By Stijn Broecke.
The concern that robots will take our jobs is one that has already been discussed several times on this blog. It is also the topic of a heated policy debate and ranked high amongst the issues discussed at the recent Policy Forum on the Future of Work, hosted by the OECD on 14 January 2016.
While claims that nearly half of jobs are at risk of computerisation (Frey and Osbourne, 2013) are likely exaggerated, it is undeniable that technological progress is having (and will have for the foreseeable future) a profound impact on work organisation and the task-content of jobs.
Undoubtedly, certain jobs will disappear. But technological change also offers opportunities for productivity gains and the emergence of new types of occupations. The overall impact on employment is therefore difficult to predict.
The effect of technological change is also likely to vary by industry and occupation, and global estimates of the number of jobs that will disappear are bound to hide much heterogeneity. It is worth, therefore, to home in on specific occupations, and have a much closer look at the tasks that workers perform and analyse the extent to which those can truly be performed by robots.
This is exactly what Dana Remus and Frank Levy do in a recent working paper, where they focus on the law profession – an occupation where it has frequently been argued that automation will soon replace much of the work currently performed by lawyers.
What is particularly interesting about focusing on lawyers is that this is not your standard blue-collar profession with high routine task content. Instead, we are talking about highly skilled workers who often require years of training and perform jobs with a high level of social interaction.
Yet several pieces of software already claim to be able to perform tasks traditionally carried out by lawyers. For example, Ross Intelligence, a legal application of IBM’s Watson, provides concise answers to natural language legal questions, and online legal service providers (like LegalZoom and RocketLawyer) produce basic wills, divorce agreements, contracts and incorporation papers without any lawyer’s involvement.
However, while several people have argued that cheap software will replace “armies of expensive lawyers” Remus and Levy take a critical look at these arguments and identify three key weaknesses in the debate to date:
- First, this debate is frequently held in very general terms, and there is a failure to engage with technical details to appreciate the capacities and limits of existing and emerging software. The authors review a wide range of tasks carried out by lawyers and argue that computer technology may be successful in addressing some relatively structured and repetitive tasks (e.g. aspects of document review and management), but also that it will be far less effective in areas where interpersonal communication and interaction are required (such as advising clients, negotiations, and court appearances).
- Second, even if some tasks performed by lawyers could be automated, there is a lack of data on how much time lawyers currently spend on such tasks. This, in turn, makes it very difficult to obtain robust estimates of the share of lawyers’ work that could be automated. Using new data on time allocation in large law firms, Remus and Levy find that tasks that can be automated actually represent a relatively modest percentage of lawyers’ billable hours. The authors therefore conclude that the impact of automation on employment among lawyers is unlikely to be as drastic as headlines frequently suggest – although it could still result in a 13% fall in lawyer employment.
- Third, the authors argue that the debate should move beyond a focus on the number of jobs and look at how computers perform various lawyering tasks differently than humans. Sometimes, these differences will be beneficial (e.g. by eliminating human error and increasing accuracy), but sometimes they will be detrimental (e.g. by threatening creativity and flexibility in the law). Importantly, these differences will inform the desirability of automating various aspects of legal practice.
In their conclusion, the authors reiterate that technological advances are changing the structure and composition of the legal profession, but that much remains unknown about the nature and extent of their impact. They therefore argue that much more research is needed and, importantly, that this research should be grounded in technical expertise and empirical data, rather than headlines with definitive but unsupported claims. This is important advice that should be heeded by all researchers working on the future of work.