The Social Credit System is a data-driven reputation system which draws on several sources to label various entities, namely businesses and individual citizens, with a trustworthiness score. One can only guess that Facebook implements similar systems, albeit for different goals. What these systems have in common is that they're based on an ontology, i.e., a finite set of entity-relations, which fix the constraints on the correlations between the various entities to be scored. One could think of similar ontologies in, say, cybersecurity, whereby various entities (domains, IP addresses, e-mails, etc.) can be seeded with beliefs or initial scores of maliciousness which will then propagate to their neighbors according to particular rules. The same could be done with fault prevention or forensics in mechanical systems of interacting entities.


Conceptually speaking, is there a common framework that best represents these systems? I initially thought of Bayesian propagation, but it seems that it doesn't easily account for

  • uncertainties in the scores,
  • loops in the ontology graphs (how does one avoid runaway "feedback"?),
  • non-linearities at the nodes (since scores can be generated at a node itself based on its attributes, regardless of its neighbors). For example: Bob is a successful surgeon, Bill is a drug addict. Therefore, Bob is less likely than Bill to commit a crime---and that's independently of any inference from their respective environments.
  • 3
    $\begingroup$ This isn't really an answerable question because nobody who knows is going to be telling. PageRank, Bayesian Networks, and block-chain immediately come to mind as possible (sub)systems. $\endgroup$ – Stella Biderman Jun 27 '19 at 15:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.