At a glance
- Source
- Brynjolfsson, Mitchell and Rock (2018). What Can Machines Learn and What Does It Mean for Occupations and the Economy?. AEA Papers and Proceedings, 108, 43–47. doi:10.1257/pandp.20181019
- How it was done
- The authors scored each task in O*NET, the United States government’s database of jobs and tasks, on how well machine learning could do it. Data analysis. It builds a measure from existing data, such as job and task records.
- Who took part
- 18,156 tasks across the occupations in the database.
- Where
- United States job and task data.
“Realizing the potential of ML usually requires redesign of job task content.”
What it found
- Machine learning affects different occupations than earlier waves of automation did.
- Most occupations include at least some tasks suited to machine learning.
- Few occupations can be fully automated with it.
- Getting the benefit usually requires redesigning the tasks that make up a job.
When the result holds
The scores describe what the technology could do with a task. They say nothing about cost, data, regulation or whether customers would accept it.
Limits
- The scoring dates from 2018, before today’s AI chat tools.
- The paper appeared in a conference volume that is reviewed more lightly than a standard journal article.
Practices that use this study
This study supports:
- Map tasks before deciding what to automate Break jobs into tasks and look at each one. Published scores rate how much of a job AI could do, its “exposure”. Treat a high score as a place to look, not as a decision to automate. Do first · Effort: low (days, one person)
- Redesign the workflow instead of adding a tool to it Giving people an AI tool inside an unchanged process changes little. Decide again who does each step, in what order, and where a person checks. Do next · Effort: high (months, several teams)