From Metrics and Sentiment to Scores and Data as Labor

From Metrics and Sentiment to Scores

My previous post was Tau specific. This post pursues the topic from a more technologically agnostic position. The current paradigm of data mining is problematic. The data which is collected is often in databases in an insecure manner. On blockchains all the data collected is public. The big data is inaccessible, hard to secure, and it is very difficult to maintain privacy even in a world of transparency.

The technology which is game changing is the Trusted Execution Environment. This technology is not flawless but it does introduce a paradigm shift. Using TEE it is now possible to compute data from within a hardware encrypted environment. The nodes doing the computation cannot interfere with or see what the data is. This would allow for the best benefits of both transparency in the sense that no one can see your data but also the benefits of transparency in that you can have a reputation score which you earned, you can have a loyalty score which you earned, in accordance to the data.

Data as the new Labor

Proof of Work introduces the philosophical concept that "work" is represented not in "human labor" but in the physical sense of the word. The more electricity and computation you spend the more "work" you produce and the greater your probability of being rewarded. This presents an economy based around mining and produces behaviors which in my opinion at least for now are inefficient (lots of expensive mining rigs, centralization at scale, etc).

The new "work" in a data market is data. The new "mining" will be providing useful data. In order for this new economic paradigm to be possible it is necessary for the data to be under the full and complete control of the owner. The Trusted Execution Environment allows for concepts such as the Data Wallet. This Data Wallet allows for digital "self sovereignty" which is an important concept to remember going forward.

The Self Sovereign Digital World

The Self Sovereign Digital World is a world where you have control over your data. You will be able to share your data to the blockchain without your data being accessible to any human being. The data in other words would remain private even while you share it and even while decentralized apps (and algorithms) make use of it. In other words you will be able to have scores produced by algorithms by leveraging the Trusted Execution Environment. A simplified overview on how this might work:

• Nodes (similar to miners or witnesses) are provided with a token incentive to compute encrypted data. These nodes must be specifically Trusted Execution Environment hardware and must run the necessary software to do secure multiparty computation (shared computation). The input is sent from the client to these nodes encrypted (undecipherable by any human or machine) until it reaches the Trusted Execution Environment of the node where it's computed. The output of this computation is then sent out encrypted to the client.

• The Trusted Execution Environment can handle any kind of computation or smart contract which means you can use any design pattern such as for example the Token Curated Registry. This design pattern would allow nodes to function like voting booths where the votes are counted in secret. In other words the Trusted Execution Environment presents us with a black box which no one (not even the owner of the box) can see into. Votes can be in the form of collaborative filtering (up and down) or in the form of reviews (points). Votes can be stake weighted or moderator style or anything else.

• The data (specifically big data) allows for predictive analytics, and allows for the training of machine learning algorithms. This immediately gives a price to the data because the data is the only way to make the algorithms useful. For machine learning the more examples it has to work with the more it learns. It's a process of statistics basically.

• Metrics and sentiment data are some of the most sensitive data. People do not want to share their true thoughts and feelings due to social desirability bias. If people could be sure that their inputs remain encrypted then they can give their true sentiment on surveys even if their true sentiment would if done on an unencrypted blockchain such as Steem cost them socially. In other words it removes the problem of social desirability bias in sentiment analysis if there is true privacy of the data. The Metrics such as account age, engagement level, and various other useful metrics a community decides to use for their algorithms or machine learning, can be kept confidential or made public. In other words you can create loyalty scores which cannot be gamed because the process is a proprietary algorithm (due to the TEE).

• You can have social rank, and this social ranking can be by scores. To provide a pro-social example we could reward creatives who introduce new ideas such as a new word which becomes popular. The first to use the word would be on the blockchain and if the usage of the word spreads then a reward pool could split the rewards according to an algorithm. A more creative example is a secret keyword lottery. It is possible to create a bot which scans the entire Internet for certain keywords and which rewards daily the users of these "secret keywords" which would create a word lottery. The secret word list could be kept encrypted via the TEE and the creators of the list could use a random number generator to sort the list and as a result produce the formula for reward distribution.