This is the second of a three part series detailing core ideas JPG has been exploring and building towards this year.

To maintain quality over spam, a TCR can use a reputation system.

There are implicit reputation systems, like the IRL “word of mouth” networks that have been around ever since humans developed speech. A more Computer Age example would be info gleaned from friends of friends on social media, such as NFT alpha seen and trusted via a peer’s peer on Twitter. These sorts of implicit systems tend to be largely, if not totally, structureless.

Conversely, there are explicit reputation systems that are specifically designed to generate and facilitate trust among people unfamiliar with each other, like members of online communities who have never met and in many cases never will meet.

9 important elements of explicit reputation systems

• Agents — the users to be scored, e.g. NFT curators or DAOs

• History — how past activity by users is tracked

• Control — how users are incentivized to perform desired behaviors, e.g. curating

• Presence — how available a reputation score is, e.g. very available = within a dynamic NFT publicly reviewable on an anti-fragile, always on blockchain

• Collection — how user activity info is actually gathered

• Governance — how authority is exerted over the reputation system, e.g. community votes

• Computation — how user activity info is computed into a score

• Representation — how reputation score is described and interpreted, e.g. a floating point number

• Interoperable — how openly a reputation system’s information is shared with the public

Examples of reputation systems

Academic (Non-Web3)

• GRAFT — a distributed reputation framework scoring online source recommendations

• P-Grid — a peer-to-peer (P2P) distributed info management platform

• XRep — a reputation-based trust management system for the Gnutella file sharing protocol

Commercial (Non-Web3):

• Amazon — registered users are able to write product reviews for the e-commerce giant’s shopper community

• Slashdot — a tech news outlet that uses an online reputation system to track its users’ “karma” based on their posting habits

• Turkopticon — a reputation system that helps crowdsourced workers find high-quality gigs on Amazon’s Mechanical Turk service

Web3 projects:

• Orange — a protocol for generating decentralized and portable reputation in Web3

• Gitcoin Passport — a decentralized identity record that lets people collect identity attestations dubbed “stamps” to mitigate blockchain-based Sybil Attacks

• RaidGuild Skill Forest — a Discord-based reputation system that lets RaidGuild members gain and track reputation from participating in “Raid” gigs and beyond

The use of reputation systems in online communities

The rise of the Web has given rise to an endless stream of online communities and user-generated digital creations.

These huge activity flows pose challenges for online communities, like 1) figuring out how to manage the sheer breadth of creators and content, 2) how to moderate bad actions, and 3) how to discern, surface, and reward the great things.

As such, reputation systems are a way for online communities to attempt tackling all of these challenges simultaneously.

Simply put, reputation systems can help achieve order and trust in online communities by reinforcing positive social behaviors. They track the actions within a group over time and then assess and score these actions so members can more adequately judge their peers and their contributions.

The vulnerability and defensibility of reputation systems

Depending on its design a reputation system can be attacked by an insider — someone who has permission to access the system — or by an outsider — someone who is unauthorized to access the system.

The motivations of reputation system attackers can be boiled down to one of two elements: malicious intent or selfish intent. A malicious attacker aims to degrade the availability or utility of a reputation system, while a selfish attacker aims to manipulate their own reputation for personal gain.

For example, there are “slandering attacks” where a malicious agent makes many false reports about other agents, and there are “self-promoting attacks” where a selfish agent attempts to artificially inflate their own reputation for greater power.

These sorts of attacks threaten the integrity of reputation systems, but fortunately there are defenses. For example, there are Sybil-resistant mechanisms like identity scores via non-transferable NFTs that make it hard for attackers to nefariously commandeer multiple “identities.”

JPG’s approach to reputation

To support the cultural value of NFTs, JPG has designed infrastructure to underpin JPG Canons, which are community curated and community governed lists of NFTs that are stored on-chain and incentivized through a bespoke reputation system.

To participate in this system curators mint a dynamic, non-transferable NFT that automatically tracks reputational score, and evolves visually accordingly, over time.

Holders will then be in charge of submitting proposals through this reputation system in order to create or delete Canons and submit or delete NFTs from existing ones. The more reputation points you accrue, the more voting power you have on these proposals. The system is one of positive reinforcement, as the only time one loses reputation points is if they have a proposal get voted down as spam.

If this type of curatorial effort is of interest to you, please reach out in the JPG Discord with questions or feedback!

Additional reputation system resources:

• Awesome Reputation Systems by MantisClone

• A Survey of Attacks on Reputation Systems by Kevin Hoffman, Cristina Nita-Rotaru, and David Zage

• Building Web Reputation Systems by F. Randall Farmer and Bryce Glass

• Reputation in Artificial Societies by Rosaria Conte and Mario Paolucci

• Reputation Systems: A Survey and Taxonomy by Kris Bubendorfer, Ryan Chard, and Ferry Hendrikx