The 2020 spending review included £2.9bn over three years for the Restart scheme, which promises “intensive and tailored support” to more than one million unemployed people. With further details still to be announced, it is crucial that thought is given to the significant challenges involved in providing careers support at such a large and unprecedented scale.
With many workers facing job losses as a result of the triple threat of Covid-19, automation and Brexit, the demand for careers advice is greater than ever. The success of the Restart scheme will depend, in large, on the quality of this advice. Job and training recommendations need to be based on data that reflects the range of shocks impacting the labour market today.
Advice must also be tailored to the needs and skills of the individual. The consequences of poorly tailored advice were seen in the recent resurfacing of a government-backed advertising campaign (from 2019) that encourages a ballerina (named ‘Fatima’) to give up her job and retrain in cyber security. The fiasco demonstrated how poor careers advice can linger and, in turn, undermine public confidence.
At Nesta, we have been exploring how machine learning might help to inform careers advice. By using thousands of descriptions about the skills and work activities required in different positions, we have created a map that shows the similarities between jobs. This map allows us to identify roles that are similar to a worker’s most recent position, while taking into account risks such as exposure to Covid and the likelihood of automation. It can also be used to spot the new skills that a worker may need to move into a different role, as well as suggest skills that are likely to increase the transferability of a worker’s skillset.
The scale of advice needed for the Restart scheme means that online tools (driven by algorithms) will be increasingly needed alongside other forms of delivery. The effectiveness of these tools depends critically on how they are built and used. Our own experience has led us to recommend four essential features:
- Focused on the long term. There is little point in retraining workers for jobs that are likely to be automated in a few years time. The current crisis provides a unique opportunity to help workers transition into lower-risk jobs, and a recommendation system would ideally take automation risk into consideration.
- Viewed as complementary to existing tools. Recommendation tools should be seen as an additional source of information for careers advisers and jobseekers. They cannot replace the role of advisers, as recommendation tools will always struggle to capture the unique needs and preferences of individuals.
- Open to scrutiny. The consequences of poor career recommendations are high and include prolonging spells of unemployment. The underlying algorithm should be open to scrutiny and it should be possible for a jobseeker to explore why a particular job or training course has been recommended to them.
- Used to broaden options. Recommendation systems for job transitions and training are still in their infancy. As such, they are best used as a way to provide new suggestions to jobseekers, which individuals can then dismiss or explore. These tools should not be used to define or constrain a jobseeker’s options.
Cath Sleeman is head of data visualisation, creative economy and data analytics at Nesta