Executive summary
Computer scientists are working on reproducing all human skills with computer capabilities. The development of these capabilities will have far-reaching implications for work and education.1
This report describes the results of an exploratory project to understand current computer capabilities with respect to one set of human skills: the three general cognitive skills of literacy, numeracy and problem solving with computers.2 The project uses the OECD’s Survey of Adult Skills, derived from the Programme for the International Assessment of Adult Competencies (PIAAC), to understand changes in skill demand in the recent past and then to assess the computer capabilities that could change skill demand further in the near future.
The three cognitive skills measured by PIAAC are developed during compulsory education and broadly used by adults at work and in their personal lives. The test involves practical problems that would be familiar to most adults who have completed secondary education and live in developed countries. The test does not assess actual tasks used in specific occupations, but the questions are designed to be similar to the kind of tasks that occur in many different occupations that require use of the three skills in question.
Key findings
A comparison of the PIAAC results with the results of a similar international test from the 1990s shows that workers at all different levels of proficiency are more likely to use their literacy skills at work than was the case two decades ago. However, despite this increased frequency of use, there is now a smaller proportion of the workforce with a high proficiency in literacy than in the 1990s in most OECD countries. As a result, the increased use of literacy skills primarily reflects increased use by workers with low or moderate levels of proficiency. These findings contrast with many analyses in economics that use wages to measure skill and conclude that more workers now work with high skills because more now have jobs with high wages. That interpretation of the economics findings appears to be incorrect, at least with respect to literacy skills.
To understand potential changes in the demand for the PIAAC skills in the future, the project worked with a group of experts to assess current computer capabilities using the questions from the test. The goal was to identify what questions could be answered by current computer techniques and then to compare that computer performance with the performance of adults with different proficiency levels. The exercise focused particularly on computer techniques that have been demonstrated in the research literature but not broadly applied in the workplace.
The expert assessment showed that 62% of workers in OECD countries use the PIAAC skills on a daily basis at work but with proficiency at a level that computers are close to reproducing. Only 13% of workers now use the PIAAC skills on a daily basis with higher proficiency than computers. The other 25% of workers do not use the PIAAC skills on a daily basis at work.
Interpretation of the results
It is important to note that this exploratory project analyses only one set of work skills and does not provide a complete basis for forecasting how computers will affect employment and skill demand. Different mixes of skills are needed for different work tasks: for some tasks the PIAAC skills will be of primary importance, for some tasks they will be peripheral, and for some tasks they will be required in combination with other skills, such as common sense, expert reasoning, vision, physical movement, or social interaction. A comprehensive programme to understand how computers will affect employment would need to assess these other skills as well. Though the current study is limited to three skills, the approach it develops could be extended to assess computer capabilities across the full set of human work skills.
In addition, the analysis provided here does not address issues related to the application or diffusion of the computer capabilities in question. Studies of technology adoption find that widespread application of a new production technology often takes one or more decades and sometimes never occurs. The current economy reflects the economic impact of computer techniques developed several decades ago, but not the capabilities demonstrated by recent research. The full implications of current computer capabilities for reproducing the PIAAC skills will probably not be seen in the economy for several decades.
Conclusions
Despite its limitations, this study can provide key preliminary conclusions about the implications of future skill demand for education. Over the coming decades, it is likely that there will be strong economic pressure to apply the computer capabilities for the PIAAC skills across the economy. This is likely to reverse the pattern of the recent past of increasing proportions of workers using low and mid-level literacy skills. Without knowing where new applications will be successful, it is reasonable to conclude that there will be an overall decrease in demand for those workers – the vast majority – whose proficiency in the PIAAC skills is no better than that of current computer capabilities. This does not mean that these workers will become unemployed, but they will become less valuable for many work tasks, and that will reduce employment in some cases and reduce wages in others.
A standard policy response might advise increased levels of education so that workers are able to move into new types work. However, this study suggests such a response may not be viable, at least with respect to the PIAAC skillset. This is because there are no examples of education systems that prepare the vast majority of adults to perform better in the three PIAAC skill areas than the level that computers are close to reproducing. Although some education systems do better than others, those differences are not large enough to help most of the population overtake computers with respect to the PIAAC skills.
Ultimately, it is likely that the employment prospects for most adults one or two decades from now will increasingly depend on other types of skills that are not measured by PIAAC. To figure out what policy responses will be helpful in the years ahead, we need to assess computer capabilities across all skills used at work, not just those assessed in this study.