What Are the Odds That Teaching Will Be Automated in the Very Near Term?

Recent months have brought a great wave of news stories about a shortage of teachers approaching crisis levels--and the possibility that even if such a shortage is not already underway (a difficult thing to establish one way or the other given the scarcity of really comprehensive educational statistics) it may be imminent as exhausted instructors leave the profession much more quickly than anticipated, new entrants are deterred from joining the profession in the expected numbers by the conditions of the job, or the combination of the two widening the gap between need and supply.

One question I have found myself wondering about, given the talk we have been hearing of automation, has been the expectations regarding the automation of teaching specifically. Not long ago I considered Ray Kurzweil's thoughts about the matter at the turn of the century--which, as with many of his predictions in the relevant areas, were premised on forecasts of advance in particular technological areas that have since appeared overoptimistic (notably the speed at which pattern-recognizing neural nets and all premised on them would develop) and a naiveté regarding the social dimensions of the subjects about which he wrote (in this case, the school's function as "babysitter").

However, not everyone has been so optimistic--even those who have, by any reasonable measure, been optimists about automation. Exemplary is the study Carl Benedikt Frey and Michael Osborne produced back in 2013, which played so important a part in the conversation about automation and employment in the '10s. That study included in its appendix a table listing over 700 occupations and the chances of their being "computerized"--"potentially automatable over some unspecified number of years, perhaps a decade or two."

The authors determined that the jobs of data entry keyers, telemarketers and new accounts clerks had a 99 percent chance of being "computerizable." Contrary to what might be expected by those who make much of "high-knowledge" occupations, Frey and Osborne even anticipated fairly high odds of a great deal of scientific work becoming automated (with atmospheric and space scientists having a 67 percent chance of having their jobs automated), with, in spite of what may be thought from the popularity of the sneer "Learn to code," a near-even chance of the same happening with computer programming (48 percent). But, teaching assistants apart, they put the odds of computerizing any teaching occupation at not much better than 1 in 4 (a 27 percent chance of middle school technical teachers), while the odds of computerizing postsecondary school (college) teaching they put at 3 percent, the odds of computerizing preschool, elementary and secondary school teaching at under 1 percent.

In short, far from being easy to automate, their analysis suggests that teaching will, to go by their assessment of the potential for computerizing the task, be exceptionally difficult to automate satisfactorily. The result is that even if a great wave of automation swept through the rest of the economy—for what it is worth, Frey and Osborne calculated that nearly half of U.S. jobs were, in the absence of significant political or economic obstacles (legal barriers, particularly poor investment conditions, etc.), at "high" (70 percent-plus) risk of such computerization by the early 2030s--automation would have little impact on a great many teaching jobs. The result is that one can easily picture a situation in which job-seekers would find themselves with fewer alternatives to teaching--meaning relatively more people pursuing such positions, not less (at a time in which an aging population structure would likely mean fewer students, and fewer job openings for that reason). In the nearer term, in the absence of any such pressure sending people toward the occupation, it seems additional reason to think automation unlikely to be a solution to the problem.