[DRAFT] Ubiquitous voting and data-mining, PoW recommendation agents

[ganzuul]

Tech

Non-Negative Matrix Factorization, AKA. The Netflix Prize algorithm, can give users recommendations for new polls to vote on based on what they have voted in the past. It does this by clustering voters in a space with a hundred or more dimensions. It is NP-hard, so it can be used as proof-of-work rate limiting, making manipulation with bots reasonably difficult.

Sociology

The user would gain an app which gives them recommendations such as URLs to visit. E.g. a researcher would get recommendations for publications to read. When voting, there would be a delay for PoW as the network is mined for the next set of recommendations. If the user does not like the recommendations they see, they can rescind their vote, not submitting it to the network, and perhaps better inform themselves on what they are voting on. Since the voter is the first to see the mined data, they are better in control of their privacy.

Security

To submit a vote to the network, the user needs PoW of their own recommendation and PoW of someone else's recommendation. This way a bot can not simply ignore recommendations altogether.

Science

A vast trove of perhaps pre-mined and therefore accessible data would be collected which could be of great interest to researchers of sociology. Definite personality groups could emerge out of the data.

Notes

This proposal emerged after discussion with @RoboticMind about #216