Creating safe and engaging consumer experiences in web3 is hard. And it's not just because of 'UX problems'. Developers have to navigate through the motivations of users, the utility users want and their revealed preferences — when data is open and on-chain, yet very noisy.

There is massive value in helping on-chain communities navigate through clear reputation signals for everything and everyone on-chain — users, transactions, apps, and smart contracts. Imagine if users in web3 could discover, read, fund, use, or buy something on-chain without worrying about getting spammed or scammed.

Developers and community leaders who are building novel social, consumer, or community experiences still face significant challenges in defining and using trust signals, heuristics, or algorithms for reputation in their contexts. However, we are seeing vast experimentation around primitives such as “Community Notes” to enable a better model for what (or who) can be trusted on the internet.

We‘ve learned some key things from our work with partners and fellow builders in this rising era of consumer and social applications in web3.

• Social and Consumer app feeds as we know them today, will go beyond content discovery. They will enable social discovery spanning different social graphs, networks and types of actions for a user.

• Reputation will be key to keep users safe, make experiences personalized and engaging and ultimately power permissionless innovation. There are emerging examples to identify and use reputation signals and heuristics.

• As a primitive, reputation graphs can help drive the desired utility for apps and marketplaces. An open and verifiable compute layer for these reputation graphs will power ideas and business models that weren't possible in web2.

Let’s unpack these.

From Discovery —> Social Discovery

Online discovery, which is traditionally a tab or a feed in an app that helps people browse based on some recommendation algorithms, may actually manifest very differently in web3, where data is open and interactions can be embedded across any user interface via contracts.

Usually, effective discovery is based on a collection of massive user data, and sometimes, it is based on curation of a user's social graph which helps in recommendations in that context. For example, we rely on foodie friends to recommend restaurants but might also rely on a centralized marketplace to give us their recommendations for the top restaurants.

The innovation surface for social discovery remains largely an uncharted territory in web3. Presently, social feeds are simplistic, primarily aggregating historical posts. The introduction of Lens Open Actions, MetaMask Snaps, Farcaster Frames, and Uniswap Hooks can transform these feeds into dynamic activity hubs, prompting users to discover and engage with many different artifacts or utilities. While the current focus in the space is mostly on transactional activities like mints and votes, these developments are just the beginning. We're moving towards making social feeds gateways to meaningful, interactive, multiplayer experiences.

We envision a future where a user can curate their own feed. We'll have the tools to empower developers and users to build and leverage their revealed preferences, instead of relying just on singular and opaque walled gardens. With an open and composable data layer, users can even see the feeds of people in their network or any on-chain user. Imagine you can see Elon Musk's Twitter recommendation feed and see the information that shapes his worldview (hah).

The challenge and opportunity is in inspiring clients and applications to evolve their approach to social discovery on-chain. We need to move beyond replicating the standard models of platforms like Twitter or TikTok and create something uniquely engaging and valuable in the context of our evolving digital landscape.

The Composability paradigm

A core challenge in web3 is how to best acquire and engage new users. Besides continuously lowering the delta between web2 and web3 UX, the emergence of embedded wallets, and interfaces such as Uniswap Hooks, Metamask Snaps, Intents, Farcaster Frames or Lens Open Actions, manifest a paradigm shift in UX in web3. These mini apps, akin to OpenAI's GPTs or Apple's App Clips, offer smaller, more focused, and bespoke interactions.

What does this innovation imply?

• The understanding that a feed in a social app is only about content is becoming an outdated notion since every social post can be a smart contract itself;

• The line between Dapps and bots is getting very blurry;

• Successful Dapps do not have to look like Dapps at all; they might even not have their own client but thrive through smart contracts therefore meeting users on whichever platforms users are. Imagine an open version of Facebook as a distribution channel without ads.

We envision a vast design space for composability among use cases, without the need for individual apps or clients. Bountycaster.xyz on Farcaster is an excellent example of this capability. Any user can post a bounty in a Farcaster post tagging Bountycaster’s robot, and access its service functionalities without leaving Farcaster’s interface. Lens Open Action bears the same big unlock, as any Dapp could test out product-market-fit through the existing social graph on Lens much more easily.

Reputation the 🗝 for Permissionless Innovation

Reputation for safety and security

Mini apps or bots performing automated smart contract actions represent a significant opportunity for designing novel consumer experiences. However, being permissionless, users also face a different level of challenge in terms of safety and security. Moreover, the noise and scams could discourage users to stay or come back.

Users need reliable signals to confidently gauge the safety when interacting with a particular Snap, Hook, or Open Action. The central quest here is whether we can have effective and evolving reputation mechanisms for permissionless consumer experiences to keep users safe and engaged without centralized gate-keeping?

Our exploration into social peer-to-peer mechanics has shown promising results in detecting likely spam and sybil clusters. One such interesting exploration was when we computed and visualized the reputation graph of a sample of Lens users, showing the interactions between users of different reputation scores. We were able to highlight sybil and potential spam clusters based on their network behaviors. Additionally, we've explored different ways of bootstrapping and growing a permissionless trust graph for users of MetaMask Snaps, which could potentially be adapted across various marketplace use cases.

Social Graph as the organizing structure of Trust and Reputation Signals

Generally, people interactions are either socially generated, meaning they happen within their social networks and related contexts, or generated when people interact with various institutions in which they (have to) place their trust. Whereas, the central notion of web3 innovation is to replace opaque and often flawed centralized institutions with decentralized networks and communities.

Despite the early innovation and potential of web3 social, most of the existing data in web3 is not peer-to-peer, let alone socially generated. However, with rapid innovation around consumer experiences, the data layer is going to get increasingly rich. Data will be available from peer-to-peer interactions, and on-chain attestations and many of the scenarios will be social. For example, your engagement on Farcaster or Lens will lead to discovery opportunities for certain NFTs, and further, the ownership of certain NFTs will boost your new social circles and experiences.

All of this already happens in the physical, off-chain, and web2 world, but within siloed contexts and data layers. web3 composability finally gives us a chance to form new social graphs for these interpersonal trust and reputation signals. Besides, to a certain degree, every chain is already a primitive social graph, every entity has an identifier and has been forming interactions of all sorts.

Implicit Signals versus Explicit Signals

Trust and reputation signals are natural reactions occurring when people interact with each other or with certain objects. Signals are embedded in the user interfaces of applications. For example - Follow, Like, Mint, and Upvote on any social/consumer app. This is why interface and UX design have a significant influence on how users engage in an app and produce a trail of these useful signals.

Differentiating between implicit signals from explicit signals is important. An implicit signal could be something like owning an NFT or sending your friend some stablecoins. In both these cases, we can derive an implicit trust link between you and the NFT you owned, or the address you sent money to. An explicit signal would be you issuing a specific attestation to a contributor in your community, or Following someone on Farcaster or Lens. In both these cases, there is a direct trust assertion from you to the recipient, in that specific context. Both signals are important, but implicit signals are overlooked and might open up untapped user stories.

Using Existing on-chain transactions as Trust Signals

In a mature community, that has sufficient peer-to-peer signals, transitive trust can be computed and used to build reputation. However, bootstrapping until we get plenty of p2p signals is a different battle to fight. For the bootstrapping phase, existing on-chain transactions such as voting history, NFT ownership, credentials, and so on can be valuable signals to help define the initial reputation of players in a system. For example, in bootstrapping MetaMask Snaps trust graph, MMG token ownership is essential to depict the value-alignment of a user.

In some scenarios, financial transactions can be much stronger signals of trust. In our survey of Lens developers, we found that financial signals such as paid-collects might be most useful to determine peer-to-peer trust and interest.

Existing on-chain transactions could also be good approximations of trust signals when they happen between EOAs. Similar to the use case of Venmo, transactions around income and spending could be associated in a social context, what if ethereum itself is a social graph, once you discount the noise? Some consumer apps, such as Interface.xyz and Zapper.xyz, are trying to utilize such data to show users meaningful activities to discover from on-chain transactions. These are valuable experiments. If we further marry social graph data with on-chain transactions, we could portray the trustworthiness of transactions, addresses, smart contracts, Dapps through an N-degree consolidated social trust view, and help users to navigate more safely and effectively.

Emergence of Reputation Graphs

Gathering rich and relevant reputation signals will help build Reputation Graphs which can power novel consumer experiences. Although bootstrapping the graph is inevitable but an often underestimated difficult phase. Figuring out how incentives could work is important — for example understanding implicit motivations that harness social capital instead of a sole focus on financial incentives. We have discussed this at length in a previous essay.

Our recent exploration evolves around surfacing a potential trust graph using existing on-chain transactions.

Building Communities using Reputation

When we initially started building decentralized reputation systems, we imagined it as a challenge to perfectly install a technology infrastructure, an intuitive and organic interface, and a set of rules that captures interaction signals and spells out effective indicators of reputation for each player.

However, as we went deep into the weeds of designing and implementing such systems with our design partners, we realized that an important aspect is also to build communities using reputation. For example, starting an on-chain community or a channel on a social platform , with a reputable host or moderator curating the group. The bespoke construct of reputation becomes the differentiating identity and culture of the community, attracting the intended crowd into the intended roles. Another example is building a community of security experts who can point out loopholes in applications or protocols. This can be built using contextually relevant peer attestations. The community takes shape and evolves to mature around the reputation of desired security assurance efforts. The right people will be attracted and encouraged by their behaviors in line with the goal of the community.

Together with IDEO Colab and We3, we are experimenting a Lab Approach to provide a design framework for communities, apps and protocols that want to build or use a decentralized reputation system. We believe social and consumer experiences will thrive when built around an intentional process of understanding reputation in their respective context.