THE IMPACT OF ARTIFICIAL INTELLIGENCE ON WORK
THE IMPACT OF ARTIFICIAL
INTELLIGENCE ON WORK
An evidence review prepared for the Royal
Society and the British Academy
September 2018
Evidence on AI from theoretical economic models Recent theoretical work in economics aims to provide a framework to specifically understand the impact of AI on employment, including both its immediate, firstorder effects, and following, second-order effects. The main findings from this literature suggest the following:
These models suggest that automation can lift living standards for all, but this is not necessarily true in the short term, and even in the long term inequality may increase as some workers benefit more than others. This is a departure from earlier economic theory, where technological progress was typically seen to benefit all workers under a broad set of conditions. It is worth noting that the nature of these models is to consider second-order effects of automation driven by price changes (the ‘productivity effects’ and impacts on ‘consumer demand’ mentioned above), but not to consider possible social or institutional changes and the effect these may have, in turn, on economic outcomes.
Access to full report here.
Evidence on AI from theoretical economic models Recent theoretical work in economics aims to provide a framework to specifically understand the impact of AI on employment, including both its immediate, firstorder effects, and following, second-order effects. The main findings from this literature suggest the following:
- A number of factors can counterbalance initial declines in labour demand due
to automation. As automation increases productivity (leading to better or
cheaper products), increasing consumer demand, greater investment, and
innovation can lead labour demand to rise.
- In the short term, it is not clear whether countervailing effects will be sufficient
to offset potential job losses from automation. Even if they are, transitions could
be challenging. But new jobs could be generated, in principle, in the same
industries where automation is taking place.
- In the long term, as production processes are re-organised, countervailing
effects are expected to become stronger and fully compensate the initial
decline in work demanded by businesses that have adopted AI to automate
production. However, workers who have been directly displaced could
experience a fall in their earnings relative to other workers (and potentially in
absolute terms). This would increase inequality if the displaced group is mostly
composed of low earners.
- The competitiveness of product and labour markets is important to drive better outcomes for workers. The results above typically assume that product and labour markets are competitive, that is, consumers can choose between different products and workers can choose between different employers. This is not always the case, as acknowledged in some of this literature. In particular, lack of competition in labour markets would mean that productivity benefits from automation flow into greater profits rather than into higher wages. There is some concern that digital technology may enable large firms to increase and maintain their market power. There is on-going research on this but evidence is still limited.
These models suggest that automation can lift living standards for all, but this is not necessarily true in the short term, and even in the long term inequality may increase as some workers benefit more than others. This is a departure from earlier economic theory, where technological progress was typically seen to benefit all workers under a broad set of conditions. It is worth noting that the nature of these models is to consider second-order effects of automation driven by price changes (the ‘productivity effects’ and impacts on ‘consumer demand’ mentioned above), but not to consider possible social or institutional changes and the effect these may have, in turn, on economic outcomes.
Access to full report here.
