Q: Data is not just a product of “labor” but also of the tech companies’ investments in research and infrastructure. Why do its returns all belong to individuals?
A: Like anything else, data is a mixed product of capital, labor and natural endowments. However, it seems plausible based on the role of human actions in producing most useful data that the productive share of labor in creating data is very high, while at present the share it is being compensated for is extremely low. So “data as labor” is a simplification of “data as mostly labor”, which is more accurate. We doubt that, given our society, it is likely to be true any time soon that excessively high rents are allocated to individual data producers rather than the owners of capital, so we are not too worried about this simplification.
Q: Isn’t the amount that any individual would be paid for their data at present tiny?
A: Calculations by experts of the value of data for a typical person reading this book at present range from $100 to1000 a year. This value would grow if people were compensated as AI would be more productive, but AI is also growing generally and a huge part of the value of the tech companies is based on this expectation. If this proves even modestly right (say if 10% of the economy becomes AI-dominated), then shifting from data as capital to data as labor could generate tens of thousands of dollars for middle-income U.S. households.
Q: It seems very hard to organize a digital union given that people are not in the same physical location and are so heterogeneous. How do you address this?
A: While digital ties may be thinner, social networks allow for some quite potent connections and for much greater observability than was possible in the era of labor unions. Furthermore, digital technology can directly facilitate strikes through VPNs that can block data access. Algorithms are also more easily auditable than standard production processes, making it easier to tie down data value. Combining these different forces, we think data labor unions may be highly effective.
Q: I don’t want people spying on and commodifying my interaction with my friends at any price. Aren’t you just promoting this?
A: If you strongly value privacy, you can choose not to do data work; the union would have the responsibility to enforce this for you. But you would also have the opportunity to work if you want to. Currently your data is used without your knowledge. We want transparency and fair market values, and then you can decide what you want to do.
Q: Doesn’t this dramatically increase cognitive load? Everyone wants to just enjoy digital services and not worry about the economics.
A: Data labor unions would handle much of this cognitive load for workers. They would monitor terms and conditions, ensure data is not being abused, etc. They would also target work opportunities to workers based on knowledge of their preferences without them having to do detailed calculations to figure out if it was worth their time. At the same time, workers would be aware generally of their role and be able to adjust parameters when they chose to weigh in.
Q: Hasn’t the labor movement failed? Why would it succeed here?
A: The labor movement was successful during the nineteenth and early twentieth century. It started to fail as the labor market evolved with technology and growing diversity of the work force. A union based on data could take advantage of technological developments and the special nature of date work.
Q: People get so many great free services from digital platforms. Isn’t that sufficient compensation?
A: The question is less “how much benefits are we getting” than are the benefits we getting directly rewarding productive work in a way that encourages it. At present we get all sorts of fun services and the platforms get all the marketable value of our work. This gives no incentives for developing productive abilities or providing valuable, high-quality data. It also gives no freedom to individuals to spend the value of their work in places other than the platforms. It is a sort of techno-feudalism, which is unproductive and inequitable.
Q: Won’t making data even more proprietary and controlled stunt innovation? Shouldn’t we make data open and free?
A: Data labor unions will very likely have an interest in making data available to a variety of services on relatively non-discriminatory terms and thus would likely open up the data economy. However, the high-quality data we need for a variety of applications won’t be produced if we don’t reward and aren’t open with those who individually contribute to its production. So the best solution is for data to be reasonably compensated, protected and bargained for, rather than just made free.
Q: Why do you call it “labor” rather than just data ownership?
A: People do work when they type into the computer. Indeed, people are frequently hired to do this.
Q: Won’t the fruits of data labor be distributed unequally to those who it is profitable to advertise to?
A: Labor income of all kinds is unequally distributed, but capital income is more unequally distributed. If we eliminated the inequality in the distribution of capital, it would solve a majority of the inequality in wealthy countries. So, yes, moving towards data as labor would not solve all inequality, but it would make things less unequal than at present.