Some “social science fiction” from Jean Tirole!
We are at a point, argues Jean Tirole, where social platforms, artificial intelligence and the data they collect will either lead us to a society where law compliance will be the norm, or to a society where governments would hold too much information on individual citizens through mass surveillance. Given the grave perils posed by the second scenario, it is important to understand the channels through which a dystopic society might come about, so as to better design legal and constitutional safeguards.
Tirole is using as an example China’s recent introduction of “social score” received by every citizen and which will contain information such as credit history, tax compliance, good deeds, environmentally friendly behavior, traffic violations, fraudulent behavior in markets, the spreading of “fake news” and the posting of “inappropriate posts” (whatever this may be interpreted as meaning), the individual’s social graph, personal traits, political, or religious opinions, etc. The score will be publicly available. He then asks how can this social score be used by autocratic regimes, democratic majorities, private platforms, and religious or professional organizations to achieve social control by managing the flow of information about individuals’ behavior:
- Bundling prosocial activities and compliance with the government’s objective into a single score exploits agents’ interest in each other’s score to promote political compliance.
- The government builds such dual-purpose social scores if and only if it is sufficiently autocratic.
- Its ability to enforce compliance with its objective through bundling is higher in a weak-ties society than in a strong-ties society. The intuition behind this result is that in a society with strong ties, agents have information about each other that unbundles the information supplied by the government.
- The government must eliminate competition from independent, privately provided social ratings. Whenever economic agents are more interested in the social reliability of their partners than in their political, religious, or societal views, private platforms would expunge any information about these views from their ratings, de facto unbundling the information.
- Transparency/citizens’ awareness about the way the social score is computed (together with opaqueness about its components) is key to the latter’s effectiveness.
Digital Dystopia
Author: Jean Tirole
From: Toulouse School of Economics
Political extremism through social platforms
Acemoglu and al. present a model of online content sharing where agents sequentially observe an article and must decide whether to share it with others. They find that:
- When there is little homophily in the “sharing network”, misinformation is often quickly identified and brought to an end;
- But when homophily is strong, so that agents anticipate that only those with similar beliefs will view the article, misinformation spreads more rapidly because of echo chambers. This viral behavior is exacerbated when the article has an extreme message or polarization is strong;
- Social media platforms that wish to maximize content engagement will propagate extreme articles amongst the most extremist users, while not showing these articles to ideologically opposed users. This creates an endogenous echo chamber—filter bubble—that makes misinformation spread virally.
Misinformation: Strategic Sharing, Homophily, and Endogenous Echo Chambers
Authors: Daron Acemoglu, Asuman Ozdaglar, James Siderius
From: MIT
