How it works?

What is a reputation system?

Our reputation system allows Graphite users to build trust around phone numbers - people can vote for numbers they personally trust, for businesses they support or against spammers, frauds and malicious actors. The system is transparent and reputation is built solely by our users.

How voting works?

To avoid abuse and help build more reliable reputation we rely not only on plain user votes, but also account for each user's personal voting history and vote's historical relevance.

This way, not all votes are equal - the most just users and fresh votes have more weight applied to a phone number reputation than people who are trying to review-bomb someone or have a more outdated opinion

How weighting works?

The most simple explanation is that your clear vote is being amplified by other values

E.g., you vote +50 on some phone number. In a clear system without any weights your vote would be applied "as is" and would be counted along all the other votes equally. But in Graphite, your vote is only a part of what's applied to the reputation. After receiving your vote, we extend it from a -100 to +100 scale of votes to a more complex scale of reputation, which goes from -1000 to 1000

We do so by applying various weights, which combined can amplify your vote up to 10 times from its original value. This way, your +50 vote could easily impact a phone's reputation by +500 if the system sees you as a just person. The same way, someone with a malicious intent of, for example ruining his business competitor's reputation, could impact it down to 5 times less serious than someone with clear intents

User voting history weight

For user voting history we use a sigmoid function with custom parameters. This function calculates the weight in a range from 1 to 5 based on the user's history of giving negative or positive reviews. Currently, this only accounts for the fact if each review was either positive or negative, calculates the ratio and gives it a san input to the weight function. The closer this ratio to 1 (the more balanced the user votes), the higher weight the user will receive

Vote time relevance weight

For historical relevance we also use the same sigmoid function. This weight calculates the relevance of each vote given their relevant position in the list of calculated votes. The newer the vote is - the higher its weight will be

Calculating reputation

Given the weights of each vote as described above, reputation is calculated as a weighted average of all the votes, that were submitted for the target phone number. Each vote has its weights added to it, all votes and their weights are summed up and divided by the overall amount of votes submitted

Revoting

There is also a possibility to change your vote at any time. For example, a reliable business partner, for whom you voted positively before, suddenly changed his behavior and now you want to give him a negative review instead. In that case, you can revote - the system will automatically delete your previous vote and create a new one. This new vote will be taken into account as a fresh and relevant one, so both weights will be recalculated again - based on your current voting behavior as well