Thread by @balajis

A fundamental issue is that viral content tends to be provocative. In theory it's possible to change the incentive structure, as shown by these two graphs.
facebook.com/notes/mark-zuc…

Right now, the "policy line" is determined by community reaction or manual platform review. So there are humans in the loop.

However, if FB or Twitter's databases were open source, anyone could try ML to try predicting virality & sentiment for a post. Probably some signal there.

Of course, the problem with ML-based pre-publication content review by FB/Twitter is that these are for-profit American companies put in the role of soft censoring millions.

I think many people there have good intent, but as folks have noted the potential for abuse is obvious.

An alternative approach that is becoming more feasible is:

- open state backend, like a blockchain
- community-imposed content standards
- transparent algorithmic enforcement
- pseudonymous crypto governance

A crypto social network for each community.

In this model, all community members would consciously opt in to a crypto social network with a transparent backend, governance model, and social contract.

Content standards would be arrived upon by a community, with enforcement by transparent algorithms.

Importantly, because these algorithms would run upon a blockchain (or similar decentralized datastore), they would not just be open source. They'd be open state and open execution.

A glass box where you can see every database entry & replay every opcode.

https://twitter.com/balajis/status/1065462749281435648