Automation, work, and jobs

The effect of automation on employment and jobs has engaged thoughtful computer scientists and economists since the earliest days of computing. Yet there have been concerns about the effects of technology on employment since ancient times, and notably during the First Industrial Revolution in the nineteenth century by a group of workers known as the ‘Luddites’. Our first topic is the role of algorithms in enabling more efficient processing of job applicants and the selection of candidates to interview. This now includes the automatic filtering out of huge numbers of résumés that are never seen by human resource professionals. Next, we look at how technology is used in monitoring job performance, with the goal of encouraging or requiring enhanced performance. Oftentimes, these practices have the opposite effect, as it makes workers feel like ‘Big Brother’ is watching. Companies have long used contractors to provide flexibility in the availability of workers as well as to circumvent costs such as medical benefits and liabilities such as severance pay. This practice has recently changed dramatically: internet communication can now rapidly link seekers of services to providers of the services. This is typically called the gig economy or sharing economy, yet a better name is on-demand services. We shall then examine areas where automation threatens to replace human workers with machines. Fear is rampant, as typified by a 2017 New York Times article, ‘Will Robots Take Our Children’s Jobs?’ Between 2014 and 2016, future prospects were analysed in five scholarly books. We examine the phenomenon of unemployment by looking at specific areas: agriculture, manufacturing, service industries, and the professions. We highlight how new robotic technology, incorporating sensing, reasoning, and manipulating abilities, is enabling significant automation. Of particular importance is the extent to which new machine learning systems are enabling the automation of thinking and reasoning, which were previously considered infeasible for machines. Arguably the most interesting, challenging, and risky application is that of automatic diagnosis of disease, and, more speculatively, robot doctors.