Sunday, April 12, 2009

Reputations

I do not claim to be an expert on reputations systems. There are many who have gone way deeper on this subject than I. Instead, I hope to seed a conversation about the complex interplay between reputations currencies, exchange currencies, and the marketplace. While there are a whopping variety of reputations currencies in play today on the web, I see two basic categories.

Do I live up to my promises?


When two parties come together to make a transaction, both sides make a variety of promises. As a player in a given marketplace I get a reputation about whether or not I live up to my promises. Some of these promises are explicit and some are implicit. The seller explicitly promises things about their goods or services, such as “This mouse works with USB.” If I get home and the box has been incorrectly labeled, I am legally allowed to return it, since the seller’s explicit promise has been broken. If the seller frequently breaks their explicit promises, they earn a bad reputation and buyers avoid them (they may even be sued). Ebay makes excellent use of this principle with their “thumbs up / thumbs down” system that allows buyers to rate sellers.


On the buyer’s end, the story is a little more complicated. In some marketplaces, the promise might have to do with timely payment. When I receive an invoice, I have until a certain date to settle my account. If, as a buyer, I stop settling accounts on time, I develop a bad reputation.

However, if we look a little deeper we see there are explicit promises backing up the exchange medium itself. Money is an IOU, and, as such, someone had to make an explicit promise to a bank to pay back a loan (with interest) in the future for it to be issued in the first place. While that kind of explicit promise is at the foundation of exchange currency, we rarely perceive money that way. If I develop a bad history as a borrower of money, my credit rating (a reputation currency) is negatively affected, and I lose access to exchange currency.


Reputations outside the explicit promise


Let’s say I own a pizza place. I advertise that, because I care about the environment, I will only employ people who deliver pizzas with their bicycles. You care about the environment too. You order a pizza, but when you get it, the delivery person is driving a Hummer. You feel cheated. Part of what you were paying for was benefit to the environment. This is an example of the first kind of reputation. I as a seller broke my explicit promise.


But what happens if everything is the same, except I didn’t promise bicycles for delivery vehicles? You might still be outraged that a Hummer was the delivery vehicle, but I didn’t promise anything to the contrary. This situation is an example of the second kind of reputation. This kind of reputation may have been what Adam Smith was talking about with his prescription for successful markets: there must be symmetrical information between buyers and sellers. I didn’t know I was supporting a pizza place that used Hummers. I accidentally caused an environmental externality that could have been avoided if I had had access to better information. We could solve this problem by introducing a reputations currency that measured buyers’ perceptions of sellers’ ecological impact. Buyers who cared could go to only those sellers who also cared (or more precisely, imbued buyers with that perception).


Complications


This situation gets murky very quickly, however. What happens if, in the above scenario, I advertise that I, as a pizza seller, care about the environment, but don’t make a promise about delivery vehicles? Perhaps I compost all my waste, or recycle soda cans, or use environmentally friendly cleaners. Since, you as a buyer sitting at home don’t see those behaviors, you are not getting the whole picture. Again, we have asymmetric information.


I might incur an unfair reputation as an ecological monster, since the people who contribute to my reputation don’t see most of my environmentally friendly behavior. A buyer-issued reputations currency would therefore bias me towards enacting environmentally friendly practices that were visible to buyers. However, many of the most important environmental practices might not be things that buyers normally see.


Enter the third party rating service. I, as a seller, can choose to be environmentally audited by a disinterested third party. An auditor gets a thorough look at all my practices and issues credentials to me based on them (another reputations currency). However, there is a tricky balance in this transaction. Since I, as a seller, probably have to pay for this service, I want to make sure there is a good chance I will receive the credentials in question. Therefore, I am biased towards hiring less stringent auditors. However, buyers are not served by this arrangement, and if word gets out that the auditors aren’t stringent, the auditor loses its reputation as an auditor. This means that auditors are trying to appeal to sellers as likely to grant credentials, while simultaneously trying to appeal to buyers as stringent.


One way to solve this dilemma might be to have the buyers rather than sellers pay for the auditing. However, sellers still need to grant access to the auditors. They might only want to grant access to auditors who are less stringent. And if auditors don’t get access to a broad enough swath of businesses, how can they charge buyers for their service?


Also, there are clearly situations where sellers make implicit rather than explicit promises. I might, as a seller, imply something about my practices or my product without making an explicit promise about them. Into which category should we put reputations around these promises? Due to the subtle nature of these reputations currencies, an open source development process that includes the communities that make use of them is the only way to ensure they can be sufficiently dynamic.


Domains of Trust


One final complication: Let’s say we are aggregating reputations about the first kind of reputation (do I live up to explicit promises?). Because (in a hypothetical near future) we have transitioned into a fully open data context, we can see the Ebay seller rating of someone as we buy something from them on Craigslist. But, does an Ebay rating have any meaning in a Craigslist marketplace? It does, IF AND ONLY IF, the rating can travel in the other direction as well, and I incur a consequence on Ebay as a result. If I transact with someone on Craigslist, and they decided to trust me on the basis of my Ebay rating, I could very easily abuse that trust unless they have the ability to post a rating about that transaction back to Ebay. Therefore the consequences of my actions must be applied across all relevant domains.

This gets even more complicated though, because how do we know that the person on Craigslist is rating me based on the same criteria? If they weren’t, we would be comparing apples and oranges, and my Ebay rating would cease to have any meaning at all. The only way to solve this problem is to have open standardized reputation currencies that could be imported and exported across domains.
Meta-currency project, here we come!

3 comments:

Arthur Brock said...

This is a dicey topic which probably merits another whole post to sort out some of the issues.

I think we should acknowledge that we already use reputation currencies. Every certification and formal title is a kind of reputation currency (UL Certified, FCC Compliant, Certified Organic, Justic O'Connor, President Obama, BBB Gold Star, Olympic Gold Medal, Academy Awards, Harvard MBA, etc.)

We manage to use most of these currencies just fine without much the confusion you refer to.

All currencies are actually a social contract. This is most obvious in reputation currencies. But we have done our damndest to try and pretend that dollars are devoid of social contract and are "just business." We then commit all sorts of folly by creating reputation currencies and confusing them with medium of exchange currencies. (Examples: Facebook's new "I'll give you 5 credits for that cool post." or
Twollars being a "thank you" currency which is configured as a medium of exchange with a credit limit, current balance, and convertibility to donations.)

Reputation is always context dependent based on the norms within a community with which you hold the social contract. I could be known in the business community as a sharp, reliable, mover and shaker, but in my son's Little League community as the bozo who picks his kid up late and never comes to the games. Different needs, expectations and social contracts.

A very few of those factors might be able to be translated across community contexts, but it's dangerous to do because how they are measured embeds all sorts of unspoken assumptions from the social contract which may be very different in another context.

Alan Rosenblith said...

Thanks for another very insightful comment! I completely concur. There are hidden assumptions embedded in social contracts that make translation to other communities very difficult. That is why I was distinguishing between explicit and implicit promises. I think many of those hidden assumptions fall into the implicit promises category.

If I say I will send you a product within 24 hours, and it takes me a week, I have broken an explicit promise. That seems to be something that could translate across boundaries. But if I say I will send this product in a timely manner, that might mean something different to you than it does to me. In my community a week IS in a timely manner.

That is why if we want to leverage these reputations currencies to the fullest extent possible, we will need open standards. Many people don't know how their credit ratings are calculated, and as a result they have fear around it. They don't know the rules of the game, so they are in a condition of learned helplessness.

Therefore we need to make explicit the rules of each reputations system. In an explicit context, I suspect there might be more opportunities for translating reputation across boundaries.

O Newhouse said...
This comment has been removed by the author.