article 1

The Economic and Technical Feasibility of AI Substitution of Harvesting Jobs in the United States During COVID-19 4 difficult to visually determine the location and suitability of the vegetable. An AI harvester must be able to select from among individual lettuce heads whose appearance is highly dependent onweather conditions, maturity, and surrounding vegetation. Two, in an area with uneven soil, the lettuce stalk must be carefully trimmed at a specific height in accordance with commercial standards. And three, lettuce stalks can be easily damaged by rough handling. A salad-quality picking solution must include “a high-precision, high force cutting mechanism” to ensure gentle handling of the vegetable (Birrell et al., 229). Finding a solution to improve the crop tracking accuracy is an issue still to be solved. Technological bottlenecks like these make it very difficult to implement robots in agriculture at present from a technical standpoint. Therefore, from a shop-floor perspective, it is reasonable to state that robotic substitution of labor will only be technically feasible for a few specific types of crops. Such technical limitations help to explain why not all industries have so far engaged in robotic substitution to cut costs during COVID-19. An interviewwith the CEO of Harvest CROO, Dr. JosephMc- Gee, further helps to gauge the accuracy of conclusions made about the feasibility of automation. On the issue of COVID-19 and agriculture, the result was slightly different than was ex- pected. Even though theO*NET “risk of infection” projection was low for harvesters, Dr. McGee stated that since much of agriculture is based on cross-country travel, many workers could not get to their workstations without risk of infection, leading to a great decrease in the number of harvesters will- ing to work or show up at the jobsite, which at least partially explains why some industries have engaged in buying robots. Additionally, H-2A visa application rejections have accelerat- ed the demand for harvesting robots during COVID-19 peri- od to meet productivity and labor demands in the harvesting sector. Similarly, with COVID-19 accelerating this trend, an aging population is already leading to a “chronic shortage of labor” in young persons with the ability to be farm workers. Furthermore, Dr. McGee mentioned multiple reasons why robots are more capable of maximizing revenue than hu- man laborers, including the fact that they can measure the weight of the crops and pack/store them according to speci- fications that are usually not met, leading to loss of revenue. He stated that these factors are leading to a direct increase in demand for agricultural automation. He elaborated on this fact by stating that “70% of the US strawberry business has invested in Harvest CROO” and is currently waiting for the rollout of the final product. So why is this solution not being immediately implemented?The fact that thefinal producthas not been perfected yet is, as Dr. McGee stated, a function of technical bottlenecks that would take “hours to present” including visual sensors, as mentioned above. However, he expects that these can be dealt with in a few months to a year ahead. In summary, Dr. McGee believes that although technological bottlenecks stand in the way of complete au- tomation in the short run, the aging worker population and accelerated effect of COVID-19 on the current carrying-out of agricultural business will motivate exciting new innovations that have the potential to completely substitute agricultural labor, including harvesters. This prediction, even if not 100% accurate, leads to questions policy-makers should consider regarding the rights of these workers (McGee). Policy Options At the federal and state levels, similar to other occupations in the United States, there should be worker protection laws and regulations to ensure the involvement of harvest workers in decision making related to digitized automation going forward. For example, when subsidizing research and development of technologies that could possibly yield robotic augmentation of, or substitution for, human labor, harvest workers should be engaged in the process. As a way to help offset or smooth the expected transition for harvest workers, companies that work on labor augmentation technologies could be incentivized by means of tax breaks to offer robot maintenance jobs to migrant workers. Beyond this, there could be corporate tax breaks for companies that reskill displaced harvesters regardless of immigration status. Establishing a joint program with other governments that supply seasonal workers to the US to reskill harvest workers with short-term training to enable them to operate augmented/semiautomatedharvestingequipmentwouldalso smooth the technology adoption process. Finally, taxation of full automation harvesting systems to support government reskilling programs and facilitate H-2A visa programs in the future would be beneficial. At the corporate employer level, investing more into the educational infrastructure of communities that rely on agricultural employmentwouldbehelpful, botheconomically and socially. In this regard, employers should further the education of their workers so they can upskill themselves in order to work with robotic machinery in a complementary manner as opposed to being replaced by the technology. Offering grants to local and regional higher education institutions specifically for research groups that are working