Should We Treat Data as Labor? Moving Beyond "Free"

Should We Treat Data as Labor? Moving Beyond Free E. Glen Weyl Microsoft and Princeton Competition Conference 2018 Boston University School of Law July 23, 2018 Imanol Arrieta Ibarra Diego Jimnez Hernndez Stanford Stanford

Leonard Goff Columbia Jaron Lanier Microsoft Free Services for Free Data The digital economy encompasses many of the most popular products ever. It creates surplus for end users and tech companies. Data

Services 3 The High Cost of Free Data Despite these advances, there has been backlash against tech companies. 4 An Anxious Time For Labor

Fears about future of AI and job displacement Growing inequality (Autor et al. 2017) AI and Productivity The free-data model has failed to induce the AIrevolution in productivity. 6 The High Cost of Free Data

The High Cost of Free Data Root of all of these problems is that returns to data are being treated as payments to capital rather than to labor. Data as Capital 1. Data is exhaust to be picked up by whoever is smart enough to make useful 2. AI success mostly driven by algorithms, computational power, brilliant programmers 3.

AI will flourish if there are big payoffs to entrepreneurs and innovators 4. AI will displace workers, so either they will work in other areas or we will give them a basic income now that they are useless 5. People will have to find meaning in areas of life other than work and get over themselves 6.

Individuals are happy to consent to surveillance in exchange for free stuff, so companies deserve to use their data vs. 1. Data as Labor Data created by user actions and efforts, so fruits belong first to individual contributors

2. AI is really collective human intelligence, driven by innate knowledge/data in ordinary people 3. AI will flourish if the individual contributors who make it possible have a real economic stake 4. AI is just another production technology that raises individual productivity by allowing people to amplify the value of their data

5. The value of data will support sense of meaning and digital dignity 6. Individuals on their own have little bargaining power in face of digital monopolies, must be protected by countervailing power (competition, unions or governments) How Did We Get Here?

Current status quo comes from a mix of: Prejudice (weight of precedent created by historical accidents) Privilege (entrenched interests that derive rents from the inefficient equilibrium) The internet business-model: convince users to give their data for free in exchange of a digital service. Users grew accustomed to this model. Few users are even aware of the productive value of their data or the role they play in enabling ML to target advertise them.

Recent evidence suggests significant monopsony power in online task labor markets (Dube et al (2018)). 10 Sources of Countervailing Power There are three main channels to balance out the scale: Competition

Data Labor Union Government Action 11 Recent policy developments GDPR came into effect First Data Labor Union in Holland Merkel considers data tax

12 13