Universities play an indispensable role in developing artificial intelligence talent, but mounting evidence suggests that computer science departments across the United States do not have enough faculty to meet the growing demand for AI education. The goal of this paper is to help policymakers better assess the scale, causes, and consequences of these “teaching capacity gaps” in AI, and to present policy levers that could help increase the capacity of universities to train the next generation of AI specialists.
While it is difficult to measure the potential mismatch between the supply of instructors and the demand for AI education, available evidence suggests there is indeed a gap. Over the last decade, the increase in computer science enrollments has far outpaced the growth in computer science faculty, who are responsible for much of the AI instruction at U.S. universities. Universities have started restricting access to CS courses, and academic leaders have publicly lamented the difficulty of recruiting and retaining qualified AI faculty.
These teaching capacity gaps can have detrimental effects on students’ educational experience, change the quantity and trajectory of academic research, and hamper the country’s efforts to build a robust AI workforce. Despite its wide-ranging repercussions, the lack of AI-teaching capacity at U.S. universities has received relatively little attention from policymakers and analysts. While we cannot say for certain that teaching capacity gaps exist, we consider this evidence to be sufficiently strong to warrant further investigation and policy attention.
When teaching capacity gaps are discussed, policymakers often attribute them to a shortage of AI faculty caused by a combination of tech companies “poaching” professors and recent PhD graduates losing interest in academic careers. This argument is based on the implicit notion that AI experts are choosing to forgo academic jobs for better-paying, less-restrictive careers in industry.
While these experts may have correctly identified teaching capacity gaps at U.S. universities, available data suggests they may have misattributed the root cause. Little evidence suggests the outflow of AI experts from academia to industry has distorted the job market, or even that industry hiring is uniformly negative. Furthermore, while a greater share of PhD graduates is indeed flocking to industry, the available data shows that many are still interested in pursuing academic careers, and the total number of PhD graduates who enter academia each year has remained relatively constant. However, we did find evidence that universities have not created enough new faculty positions to accommodate students’ growing interest in the field. These findings suggest that universities would be able to close their teaching capacity gaps if they created more faculty positions, though budgetary constraints may limit their ability to do so.
Understanding the specific factors creating teaching capacity gaps is critical, as different causes demand different policy solutions. In pursuing such measures, federal agencies must consider whether their interventions are intended to increase research capacity or teaching capacity. While both research and teaching are integral to universities’ missions, they can at times be at odds with one another. If targeted incorrectly, policies meant to increase teaching capacity can in reality exacerbate the problem, and potentially create AI faculty shortages. Still, there are a variety of measures policymakers can explore to close teaching capacity gaps, such as creating federal grants to facilitate faculty hiring, incentivizing industry to support university education, and expanding access to government data and computing resources.