affect wages and employment, and in turn the underlying mix of skills demanded in the economy. Over the past 20 to 30 years, a large fraction of income disparity has stemmed from changes in demand (due in part to changes in technology) and changes in supply (due to changes in the quality, quantity, and types of education). Employers report shortages of job candidates with needed skills at the same time that salaries for employees with high school degrees have dropped. The wage premium for college degrees remains quite large, even as salaries for bachelor’s degree holders have leveled off on average in recent years. Companies that are heavy users of technology and automation often require dramatically fewer workers than the dominant companies of the past century. This reduced need for labor has the effect of further skewing the distribution of income and wealth created by new companies in this category and could limit the ability of reshoring to create new jobs. At the same time, while automation has contributed to inequality, it also is a key driver of productivity and economic growth. This growth can enable more options for easing the implementation of public policies that improve the equity of economic outcomes in the United States. However, such policies cannot be implemented without public and political support.
4. IT is enabling new work relationships, including a new form of on-demand employment. Although current digital platforms for on-demand work directly involve less than 1 percent of the workforce, they display significant growth potential.
Many employers are increasingly viewing their relationship with employees as a short-term commitment rather than a lifelong investment. As Manpower Group CEO Jonas Prising recently put it, “Employers have gone from being builders of talent to consumers of work.”1
While freelance or on-demand work has long been part of the economy, new IT platforms have changed this aspect of the economy significantly. Internet platforms now match drivers to riders (e.g., Uber, Lyft), producers of goods to buyers (e.g., Etsy, eBay), computer programmers to employers (e.g., Upwork), and high-end consultants to businesses (e.g., HourlyNerd). This “on-demand” or “gig” economy has expanded rapidly, although current government labor statistics make it difficult to track.
Some workers report that they enjoy the flexibility of working only when they wish to, and some workers use these platforms as a secondary form of income. For workers for whom it is their only job, the question of how to obtain health and other standard benefits is important, and some policies are being considered to make it easier for such freelancers to obtain and carry benefits from job to job.
Beyond enabling this on-demand economy, advances such as more widespread access to the Internet have also enabled more full-time employees to work at home part of the time, enabling more full-time employers to offer their employees flexible work hours and work locations.
5. As IT continues to complement or substitute for many work tasks, workers will require skills that increasingly emphasize creativity, adaptability, and interpersonal skills over routine information processing and manual tasks. The education system will need to adapt to prepare individuals for the changing labor market. At the same time, recent IT advances offer new and potentially more widely accessible ways to access education.
While education must change to deliver new content and to teach new skills, the details of exactly what should be taught, and how, are not well understood. There is evidence that “softer” skills have been increasingly valued in the labor market. At the same time, the 20th-century model of degree completion followed by a semipermanent job based on that education is yielding to a model where degree completion is followed by more specialized on-demand education over one’s entire career (which may include multiple occupations). The most logical developers of new educational approaches are local, state, and federal governments, as well as researchers and research agencies.
There are significant opportunities for IT to be used to advance educational strategies and delivery. For example, online education companies such as Coursera, EdX, and Udacity have begun to experiment with new modes of Internet-based continuing education. While IT is likely to enable broader access to education, individuals without the opportunity or incentives to access it are at risk of being left even further behind, potentially reinforcing existing racial, ethnic, and socioeconomic disparities in society. While these tools show promise, their efficacy and the extent to which they can be truly democratizing remain to be seen.
6. Policy makers and researchers would benefit significantly from a better understanding of evolving IT options and their implications for the workforce. In particular, (1) sustained, integrated, multidisciplinary research and (2) improved, ongoing tracking of workforce and
technology developments would be of great value for informing public policies, organizational choices, and education and training strategies.
Despite much anecdotal evidence suggesting that big changes are under way, surprisingly little data are available to help determine which anecdotes correspond to significant country-wide or economy-wide trends and understand the nature of these changes and how potential policy choices can influence them.
For example, although there is much anecdotal evidence that the on-demand economy is growing, better data are needed to understand issues such as what is driving it (e.g., how much is due to the business cycle versus improvements in technology or the invention of new business models?); whether it is particularly important or attractive in specific sectors of the economy (e.g., computer programmers versus taxi drivers); and whether it is lowering the barrier for creation of new businesses (e.g., by reducing the need to hire full-time employees in early stages).
New data across all aspects of technology’s progress, its diffusion into firms and products, and its impact on the economy, along with tools to analyze these data, could shed important light on the types of public policies that might optimize the benefits of technological advances for the workforce and society. As new data sets emerge, it will be important to design them to accommodate potential future research needs.
Achieving the goal of a more evidence-based understanding of the forces at work depends on overcoming barriers to data access for the research community. A critical step is enabling collaboration between U.S. statistical agencies. Providing these agencies with access to and integration of (1) federal tax information for statistical purposes and (2) additional administrative data from federal and state sources would provide new opportunities for economic research. In March of 2016, the federal government passed a law to establish the Commission on Evidence-Based Policymaking; this group is likely to address some of these challenges and may facilitate enhanced data access for the research community.2
The interplay between technology and work is complex and changing, and important public policy issues are already arising. There are many open questions that policy makers may face. For example:
- How can the United States make use of technology to maximize overall economic growth while maximizing access for everyone to the economic and other benefits afforded by new technologies?
2 The Evidence-Based Policymaking Commission Act of 2016, Public Law 114-140.
- What policies, resources, and practices would ease transitions for workers forced to change occupational fields due to technological change?
- How can we anticipate and actively guide the future impacts of newly developed technologies before they occur?
While it is not within this committee’s charge to recommend specific policy actions, it is within its purview to advocate for well-informed policy discussions about how IT is affecting the workforce, including job opportunities and workers’ quality of life. The committee believes that the foundational knowledge and insights essential to an informed policy debate can best be attained through a strategic research program to better track the changes that are occurring and to understand the mechanisms by which advances in technology influence our economy, workforce, and society. Such a research program will be important for helping stakeholders address productivity growth, job creation, and the transformation of work.
The committee recommends that federal agencies or other organizations that sponsor research or collect data relevant to technology and the workforce establish a sustained, multidisciplinary research program to address the many important questions about how IT is changing or how it might change the nature of work and the U.S. national economy. This program should
- Target a deeper understanding of how choices about technology use or functionality can affect the workforce in order to inform the design of technologies and policies that will benefit workers, the economy, and society at large;
- Emphasize integration of micro- and macro-level research methods from disciplines including the social sciences, economics, computer and information sciences, and engineering; and
- Establish and facilitate the use of new data sources, tools, methods, collaborations, and infrastructures to facilitate such research while protecting privacy with appropriate data-management practices.
The committee envisions this as a highly multidisciplinary, sustained research program that integrates a range of strategies, from bottom-up ethnographic studies to large-scale quantitative approaches, including survey data, administrative data, research-generated data, and privately collected data. This research should take advantage of what can be learned from the experiences of other countries faced with similar trends, to learn
from their experiences with different policies regarding education, worker benefits, and so on, as well as learn from historically analogous events, such as the advent of farm and factory automation during the early 20th century. There are myriad opportunities for cross-fertilization across disciplines such as machine learning, data science and statistics, economics, sociology, anthropology, and other social sciences.
In some instances, ethnographic components will be crucial for understanding the right questions to ask and provide insight about how worker experiences may vary. The quantitative components are critical for understanding the breadth of developments and, perhaps more importantly, where within the economy effects are located (which occupations, types of organizations, and geographical regions). It is critical that research on work consider variations between different demographic groups, geographic regions, industries, technologies, and occupations.
Such a research program should span a range of research themes, which the committee describes below. These issues fall under the U.S. National Science Foundation’s “Big Ideas” for investigation, “work at the human-technology frontier,” announced in 2016.3
Theme 1: Evaluating and Tracking Technological Progress
While many researchers or individuals who follow technology trends may have a strong understanding of current technological capabilities, a uniform strategy is currently lacking for tracking new developments across a broad range of technology fields, their capacity for automating or augmenting human work functions, and the degree to which they are diffusing into industry and firms. More rigorous measures and awareness of the state of the art of technology would help to signal the potential for corresponding workforce impacts.
Research to develop new ways of evaluating, tracking, and projecting technological progress would help enhance understanding of the impacts of technology on the workforce, inform strategies to help prepare for likely changes, and reveal what kinds of indicators might point to upcoming disruptions.
Such research should include objectives such as the following:
- Further develop, refine, and test strategies for classifying technological capabilities in terms of the human skills and tasks they can or could replace. Several strategies making use of the O*NET (Occupational Information Network) classifications have been reported, which could
3 National Science Foundation, 2016, “10 Big Ideas for Future NSF Investments,” https://www.nsf.gov/about/congress/reports/nsf_big_ideas.pdf, accessed December 2016.
serve as a starting point. New strategies should also be explored—in particular, to consider not only the tasks that might be automated but also the context and systems within which they are conducted.
- Identify key indicators that could signal the extent of the impact of developments in a given technological field.
- Develop new mechanisms to track and forecast technological and economic changes of particular relevance to the future of the workforce, potentially via the development of metrics that would quantify such changes and their impact.
Key research questions include the following:
- What technological fields are advancing the most rapidly? What new fields are emerging?
- How can we track and predict the types of human tasks that can be automated? Which technologies are having the biggest impacts in industry, and how might we best track their adoption?
- What are key benchmarks that might indicate the imminence of groundbreaking progress with significant economic impact?
Multidisciplinary research, including economists and technologists, should investigate possible input categories, measures, and output scales for a set of useful indexes, such as the following:
- Technology progress index. Analogous to the Consumer Price Index, this index would measure and summarize the current state of technology and its impact on the economy and workforce, aggregating quantitative data from diverse sources. For instance, it could track progress in computational hardware (Moore’s Law), data storage costs and speeds, communication speed, saturation of high-speed Internet coverage across the United States, progress on specific technologies such as speech recognition and computer vision using standard benchmarks (e.g., the ImageNet benchmark for testing computer-vision algorithms and the COCO benchmark for image captioning), and so on. Such a generalized index could make it easier to understand how technology is advancing, enabling better determination of how and whether rates of technological progress translate into productivity growth.
- AI progress index. The potential for AI to significantly advance automation capabilities make this a field of particular interest to track. An AI progress index could focus specifically on the progress in AI and machine learning. This index would track progress on specific technologies, such as speech recognition and computer vision, robotic dexterity and mobility, autonomous vehicles, medical diagnosis, legal and investment advice,
ability to converse with humans in formats like the Turing Test, proficiency in various games, structured and unstructured problem solving, and other areas of AI, using standard benchmarks (e.g., the ImageNet benchmark for testing computer-vision algorithms). Relatedly, a Turing Olympics or Turing Championship,4 consisting of a variety of tasks, could be organized, providing both a structured way to track progress and a set of milestones to motivate future work.
- Organizational change and technology diffusion index. This index might look at the gap between the productivity of frontier firms and the median firm, adoption rates of technologies, the rate of start-ups, patenting activity, business dynamism, labor shortages and skills mismatching (e.g., the Beveridge curve), and the time it takes for technologies to go from invention to implementation across industries. Diffusion of AI technologies could be tracked via a market-based approach, perhaps analogously to that used in Robo Global’s index.5 Various measures of performance of current AI systems could also be used to track foundational capabilities in AI.
In addition to the above quantitative indexes, methods for modeling and predicting technological progress at various confidence intervals should then be developed for specific fields. In addition, a forward-looking panel of technical experts could be created to forecast technological progress and its impacts, analogous to groups of expert economists who attempt to forecast the economy. These forward-looking subjective forecasts can be combined with data from the above quantitative indexes to form a broad view of the state and future direction of technology and its impacts. While expert forecasts, whether of technology or other topics, generally have a poor record, practices exist that are correlated to useful predictions.6
Together, such indicators could provide important knowledge for many decision makers, from individuals considering possible training and job changes, to educational institutions designing their curricula, to the Federal Reserve Board assessing the economy. Furthermore, these methods of quantifying technological advances and their diffusion into
4 See, for example, the webpage for the “Beyond the Turing Test” workshop at the Twenty-Ninth AAAI Conference on Artificial Intelligence (AAAI-15) on January 25, 2015 at http://www.math.unipd.it/~frossi/BeyondTuring2015/.
6 Such practices are discussed in the context of economic and political predictions in P.E. Tetlock and D. Gardner, 2016, Superforecasting: The Art and Science of Prediction, Random House, New York.
industry could provide a basis for testing the various hypothesized explanations for the slowing of productivity growth discussed in Chapter 3.
Theme 2: Technology Adoption and Impact Within Organizations
Technologies can greatly affect the type and nature of tasks conducted by individual workers. Whether they can completely replace a human worker depends on the specific context in which individual tasks are conducted. Understanding the impact of technology on workers within a firm requires an understanding of the organizational systems through which individual tasks combine to meet the company’s goals. Research on how different industries use technology to organize their operations, allocate tasks, and perform specific functions, at both the micro- and macro-level scales, should be undertaken to provide a firm- and industry-level window into the impacts of technology on employees in a given industry or at a given organizational level.
Key research questions include the following:
- How are decisions relating to technology and labor made within these organizations? How does this vary across firm/enterprise types and life cycles, and over time?
- Is the increasing importance of online data (e.g., for targeted marketing) creating a less even playing field for new start-ups that lack the legacy data of incumbents?
- Are businesses’ production processes being segmented differently? What kinds of changes are emerging, for example, with respect to outsourcing?
- How does the adoption of IT into work processes affect the distribution of work tasks for employees in specific industries?
Theme 3. Impacts of Policy Choices
Important public policy issues are already arising as a result of the changing technological landscape and its impact on the U.S. workforce; the urgency of these issues may increase. Research to inform national and regional policy decisions and strategies to enable innovation and the use of IT to maximize overall economic growth while maximizing access for everyone to jobs, opportunities, and the economic and other benefits afforded by technologies is critical.
This research should aim to understand which policies, resources, and practices would (1) mitigate technological unemployment, (2) ease transitions for workers forced to change occupational fields due to technological change, and (3) provide opportunities for actively guiding the future
impacts of technology development and deployment before they occur. In particular, this could include assessment of the efficacy of new social policies—for example, the impact of a guaranteed income or specific proposals for retraining and continuing education initiatives—in the United States. It could also include evaluation of strategies for providing incentives to private companies to share data that will enable better monitoring of the impacts of technology on jobs and workers.
Although this report focuses on the U.S. workforce, many of the same issues regarding the impacts of IT are being faced in other countries, whose circumstances could serve as a pool of diverse case studies to help illustrate the variety of mechanisms by which IT can impact the workforce and how related social and economic policies can mitigate or exacerbate any negative impacts. Similarly, U.S. history includes episodes of automation and technological developments that have had major impacts on the workforce and can demonstrate how related social and economic policies can mitigate or exacerbate any negative impacts. Although today’s software-driven changes are different in many ways from the impact of earlier technological advances, there are lessons to learn from both historical and international case studies.
Key research questions include the following:
- What are differences across existing case studies, in their demographics, pool of available jobs, training level of their workforce, exposure to international trade, and other factors that may influence how technological advance leads to augmentation or replacement of workers? How can these different case studies be used to build a more robust model of how technologies can impact the workforce and how policies can influence these impacts?
- What is the inventory of policies attempted across different countries and throughout history regarding continuing education, workforce benefits, social safety nets, and other relevant issues?
- What data exist regarding the impact of past policies adopted in different contexts? What projections might be made regarding the impact of such policies on particular sectors of today’s U.S. workforce and economy?
- What are the global interactions that influence how technological advances impact the workforce, given today’s highly multinational economy? How do differing labor pools and labor conditions in different countries lead to different types or rates of influence of new technologies on the workforce? Do these differences result in an uneven playing field in the rate of adoption of new technologies?
Theme 4: Working with Emerging Technologies
As emerging technologies are increasingly implemented in different industries, individuals must learn how to interact with and successfully complete tasks with them. One key implication of this shift is a change in the nature of decision making, teamwork, and organization. In some cases, robots are displacing positions that would have previously been assigned to people,7 and other members of the team must learn to collaborate with nonhuman members and interact with technology in the place of human intelligence. Organizations are also increasingly relying on teams that primarily use virtual tools to collaborate.8 These are just two examples; there are many current changes occurring that are affecting the way individuals collaborate.
More research is needed examining how these teams may function differently than traditional teams and how this will impact organizations. Research is also necessary to uncover under what conditions these teams are as effective as, or more effective than, traditional teams.
Key research questions include the following:
- How do teams using these technologies function, and how is that different from teams without them?
- What impact do different technologies have on the functioning of teams? For instance, are there discernible differences between those physically augmenting human actions, such as robotic arms, versus those providing data or making decisions?
- How do emerging technologies impact organizational outcomes and, ultimately, high-level organizational trends?
- What conditions affect the effectiveness of teams using emerging technologies? When are they as effective as or more effective than traditional teams?
Theme 5: Societal Acceptance of Automation Technologies
While tasks cannot be automated without mature technologies capable of performing a given function, the mere existence of technological capabilities does not guarantee that a given technology will be deployed. Economic costs and benefits will influence decisions to deploy technolo-
8 T. Minton-Eversole, 2012, “Virtual Teams Used Most by Global Organizations, Survey Says,” Society for Human Resource Management, July 19, https://www.shrm.org/resourcesandtools/hr-topics/organizational-and-employee-development/pages/virtualteamsusedmostbyglobalorganizations,surveysays.aspx.
gies, but other factors may also be at play. In some contexts, people (either workers or consumers) may prefer to interact with a human over a machine. This may reflect the existence of important yet largely invisible and unremunerated human skills that can easily be missed in existing skill categories and national statistics but are valued by individuals. Consumer behaviors and worker preferences and bargaining power drive markets; understanding automation’s behavioral economics will be important for understanding its adoption patterns. Additional human factors and the social, philosophical, and psychological dynamics of automation should be explored.
Key research questions include the following:
- What factors determine whether or not individuals will accept automated analogues to traditionally human-performed tasks? How does this vary with the type of technology or task?
- Can such behavior or preferences be modeled? To what extent do current behaviors conform to existing economic models? How might this change as technologies become more established?
- Can existing deployments of automation (such as digital assistants, ATMs, or self-service checkout lines) serve as models for understanding new applications of AI, machine learning, or robotics?
- What opinions and perceptions does the public have about automation in various contexts? How are these related to established societal norms, and how might this inform decisions about what to automate and how to design automated systems?
Theme 6: Changing Labor and Skills Demands and Implications for Education and Training
Changes in technology use affect the roles of workers and contribute to changing labor and skills demands. This creates challenges for individuals as they plan and adjust their career strategies, as well as for employers, educational institutions, and policy makers. The economic insecurity felt by many workers underscores the importance of understanding the interplay of technology with jobs, wages, and opportunity, and not simply looking at technology in isolation. Furthermore, there is a good likelihood that already disadvantaged groups will bear the brunt of the costs of automation.9 Research examining changing labor and skills demands in specific industries and occupational fields over time, along
9 In addition, there is some evidence that the rise in disability rolls may, in part, reflect a lack of employment prospects for some groups. The extent to which automation of jobs contributes to this is an open question. See D.H. Autor and M.G. Duggan, 2007, Distin-
with regional variations and associated policy implications, would help to provide a basis for understanding and anticipating future trends and for informing education, training, and retraining strategies.
This research should examine existing trends in both authoritative economic statistics and emerging data sources and methods, identify correlations to changes in technological capabilities and diffusion in different fields, and hypothesize and test causal relationships using results from research about organizational practices.
Key research questions include the following:
- What types of worker skills and traits are currently becoming more or less valuable in specific sectors of the labor market?
- Do these changes correlate to the development or deployment of specific technologies that enable, augment, or replace such skills?
- How might measures and projections of technology impact and diffusion enable modeling of changing skills demands and income distributions, and how much might such models be tested?
- How do the uses of specific technologies in specific jobs impact work conditions and worker job satisfaction in the most common occupational fields?
- Under what scenarios might the creation of new jobs outpace the reduction in existing jobs, and vice versa? How is this balance likely to change over time? Which job types are likely to be most affected?
- What kinds of indicators might be useful for identifying disruptions to the types of skills required by workers in a given occupational field? What data-collection strategies would enable the development of such indicators?
The new workplace requires a workforce trained for expertise in areas that will drive the future economy and with the flexibility to adapt to rapid change. The U.S. education system has already begun to evolve in response at the levels of primary, secondary, vocational, college, graduate, and continuing education. However, it is far from clear at this point exactly what children should be taught and how to best prepare displaced workers for the future. It is unclear how to best take advantage of the new opportunities for online or other IT-enabled training, how to construct the most effective online courses, and how to best incentivize students and workers to complete the courses they begin. Because education will largely determine the success of the United States in responding to the changing workplace, a better understanding of effective strategies is criti-
guishing income from substitution effects in disability insurance, American Economic Review 97.2:119-124.
cal. Some insight into changing skill needs can be inferred from how skill demands are currently changing; additional insights might be gleaned by a partnership between computer scientists, labor economists, and education researchers to discuss the kinds of technology capabilities that are likely to emerge in coming years.
Research in this area should aim to assess educational and training needs based on understanding of skills demands driven by technological change, and ways in which technology can be best used to prepare and train the future workforce.
Key research topics include the following:
- Educational needs. What skills will be most valuable for young students and for employed and unemployed adults seeking better jobs? What are the best practices for teaching these different skills, and what new innovations are possible? What are the current and emerging technological substitutes and complements to these skills?
- Education delivery. How can education best be delivered? How can traditional classroom models be augmented by online education, workplace apprenticeships, peer-to-peer education, and other models for optimal success rates? How might new approaches that leverage IT, such as gamification or simulation-based learning, be deployed to improve learning outcomes?
- Education access and incentives. How can the benefits from educational opportunities be extended to all segments of society, including students who cannot or will not attend 4-year colleges? What programs are needed to assure that primary and secondary schools have the tools they need to help students?
- Education policy. Various countries have policies that provide for worker education. For example, workers in France accrue a credit of 20 hours of paid time for continuing education for each year they work.10 What would be the expected impact if the United States were to adopt similar policies?
Additional questions include the following:
- What are the factors contributing to gaps in educational attainment, and is the United States on a trajectory to narrow or widen current gaps in educational attainment? Where can learning technologies help to
10 Y. Lochard and B. Robin, 2009, “France: Collective Bargaining and Continuous Vocational Training,” EurWORK, http://www.eurofound.europa.eu/observatories/eurwork/comparative-information/national-contributions/france/france-collective-bargaining-and-continuous-vocational-training, accessed May 2016.
bridge these gaps, especially for women, underrepresented minorities, people with disabilities, and the economically disadvantaged?
- What is the mean time to development for the most valued skills, and how do these relate to job hopping?
- Are certificate and test-based certificate programs serving employment goals? What does the future look like as certificate programs continue to proliferate and compete for attention?
- What is the current state of vocational education opportunities? Are these meeting current needs?
- What changes should be made to what is being taught, how it is being taught, and who is being taught?
- What positive outcomes can online learning effectively offer in both general education and training programs to prepare individuals for the workforce?
- How can online or other IT-assisted learning techniques be best leveraged to assist in retraining or just-in-time learning for individuals undergoing work transitions?
Theme 7: The On-Demand Economy and Emerging Modalities for Organizing Work
The emergence of an on-demand economy, in particular for ride-sharing services and crowdsourced work marketplaces, has created great interest and excitement. However, there is little information about the extent of its impact on the economy and the workforce. Research on the ability of authoritative economic and labor statistics to capture this impact, and more comprehensive and persistent strategies for measuring it, are needed. In addition, social science research to illustrate the rights, protections, and autonomies of workers, and how on-demand jobs fit into workers’ lives and careers, is needed to understand the impact and need for policies in this domain. Through this work, the potential for on-demand jobs to effectively provide or augment employment for unemployed or low-income workers, as well as other associated advantages and disadvantages, should be elucidated.
Key research questions include the following:
- To what extent do current measures of business dynamism capture activity in the on-demand economy? How can they be improved?
- To what extent do on-demand workers in this sector lose income and legal protections (the right to unionize, the right to minimum wage, and the right to overtime pay)?
- To what extent do on-demand workers gain or lose control over their schedules and other types of autonomy?
- Where do on-demand jobs fit into the career path of workers? What are the longer-term implications of this type of work for individuals and society?
Physical and geographical boundaries of work have shifted over time, with significant impacts on worker experience and job availability and access. Technology has enabled this to a large extent, and the emergence of new technologies could continue to change this landscape. Research in this area could aim to elucidate the role of technology in shifting where and how work is conducted and lay the groundwork for anticipating future changes and opportunities.
Key research questions include the following:
- How might new technologies continue to change the costs and benefits of telecommuting, globalization, or other aspects of the geography of work? What are the technological and nontechnological drivers of these changes?
- How might developers and organizations design technologies and strategies for technology use that will create more work for those at the margins of society, make better use of existing workers, and yield deeper satisfaction for workers?
- What is the impact of new modalities of organizing work on organizational forms such as corporations, limited-partnerships, freelancing, volunteer organizations, and so on?
Theme 8: New Data Sources, Methods, and Infrastructures
As described at the outset of this chapter, the interplay between technology and work may be studied at various levels, with different lenses provided by a range of academic disciplines. Part of the challenge of developing a holistic understanding of the interplay between technology and work stems from the vastness of these interactions, which is accompanied by gaps in information gathering and a need for deeper interdisciplinary collaboration. How can researchers design comprehensive strategies for understanding the role of a given technology in a given occupational field, both from the top down and the bottom up?
New sources, methods, and infrastructure to enable the collection, aggregation, and distribution of a diverse range of data are needed to support investigation of the preceding themes to understand what technology advances are in fact occurring, along with their impact and the
mechanisms by which they affect workers and the economy.11 Below, the committee discusses data strategies that would advance understanding of the impacts of IT and automation on the U.S. workforce. The committee recognizes that new government data-collection efforts such as those discussed below would most likely require additional resources.
Updating and Augmenting Authoritative Data Sources
As discussed in Chapter 5, much useful quantitative data comes from government sources such as the Bureau of Labor Statistics (BLS) and the U.S. Census Bureau. Updates to existing government survey tools and instruments with methods (e.g., questions, sampling techniques) that better target the impact of new technologies and related work processes and task structures on organizations, workers, and management capabilities would greatly assist researchers. The Current Population Survey (CPS) (of households) and the Current Employment Statistics Survey (of establishments) could potentially benefit from new questions that can help to assess the impact of new technology on workers as well as the size of the on-demand economy. For example, the planned revival of the Contingent Worker supplement to the CPS is an important opportunity to obtain well-defined information about worker participation in the on-demand economy; sustained efforts to collect such information would yield an important longitudinal database. Only these large-scale surveys, conducted at regular intervals, have the capacity to track such phenomena over time.
Development of New Data Sources and Methods
Much data about investment, employment, and sales that could be of great value to researchers aiming to unravel technology’s role in workforce and economic trends are currently held by private-sector organizations (discussed in Chapter 4). Opportunities for obtaining useful research data via partnerships with private industry should be pursued.
Furthermore, creative efforts to mine and interpret “born-digital” data from a range of sources could lead to new, real-time monitoring of key employment and economic trends. Applications of contemporary analytical tools based on data mining and machine learning should be pursued. Researchers in this area might pursue the following:
11 See also the efforts of a panel of the Committee on National Statistics of the National Academies of Sciences, Engineering, and Medicine on Improving Federal Statistics for Policy and Social Science Research Using Multiple Data Sources and State-of-the-Art Estimation Methods at http://sites.nationalacademies.org/DBASSE/CNSTAT/DBASSE_170268.
- New methods of obtaining, augmenting, or validating the information provided by authoritative sources that may be updated in real time and do not depend on existing cost- or labor-intensive methods; and
- New types of indicators of technology diffusion, shifts in labor or skills demands, and the connection between the two.
Classifications and Indicators for Job Types and Categories
Researchers face additional challenges with respect to monitoring. Key research questions include the following:
- How might the classification of job types, categories, and characteristics be standardized to enable identification and longitudinal tracking of changes due to technology-related trends?
- To what extent are human attributes—including creativity, empathy, interpersonal skills, leadership, mentoring, and physical presence—currently valued in the U.S. labor force? How might this be tracked in the future?
- How can independent, freelance, and contingent labor be characterized and measured?
- How can the deployment of new worker-engagement models and trends, such as the more general “task markets” that compose the intellectual and physical abilities of machines and people in new, more flexible ways be monitored and understood?12
Combination of Micro- and Macro-Level Data and Methods
The importance of targeted qualitative, ethnographic, microscale, and case-study and fieldwork investigations must not be overlooked. As many researchers come to rely on identification of trends—or even predictive models—based on very large and, increasingly, born-digital data sets, the underlying causes or mechanisms behind these trends may not be clear. Qualitative and interview-based approaches—for example, investigation of the specific deployment of a particular technology into the work flow of an organization, how workers are trained on it, and its impact on their job satisfaction—could be used to test conclusions that may be inferred from macroeconomic trends, and likewise provide insights into drivers of trends that may emerge on the horizon at the macroscale. Moreover, data
12 D. Shahaf and E. Horvitz, 2010, “Generalized Task Markets for Human and Machine Computation,” AAAI 2010, July.
from field studies and related methods are likely to raise questions that will fruitfully guide more macro-level research.
New Infrastructure and Partnerships for Aggregation, Sharing, and Collaboration
Beyond simply collecting and analyzing relevant data, the development of an infrastructure to enable access to, interoperability of, and new multidisciplinary strategies for using data could be a great benefit for researchers, policy makers, and others. In particular, this infrastructure could enable researchers to access and share the very large amounts of relevant digital data being created, gathered, and shared by private companies and via emerging technologies and platforms (such as the Internet of Things and mobile phones) and integrate it with data collected by governments. Such an infrastructure could provide insights about the labor market with much more detail and at a higher frequency than ever before.
One strategy could be to establish data clearinghouses or data centers, ideally virtual, that permit integration of real-time data from U.S. statistical agencies with that of companies, such as the data centers currently used by firms like Burning Glass, LinkedIn, Google, ADP, and many others, with these combined core data sets tracking changes in innovation, changing skill demands, productivity, and the workforce. Such an approach could enable continuous monitoring of where novel jobs are being created and how skills requirements are changing for specific occupational fields, which could be correlated to trends in technology. This could also be used to facilitate matching between employers and job applicants, or current or potential workers with training or educational pathways.
More generally, many believe that the future of economic statistics lies in the integration of survey, administrative, and commercial data,13 and the use of federal administrative data was named a priority of the executive branch for fiscal year 2016.14 Such integration is expected to revolutionize key national indicators and also allow much more timely and granular tracking of economic data by detailed geography, industry, occupational fields, worker characteristics, and other classifiers. Academia can play a critical role as honest brokers between the public and private sectors with respect to these data sets. This would likely require address-
13 See, for example, efforts of the American Economic Association and the Sloan Foundation to identify sources and procedures for researcher access to Federal Administrative Data at https://www.aeaweb.org/about-aea/committees/economic-statistics/administrative-data.
14 Office of Management and Budget, 2015, Fiscal Year 2016: Analytical Perspectives of the U.S. Government, https://www.whitehouse.gov/sites/default/files/omb/budget/fy2016/assets/spec.pdf, Chapters 7 and 16.
ing organizational barriers to information and data sharing and standardization of information-sharing levels and ranks across organizations. There are also measurement and methodological issues in developing key indicators from the new data infrastructure that will emerge. These could build on the newly created Federal Research Data Centers—which now house data from the Census Bureau, BLS, the National Center for Health Statistics, and the National Center for Education Statistics, with plans to add data from the Internal Revenue Service and the Bureau of Economic Analysis—and could potentially bring private-sector sources into the mix from willing partners. The new national Commission on Evidence-Based Policymaking may reflect efforts in this direction.15
Such efforts may be frustrated by existing and potentially outdated government regulations that constrain the ability of government to share certain data sets with researchers. While regulations to protect the privacy of individuals are well justified, they may not reflect current approaches to protecting privacy while making data available for analysis. In any case, there is a general and persisting need for research on strategies for protecting the privacy of individuals’ data beyond the context of this particular research agenda.
Finally, creative strategies for integrating qualitative and quantitative data from research pursued through the lenses of social science, technology, and economics could help to build a holistic research community and transform the ability to interpret technology and workforce trends, evaluate their consequences, and inform future policies, with great potential benefit for the nation.
15 The Evidence-Based Policymaking Commission Act of 2016, Public Law 114-140.