If you have a job (say, driving a truck, or working in a warehouse) and that job gets automated, what are your future employment options? A recent analytical paper on networks by Jordan D. Dworkin looks at this situation mathematically.
It brought to my mind studies of fitness landscapes in evolutionary biology – when conditions change what are the new niches that you could move into?
Dworkin notes that many studies of the risks of automation focus solely on the potential automatability of individual jobs. His paper looks not just at how likely particular jobs are to be automated, but the projected growth in other job fields and the similarity of skill sets between them.
Given your current job, and its associated skill set, what new types of work (that hopefully are less likely to de done my machines or algorithms) could you move into (with some retraining)? He calls this “job transition.”
This is very slightly equivalent to asking “If you are a nut eating finch on a Galapagos island, how could you use your nut cracking-focused skill set to find a viable niche in a new habitat on an adjacent island” (as opposed to, “I’d really like to be a marine iguana”).
In line with other analyses, Dworkin sees that, generally, prospects for “safer” (ie, less at risk of automation) jobs in the industrial and service sectors are less hopeful than those for the business, scientific and medical-related sectors (he uses the Standard Occupational Classification System), and that it may be harder for workers in industrial and service-related jobs to have skills that help them find work in the other sectors.
Dworkin’s findings, he himself admits, tend to agree with the less mathematically oriented thoughts of think tanks and other ‘future of work’ assessments. However, he provides greater methodological rigour, and highlights the importance of mobility of workers, not just automation of jobs.
“Importantly, understanding which jobs will have a harder time successfully completing job transitions is only half of the story. Once at-risk jobs are identified, it is crucial to determine which transitions will be the most beneficial for workers seeking to make a switch. At this moment in time, the common wisdom regarding job transitions appears to be that workers in automatable occupations should consider retraining for jobs in technology … or healthcare …” Dworkin
Some good skills to have, regardless of what you do, are:
“… skills that ranked highly [for low future automatability] still tended to be associated with problem-solving (e.g. systems evaluation, complex problem-solving) and communication (e.g. instructing, management of personnel resources).”
There are four key caveats to Dworkin’s analyses. The first is basing the likelihood of job automation on Frey & Osborne’s 2017 analysis – The future of employment: How susceptible are jobs to computerisation?
I’ve noted some of the limitations of their earlier work previously – just because a job could be automated doesn’t mean that it will be, and more often tasks rather than jobs are automated. (For example, a lot of farm work is automated now, but there are still farmers and farm workers). Frey’s also not comfortable with how some are interpreting their research, and that it is more of a historical analysis rather than a future prediction.
Dworkin acknowledges some of these uncertainties. He also notes that projections about growth (or decline) of job types may be inaccurate.
Thirdly, estimating skills is rather crude, a point also recognized by Dworkin. It also fails to take account of other skills workers have that aren’t reflected in their current jobs.
The fourth limitation is that the Dworkin’s model is static, rather than dynamic. As we have seen over the last few decades, jobs, and tasks change, and different skills become more or less important. That doesn’t seem likely to change in the future.
Models can be useful for examining complex systems. We just need to be careful not to uncritically accept them. Nor to dismiss them out of hand if you disagree with some assumptions.
Dworkin’s conclusion that it’s complex is likely correct:
“Overall, this study demonstrates that automation will likely have complex effects on the job market that are not fully captured by the likelihood of individual jobs becoming automated. Instead, it is important that policy-makers, businesses and workers consider the relationships between jobs when determining who is at the highest risk of long-term displacement and which transition or reskilling opportunities they should pursue.”
Much of the discussion about the future of work focuses on numbers of jobs, and types of work. Less often factored in (because they’re harder) are “quality” and stability of work, and how equitable are pay rates.
Like evolution it’s hard to predict what’s going to happen next work wise (even with fancy fitness landscapes or network models). It won’t just be the adaptability of workers that matters, but the conditions set by policies and work place practices, and their agility too as the landscape keeps changing.
Featured image: By John Gould – From “Voyage of the Beagle” as found on wikimedia