On the Nature of Reputation

Abstract: Reputation tokens (e.g. brands, but lot of other things as well) are vessels to store reputation. They are free to create but expensive to fill with trust. Consumers use them to deal with information overload. Producers use them to manipulate common knowledge. Also, a certain kind of supply-demand-style equilibrium exists.


Speaking about reputation, one risks getting all hand-wavy and disconnected from the real world.

To keep the discussion down-to-earth, let's start with the trademark law. The trademark law was, after all, created specifically to deal with reputation. And being a law, it is not a theoretical model but a living organism evolved to solve actual real-world problems.

You may have a vague idea that a trademark is something like an Internet domain name. You can buy a name and then it's yours. It's your property and everyone else should get off your lawn.

But the trademark law is not like that at all.

First of all, you don't create a trademark by buying it. Trademark is rather created as a by-product of using a name. You establish a company. You start producing stuff. Maybe you do a little advertising. People start to recognize your brand. Et voilà, you own a legally recognized trademark. No explicit action on your part is needed.

Similarly, if you stop using the name, your claim to the name gradually dissipates. If you claim a trademark and your opponent is able to prove that you haven't used it for decades, the court will rule against you.

There is also the concept of "trademark goodwill" which loosely translates to "reputation". Interestingly, the notion of trademark goodwill tends to be phrased in economic terms. Namelt, it is the part of the value of the company that is gained through owning the trademark.

Another common misconception about trademarks is that the names are, similarly to the Internet domain names, global.

In reality, the scope of a trademark is limited to the area of one's activity. Apple, the grocery store doesn't infringe on the trademark of Apple, the consumer electronics manufacturer. Similarly, Apple a grocery store in Manchester, England may not infringe on the trademark of Apple, a grocery store in Brisbane, Australia. The rule of the thumb is that there's an infringement only when the two trademarks can be easily mistaken for each other by the customers.

Reputation Tokens

Alice is a commercial programmer. She writes software and is paid money. Bob is an open-source programmer. He writes software and earns respect.

It kind of feels that money and reputation may be, in some way, similar. Maybe if Alice is paid money, Bob should be paid some kind of generic reputation tokens.

But the analysis of the trademark law above shows why the two kinds of tokens are vastly different.

First, reputation tokens (e.g. trademarks) are not scarce. Quite the opposite, they are completely free. You can spend the entire day in a bed inventing brands. And you will, in a sense, "own" those brands. But you can't do the same with money. You can't just think it out of the thin air.

On the other hand, you can fill those tokens (trademarks) with reputation. You can build high-quality products. You can advertise. You can force people (as states do) to buy your product.

So maybe a reputation token (trademark) is not like a value token (currency unit) but rather like a purse, or a bank account.

That must be true to some extent. The trademarks are really containers to store reputation. But at the same time they do differ from bank accounts. You can, for example, withdraw money from one bank account and deposit it to a different one. But once you have built a reputation of a particular brand, you can't easily transfer it to a different brand.

Maybe we should stop making monetary analogies and look at what's really going on.

What really happens is that Apple develops iPod and a lot of people start believing that Apple builds great products.

Investment results in more reputation for the brand, which, itself, is a kind of common knowledge, stored in the brains of people.

If you are not familiar with the concept of "common knowledge", here's a great, detailed write-up by Scott Aaronson.

A side note: The game theoretic model of common knowledge presupposes not only that everyone is aware of a certain fact but also that everyone knows that everyone else knows, that everyone knows that others know that everyone else knows, and so on. Note that this is not necessarily needed for reputation. People can believe Apple is a great company without knowing that others think that Apple is a great company. Apple would still be able to sell more iPhones. Yet, when you start looking at real-world examples of reputation it's hard to find an example where this recursive kind of common knowledge is not involved.

Social media example

Now, let's go beyond trademarks and consider social media. You can definitely gain reputation on Twitter or YouTube. So how does it conform to the model above? Let's have a look.

Creating a reputation token, say a Twitter account, is cheap. In fact, you can create any number of them cost-free. Filling the token with actual reputation is more expensive. You have to be engaged in conversations, write tweets that people find interesting etc. At the same time, the reputation is locked in other people's heads

However, unlike with trademarks, the amount of reputation stored in a token can be measured. The number of followers can be used as a rough estimate. Hearts can be used to track reputation on more day-to-day basis. These numbers can be used to compare individual reputations.

It's not clear to what extent such numbers are useful. One may object that reputations are multi-faceted and in principle incomparable. Reputation of an ordinary person cannot be readily compared to a reputation of a scientist, to a reputation of a pop-star, to a reputation of a politician. Followers may also value different aspects of a YouTube video.

On the other hand, consider money. Price is without a doubt a useful measure of value. Yet, it suffers from all the problems above. A price of banana is not really comparable to a price of bicycle. Some people may value an automobile because it can move them around, others may value the prestige of owning a Ferrari.

Reputation tokens: How exactly do they work?

I see two possible answers: Either the reputation is tied to a token (trademark, brand, name) or, maybe, it it tied to actual physical reality behind the name, say, the people who designed iPod.

This question can be answered on empirical basis: What happens when the reality behind a name suddenly changes? Do people still trust the name, or does trust follow whatever used to be behind the name.

Case study: MySQL database. The creators of the database sold it to Sun Microsystems, which then got bought by Oracle. At that point the engineers decided to fork the database and create MariaDB. Here we have a situation where the reality behind a name have changed. MySQL was from that point on developed by a different team. At the same time we can compare it to the old team, building the same product, under a different name. The question is whether the former trust in MySQL remained with MySQL or has shifted to MariaDB. If the former was true, people stored their trust in name "MySQL". If the latter was true, they've trusted the database and people who have built it.

It's hard to get data on this, but my impression is that the trust remained mostly with MySQL. And that was not because people trusted Oracle more than Monty Widenius, but because they've been using MySQL as their dtaabase engine all along and have even spent a thought on changing the status quo.

My conclusion therefore is that people trust names more than than whatever they happen to refer to. And we'll see why that is important in a minute.

Operations with reputation tokens

Before we move on to more interesting stuff, let's get a better feel for the reputation token concept by asking what can people do with them. You can do a lot of stuff with money. You can exchange it for goods and services, you can lend it, you can borrow it and so on. Can you do similar things with reputation tokens?

Well, you can definitely sell a reputation token (a brand). The fact, that the trust is associated with the brand name and not the physical reality behind the name, as discussed above, means that the owner of the token may change and the trust will remain intact.

You can think of it more broadly than just as of selling brands. For instance, the reputation token "Government of Italy" is transferred each time there's an election, or, smewhat more often, when there's a change caused by a government crisis.

Two tokens can be merged to refer to refer to the same underlying reality. Vauxhall and Opel names used to refer to different cars in the past. Today they refer to the same cars. (As far as I understand, Vauxhall branding is used in UK, Opel elsewhere.)

A token cannot be easily split. After a split, one of the child entities would need a different name and it would have to build its reputation from the ground up.

Why do consumers need reputation tokens?

All of that is nice and well, but the real question is, who needs those tokens in the first place? Why can't we just have a life and not care about them? Or, taken from a different perspective, what are the forces that shape how the ecosystem of reputation tokens works?

First, let's have a look at consumers. (The naming here is quite arbitrary. By consumers I mean people how buy the branded products, that is, people who put their trust into a brand.)

Consider following questions: Should I get vaccinated? Do vaccines cause autism? Should I look into the research? The study about vaccination/autism correlation was published in a high-impact journal. Does it mean it is trustworthy? I've heard it has been retracted. How do I verify that? And what does it mean? I've heard bad things about Elsevier. Should I distrust everything they've publish? What about replication crisis? And who, at the end of the day, decides which journals are high-impact and which are not?

Navigating the world, especially the modern world, is hard. When facing the anti-vaccination movement many people, without thinking about it twice, assert that the educational standards at our schools are too low and should be improved. That would surely make people believe in vaccines.

The problem with that is that's not feasible. Understanding the science around vaccination is hard even for experts and a layperson cannot really hope to get a correct, balanced picture all by themselves.

And it's not just the vaccination. The same applies economics, politics, climate change and even to much mundane problems like which smartphone should you buy or which school to send your children to.

In fact, even a person with IQ of 150, superhuman googling skills and a 10TB hard disk plugged directly into the brain, doesn't necessarily want to care about all of that. They may care deeply about one or two issues and would be happy to delegate everything else to the experts.

And I think you can see the problem now. Which experts should be trusted? When trying to find out, one runs into the same kind of complex, branching, fractal problems.

The common sense answer is to trust the authority. In the case of vaccination, that would be your doctor. You ask her: "Hey, doc, should I get vaccinated?" And she replies that yes, you definitely should. Problem solved.

But again, how do you know that you should trust doctors? You don't have the skills to verify the trustworthiness of the medical profession.

In the end, you simply trust doctors because most people around you do trust doctors. (You may recognize the recursive nature of common knowledge here!) It's not a safe bet, people can be wrong after all, but it's better choice than trusting nobody and doing your decisions at random. Or than trusting some shady YouTube channel with three subscribers overall.

You assume, without verifying it, that doctors must have done something to gain all that trust. Maybe one day someone got treated by a doctor and they've felt better the next day. Or maybe the doctors have some kind of exam that prevents people without extensive knowledge and experience becoming doctors. Who knows. You just trust them because everybody else does.

(And it goes without saying that according to this logic, if you live among anti-vaxxers, you are going to trust their authorities and you won't get vaccinated.)

Why do producers need repution tokens?

A terminology note: By producers I mean people who own tokens (brands) and invest in putting more trust in them. In short, Steve Jobses and Italian Governments of this world.

Once I've read a story about a regional government in Africa (I a not able find the reference right now, but I think it was in Acemoglu and Robinson: Why Nations Fail) which decided not to enforce the written law of the country and to rely on the custom law. Interestingly, their argument wasn't that the customs were in one way or another better than the official law, but rather that the written law can be easily changed by the central government - whom they mistrusted - whereas the customs could not.

Think about what that means in terms of reputation tokens. There's a token called "written law" owned by the central government. There's another thing called "customs". (It's not clear whether it should be called a reputation token because it's just a rule saying: "Do as your forebears did," and nobody really owns it.) But: If the customs happen to be the same as written law, why would the central government prefer the former to the latter?

And put this way, the answer is obvious: Because they can change the law, if they want to. The customs they can't.

Let's return to the idea of reputation as common knowledge. Common knowledge is hard to change, because it's stored in people's brains and most people are not particularly keen to change their beliefs. Also the recursive nature of common knowledge slows down the process considerably. If your trust is based on other people's trust then you'll have to find out that other people have changed their mind before you change yours. But if everybody follows that algorithm, the status quo is preserved. (Consider the study showing that people living in the territory of former Austria-Hungary have more trust in courts than people from former Ottoman or Russian empires. At least some of the reputation has survived for a century!)

That resilience to change fully applies to "customs".

But "written law" is different.

People do not trust the particular wording of the laws, they put their trust into the token called "law", irrespective of what it refers to. That allows the state to change a law while at the same time keeping the trust.

The same way, if people trust in "Apple" brand rather than a particular engineer in the company or a particular product (iPod), Apple can introduce a new product (iPad, iPhone) and people will embrace it. If, instead, a different company tried to sell exactly the same device, convincing customers to even give it a try would be an uphill battle.

To sum it up, reputation tokens are a mechanism to manipulate common knowledge. Technically, it's done by separating the name from the physical reality behind it and giving people an option to trust the name instead of what it refers to.

Is there an equivalent to demand/supply equilibrium?

I haven't thought much about the topic but the actors seem to have to similar incentives to make trade-offs.

Consumers want to put as much trust into a token as possible so that they don't have to make decisions themselves. On the other hand, they can't trust it too much, because then it would be easy for the brand to make profit by manipulating their beliefs.

Producers would prefer to invest as little as possible to building a reputation, however, with no reputation at all nobody would listen to them. They have to keep a balance.

February 20th, 2021

Discussion Forum