Testing................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
Requesters, Workers, and Police
- Each account can be any of these three.
- Each account has different types of reputations that are named based from HITs and other democratic or stake-based actions they take.
- Police may evolve into Judges in future versions.
- Delegation for investors is possible but a minimum balance is required.
Projects
- A minimum stake is required to be able to make projects.
- The higher the stake that the requester has, the more rewards he can give.
- The manner of distributing rewards per HIT could be:
o Based from Dawid and Skene Model (how correct were their answers compared to others’, relevant to reputation scores or weights), or
o Based from manual or fixed-automatic reward that the requester sets. - Reputation points given are proportional to the tokens transferred.
- Projects partner with guilds for recruitment.
- Top 5 (tentative) can be supported by stake-based voting. Proceeds will go to the requester. Projects are ranked for voting based from average reputation of recruited members. Top 20 guilds are displayed for voting. The purpose is to support high-rep projects with little funding.
- Projects should give higher reputation points to workers than training in guilds. Reputation points given are determined by the number of correct answers (based from Dawid and Skene Model) that they get.
- Assists the recommendation system (refer to Daemo).
- Projects can sell their finished AI products with tokens at a discounted price or for free in exchange for the share of the ad revenues.
Guilds
- Main role is to train members to increase their reputations.
- High-reputation members are democratically voted to become trainers. There’s a minimum reputation required to be legible.
- Voting allows competition for trainer position.
- Number of trainers is directly proportional to the guild’s population.
- Trainers are regularly rewarded by the Guild funds. In exchange, they must provide training sets for the rest of the members. Their training sets are released with their own answers. If they’re proven wrong through D&S, their reputations reduce.
- Each guild is allocated with equal funds.
- Top 5 (tentative) can be supported with stake-based voting. Projects are ranked for voting based on average reputation of the top 50% members. Top 20 guilds are displayed for voting.
- Reputation rate decelerates after every training.
- Each training set is given randomly to participating members.
Constitution - Serves as the agreement among all accounts.
- States all subjective rules and current plans that are being implemented.
- Displays current “technical leader”, which should be researchers for the platform.
- Type of leadership is stake-based (weighted + representative) democracy.
- Also, displays the current allocation of funds.
The Pit
- The place where reports (i.e. spammers, fake guilds and projects, etc.) are judged.
- Where the “police reputation” is based from. Uses Dawid and Skene Model.
- Punishment and rewards are distributed when a certain percentage (10 to 30%) of the stakeholders vote to either expel or save the account/project.
- Reward: “police reputation” – based.
- Punishment:
o Divesting - for bad elements.
o Decrease of “police reputation” - for wrong reports and downvotes. - Special Note: For cases within the gray area, the Constitution is the answer.
Recommendation System
- Uses the Daemon’s Boomerang Reputation Mechanism.