Wednesday, February 08, 2006

imho James Kobielus on reputation


Pointer to blogpost where at least one person and perhaps two in collaboration (Marco Barulli and Giulio Cesare Solaroli) imho’d something that I had previously imho’d:

Kobielus self-regarding kommentary (but really, kommentary on the Barulli/Solaroli proposal—trust me—it’s not all about me):

I want to thank Marco Barulli for his kind words about my prior blogged thoughts on reputation. I wish I could bottle such beauty and uncork it on down days.

I also want to thank Marco (and I assume, Giulio as well, though Marco was the one who e-mailed me to inform me that he had commented on my thoughts) for his/their thought-provoking proposal to operationalize a computational reputation service. By the way, folks, I’m also giving a shout-out to Phil Windley’s recent blogpost on computational reputation. Good stuff—all of it. However, I’m the one who pointed out to Marco to note in his blog that my January 25 blogpost (high-level, conceptual, not implementation-oriented) on computational reputation came first. Just because. If I don’t take pains to point out these simple facts early on, the historical record might remain distorted (to my disadvantage) forever. Clearly, each of us is the number one stakeholder in and steward of our own reputation. For the record, though, I think Windley’s (and Barulli/Solaroli’s) blog(s) is/are terrific. I’m not a glory hog on such things: If Phil developed this computational reputation ideas independent of me, so be it—he’s developing them to an extent that I’m not—and I’m reading him avidly. (Besides, a quick Google on "computational reputation" shows that the concept and term precedes us both by a mile--so there's no point stoking my vanity any further with this silly tangent).

Now onto the Barulli/Solaroli proposal. They focus on the following possible use case for computational reputation (quoting from their blog):

“We focused on a special kind of reputation, one built from the whole set of comments you sent to blog authors and their acceptance or rejection. A little portion of “who you are”. Given that set, everyone could “run” his personal reputation application and compute your reputation as a commenter. Unfortunately this set, your full comment history, is not easily available. Most of the times, a blog author will have to decide to accept or reject your new comment judging solely from the content of that very last submission. This was the problem we were trying to solve.”

This use case had me scratching my head. First, I had to mentally diagram the entities and interactions. Let’s see:

  • The reputation-dependent access-control decision is “accept/post or reject submitted comment.”
  • The reputation-relying service provider is a “blog author,” who selectively accepts (and rejects) third-party comments for posting to the blog.
  • The reputed party is the “blog commenter.”
  • The reputation-asserting authority is the “reputation manager” (which may be an automated process that calculates reputation scores that get attributed to the blog commenter and relied upon by the blog author ).
  • The reputation score associated with a blog commenter is computed by the reputation manager from the transactional history of acceptances and rejections from other blog author to whom the commenter had submitted comments.
  • The reputation score for a commenter is contained in a “comment token” that is asserted/issued by the reputation manager.
  • The current/computed comment token for a particular blog commenter is issued by the reputation manager upon request by that blog commenter.
  • The blog commenter receives that comment token and includes the token—plus a fresh comment—to a blog author for consideration.
  • The blog author considers the fresh comment, in the context asserted by the comment token; makes an accept/reject decision; and asserts this fresh accept/reject decision both to the blog commenter and the reputation manager.

Did I get that right, Marco? I await further details on your approach, which you’ve promised for subsesquent blogposts.

My first reaction: This is not an appropriate use case for computational reputation.

Most important, the blog author’s decision to accept or reject a comment wouldn’t normally depend on any knowledge of the commenter’s reputation. It would depend purely on the merits of the comment itself (was it interesting, stimulating, on-topic, etc)? Why should I, as blog author, give a damn what others think of this commenter, or whether (unseen? unknown? unintelligent?) other blog authors have accepted or rejected this individual’s prior blogposts?

There are only a few scenarios in which the blog author would care about the commenter’s reputation, as asserted by the (computed/weighted/aggregatged) opinions of other blog authors. For example, if the blogosphere has identified this individual as a serial plagiarist who keeps submitting other people’s words as his/her own, then, yes, I may choose not to give that intellectual property thief a forum to rip off other people’s work.

I can’t think of another scenario where I would give a crap about someone's reputation when considering their ideas. I like to think of the blogosphere (and democratic societies in general) as intellectual meritocracies.

I choose to consider only the text (ideas/thoughts) you present to me in conversation, not the irrelevant con-text (e.g., other people’s prejudices or grudges against you) that others want me to consider.

If I were to consider loaning money to you, that’s another issue entirely. I would definitely consider your reputation as derived from your transactional history (as reported/asserted/scored by others) of paying your debts on time or being a serial deadbeat.

For the record, I don’t accept comments for posting on this blog. This is my personal pulpit. If you all want to respond to these thoughts, get your own blog. Or do what Marco and others have done. Send me a note at

If I find your comment interesting enough, I’ll do what I did here. I’ll blog on it. Respectfully. And, hopefully, the act of my doing so will add to your reputation scores as computed by others who value what Jim Kobielus has to say on such things).

My personal calculation of you is purely fuzzy math. Extremely fuzzy. Warm and fuzzy.