Saturday, December 08, 2007

Reputation economies

Reputation Economies Symposium at Yale Panel I: Making your name online!

Moderator: Laura DeNardis: Online, this panel might look different: the audience could rate the panelists, remix them, etc. The moderator could allow the panelists to pay for placement; there would be ads.

Michel Bauwens: His P2P Foundation wants to understand P2P dynamics in every area of social life. Key feature: Self-organized; structures arise from individual choice.

Emerging models: (1) sharing content, for self-directed reasons – there are weak ties among sharing communities; platforms like Google offer the ability to share the content because individuals can’t do it themselves. Most of the people in those systems are not there for monetary reasons. And there is a risk of crowding out nonmonetary incentives if you integrate revenue-sharing mechanisms (Benkler). Monetizing a sharing economy is a wrong.

(2) Commons-oriented production. People working on common product, e.g., Linux or Wikipedia. They have to engage with each other. Instead of using proprietary platforms, they use something like an NGO. The NGO creates an infrastructure with enough viscosity to sustain the project without monetizing it. (Comment: I see media fandom, where fans create fiction, videos, art, etc. based on popular shows, books, and movies, as having strong “engagement” features because in practice fanworks reflect on and engage with each other, and of course fans provide each other feedback. This is true even though Bauwens would categorize fanworks, hosted by corporate platforms fans don’t own, into (1) – the Organization for Transformative Works is an attempt to give fanworks a protected infrastructure to keep it from corporate appropriation.)

(3) Crowdsourcing: partial integration of P2P dynamics into the corporate value chain.

Rishab A. Ghosh: Economically, reputation is information about goods/people that is not encapsulated in the price. The formalization of the market price mechanism is missing in informal economies, which hinders decisionmaking, entirely separate from issues of incentives. How do you decide which open source project you want to devote your time to? This exists in formal economies too because price doesn’t encapsulate everything of relevance to your decisions.

“Exchange rates” of reputation: the bigger you are, the better of a deal you get – Linus Torvalds gets more than twice as much reputation than a programmer who does half as much useful work. (This is detailed quite intriguingly in Ghosh’s position paper.)

We need better ways to segregate reputation, so that the very best Provencal restaurant in town doesn’t get overall bad ratings because most people just don’t like Provencal food. Online reputation systems today tend to treat reputation as a single objective fact, but reputation is subjective and we need to figure out how to handle this context-sensitivity.

Auren Hoffman: The amount of data being collected on you is growing rapidly. It’s been growing every year. The rate of growth is not slowing. The number of actors collecting data is also growing hugely – it’s not just the government, as in 1984. Opt-in systems never reach critical mass. Opt-out gets critical mass and high adoption rate quickly – everyone wants in on Doubleclick and Experian’s databases. Benefits of opt-out: readily accessible credit, microtargeting (less spam), cookie-enabled websites. Consumers so far have made this bargain. The problem is these exponentially increasing curves are hurting privacy, and users are starting to rebel.

Hoffman proposes a new bargain – he thinks it will either be consumer-driven or government-imposed. Three components: (1) you should be able to find and know your data wherever it exists – no one should know more about you than you do; (2) you should own it, and be able to opt out either totally or selectively; (3) you should be able to share your data – move it to any service you want.

Moderator: we’ve heard a lot about the negatives of this shift to peer reputation. What are the positive aspects?

Ghosh: Reputation is essential to everything we do – as more decisions get made without the information provided by market pricing, reputation is more important. It’s just not now being implemented by mechanisms that work the same online as offline.

Hoffman: Reputation is contextual – Bill Clinton can be a good leader and a bad husband. Yale has a good reputation, unless you’re from Harvard. Reputation is overall an unsolvable problem. Only chunks of it can be measured well – the FICO score is good at saying how likely you are to pay back borrowed money, but nothing else.

Bauwens: We’re used to kinship groups. New mechanisms allow affinity-based groups to get larger and to trust one another in ways they couldn’t before – traveling to a strange city to stay on a stranger’s couch. (Fandom again!) What doesn’t work, and maybe shouldn’t, is going from one’s slashdot reputation to one’s eBay reputation. There’s a danger of a lowest-common-denominator effect if too much linkage of reputation occurs. If you use everyone’s rating, you get the worst overall. But if you look at what the people you most admire have been liking, you do better. (Bauwens contrasts digg to here.)

Hassan Masum: Problems of information overload – how much information can we stand? Why do we think reputation is important – to enable distributed collaboration, to complement mass media POV, to search through masses of information, to feed our egos, to enable us to decide with whom to transact, or what?

How do we aggregate trust when it’s weighted, subjective, asymmetric, and context-dependent? (Asymmetry is an excellent, excellent point. I keep thinking that everyone here should look at LiveJournal’s friends mechanism; with filters, even mutual “friends” can implement asymmetric interest/trust in one another, though if you’re filtered out of an entry you won’t know about it unless you’re informed through other means. Trust is implemented by each individual through LJ filters, but usually not expressed to others; even the people on a journaler’s filter can’t see who else is on it with them – it’s a blogging BCC.)

Audience: would you rather be highly rated by a small group of peers, or the world at large? Are there areas in which we’re all experts? (Democracy?)

Exemplars: What would it take to pick the most useful voices? Amazon is trying a most useful negative/most useful positive review system, which sounds promising.

How does creation time and number of creators affect reputation? Realtime (chat) v. long-term (book, blog); solo (writing/programming) v. collaborative.

Beth Noveck: Reputation is not simply about risk (managing identity, authentication, offline/online distinctions). People have real economic opportunities to create value by working with strangers. Increasingly, online reputation isn’t the product of the individual alone, nor is it the product of the service provider by eBay, but it’s the collective product of the members who did the rating. It is those sharing communities that should decide the rules about the disposition of one’s reputation online.

Reputation can depend on specific transactions in a marketplace – the Amazon model. Limitations: you get sockpuppets, or you can’t succeed without already having transactions. There’s the Linked-In model of social connections, which depends on your friends participating in the network. There’s the resume model: just put it out there – but again that depends on who you are.

How do individuals without extensive experience in a particular community establish reputation and shore up their promises? How do they guarantee themselves? You need reliable sanctions for defection. How can we enable different social groups to coalesce into bonding groups, providers of insurance? (I’m not sure how this would work. As discussed above, being a good knitter doesn’t mean you’ll be a good eBayer, so how would your knitting circle vouch for you?) Assurance from micro-elites, backed up by real financial teeth, is the way to go.

In discussion, Noveck elaborated (and she's also written an article, as she notes in her position paper): she's interested in ways of bonding performance, say likelihood of showing up and having a good attitude on an assignment. General reputation -- the social interactions that are common subjects of gossip -- is less amenable to bonding.

No comments: