On Monday, November 26, 2001, at 07:28 AM, Declan McCullagh wrote:
Thanks, Tim, for posting an interesting essay. You say:
Thanks for the thanks. It's just a facet of what I've been thinking about for a long time. I was bored so I just dashed off the piece, more to help crystallize thoughts than to lay out a convincing case. If others are partly convinced, so much the better. If nothing else, it's my stake in the ground that "reputation capital" is the _wrong_ focus. Since it's thrown around a lot here, someone ought to point out the many limitations of the concept.
2. When in fact different people have different assessments of some agent's reputation. Thus suggesting strongly that reputation is not something attached as simply as above.
To expand on this: It seems that using "reputation capital" to describe a multifaceted information space such as even the most wretched of cypherpunks posters does not do that person justice. Even if someone is generally correct and truth-telling, he may also be immoral or a fan of invasive police measures or immature in debate or prone to violations of netiquette -- any of which might lower his reputation capital.
Sure, one can say: let's just have a complicated reputation space (think an array of arrays) for each one of these characteristics. To use a silly example: * truthtelling [0-255] * maturity [0-255] * morality [0-255] * netiquette [0-255] * spelling [0-255] * etc.
This is not really an "array of arrays," just a garden variety n-space. And _not_ a vector space, because vector addition does not work. More of a tensor, in that the axes are independent. (The canonical tensor being the stress-energy tensor for a solid material, for example. The stress in the X axis and the stress in the Y axis do NOT "vector sum" to some resultant stress.) Anyway, I digress. I agree that a person may have various of these components. If I am betting on whether Alice will have few spelling mistakes in a post, I would look at her "spelling [0-255]" measure.
But that quickly becomes burdensome to use as a shorthand.
Though the place we need shorthand is for ordinary human to human communication, and for this it is enough to just say "Alice is a poor speller" or "Declan writes good articles." For a machine, maintaining such data bases is trivial. Not that I think this is especially useful. Think of just one of these facets, one of the most important ones: credit worthiness. Essentially, the basis of a bet on whether Bob will repay a loan.
It seems to me that reputation capital is a term that has limited value when applied to something as subjective as the areas above: having an article published in the editorial pages of the Wall Street Journal (or the Journal of Socialist Doctrine) may lower your reputation capital among some people and raise it among others. This is the nature of subjectivity.
Which is why I balk at schemes to grade Cypherpunks posts on the basis of "reputation capital" (an idea which bubbles up every couple of years, though no one ever bothers to implement it). Each person has his or her own idea of who they want to read, embodied in a) who they read vs. who they skim vs. who they hit "Delete" on, and b) their kill files. And who they write replies to.
Reputation capital is more valuable a term when describing traits that are less subjective. When dealing with an online ecash bank, you may want truthfulness and reliability and good customer service (for example), which are less subjective than "interesting political opinions."
The direction I'm going is all about digital money. My claim is that a belief ontology (risk, bets, futures, insurance, moneychanging, re-insurance, discounts, etc.) solves many if not all of the "thorny" problems often cited with digital money. The underlying mathematics, important as it is, just becomes another part of the calculation of beliefs. It doesn't matter to me much that folks use the term "reputation capital" loosely to describe actual people...that's just wordage. More important is the application of probabalistic reasoning to digital economies. Maybe I'll have something to show along these lines some day. --Tim May