Re: Wikipedia & Tor

[Jimmy Wales <jwales@xxxxxxxxx>]

Paul Syverson wrote:
> I want to emphasize a central aspect of my suggestion: The goal is not
> just to provide a filter for abusive posts, it's to change incentives.

This is exactly the right approach!

> We can't know for sure without running the experiment, but my guess is
> that if abusive posts through Tor never succeed (OK perhaps virtually
> never), and if the process of posting through Tor informs posters of
> that fact, then Tor will become worth it for your admins. The abusers
> will disappear or greatly diminish because they will know from being
> warned, and if necessary from experience, that their attempts will
> fail. Posts through Tor will then mostly have value (in the sense of
> not being abusive in the ways that prompted this discussion.)

I would say that even some fairly slight changes to the incentive
structure may help a lot.  The less desirable Tor is for problem users,
the more they will shift to traditional broken open proxies.  We can
play whack-a-mole with these as we do now, while at the same time
leaving Tor more open.

> Yes, I know (and I'm sure Jimmy knows) that this won't solve the
> longterm underlying issues. Abusive posters will just move on to
> another avenue than Tor. But I think it will be a quick, cheap, and
> big win for both Tor and Wikipedia.

Yes, but I don't really mind them moving to other avenues.  That's the
point.  If I didn't love Tor, I wouldn't care about blocking Tor either.
 Let them abuse broken proxy servers, let them do whatever, that's fine,
we can deal with it.  We just want to open up to Tor.

> Yes, as Marc Abel suggested you could implement passwords, pseudonyms,
> or hell ZKPs.  But this is stepping onto the slippery slope of trying
> to solve the more longterm problem that using IP addresses in the way
> Wikipedia does is a temporarily useful kludge. (Kludges are great, but
> function creep is dangerous and can make for bigger problems in the
> long run.)

Let me see if I can explain a bit more of the math behind this.  I'm
just going to make up a hypothetical example.

Suppose 100 out of every 1,000,000 edits to Wikipedia is malicious.  And
suppose we study them and discover, hmm, 25 of them come from Tor, which
is easily blockable.  50 of them come from static ips or dynamic ips
that are expensive for users to get new.  25 of them are from broken
proxies.

Now, our present solution is to block Tor, do various things in other
situations, and this works reasonably well.  Of the 25 bad edits we
block from Tor, some portion of them surely shift to other means, but
not all of them.  So we find it to be a net win.

Except.  Except we don't really like to block Tor.

Now, fast forward, and imagine that the "expensive ip" situation goes
away in a few years, either due to widespread onion routing, or whatever
you may want to dream up that makes our temporary kludge of using ips no
longer functional.

Then we'll still only have 100 out of every 1,000,000 edits to Wikipedia
as being malicious.  How we'll deal with that is how we'll deal with
that, but that's fine.  We'll manage.

For now the key thing to do is to shift the incentives on the bad users
so that Tor is less desirable for them than playing with the broken
proxies or just doing whatever with a dialup account or aol addresses or
whatever.

--Jimbo