[EM] Delegable proxy/cascade and killer apps

[Raph Frank]

On Wed, Sep 10, 2008 at 7:38 PM, Michael Allan <mike at zelea.com> wrote:
> Call this the "formal defense" of the modern state.  It claims that
> the constitutional structures are not at fault.  The faults or
> failings in democracy are located outside of state institutions.  But
> whether we argue that government isn't broken and needn't be fixed, or
> that it's impervious to change anyhow, the issue is the same.  The
> crucial thing is location.  The fixes and changes are needed
> elsewhere.

Right, if the people are organised, they can change the constitution.

>> A telephone is worthless unless others have telephones.
>
> Nevertheless we have telephones.  Even without the equivalent of
> Bell's money and social connections (he married well), we have other
> networks that grew from small beginnings, pretty much on their own
> (file sharing for instance).

However, the internet didn't just appear.  First you had groups
(universities and the military) that valued data communications
highly.  They were willing to pay a lot for pretty low grade
communications by today's standards.

Once that was in place, it encouraged more universities to join.  As
time passed, the value of an internet connection increased.  Large
business might have added a connection.

Eventually, it was worth putting one in every home.

Few would pay EUR 20-30 a month for an internet connection if they
could only browse 2-3 university websites.

> Much depends on the design.  I believe my own design is immune to the
> network effect.  If it works as I hope, it will have almost the
> opposite tendency - a viral effect - a strong tendency to scale up.

This is a characteristic of the network effect.  Once it is started,
the value increases, so more join and the value increases even faster.

You just need that initial push to act as a spark and as time passes
the size of spark needed decreases.  Eventually, telephones become
inevitable.

>> Adb has said that even if you just have 1 proxy who has accepted you
>> as a client, you are better off as they find info for you.  This may
>> be true, but it doesn't look like it is enough to trigger mass use.
>>
>> What does a specific user get from joining the system?  You need a 'killer app'.
>
> If they get what democracy promises, and modern democracy denies -
> participation and freedom of choice - that would be enough for a
> killer app.

No, it wouldn't.

The world is made up of individuals, you need to show that individuals
benefit from joining.  If I don't join, then I get most of the
benefits and if it fails, I haven't wasted my time.

You need to show what is in it for potential members.  For the system
to grow, it has to create an incentive for people to join and also for
current members not to leave.

I am not sure if the proxy idea is actually able to do that.  However,
clearly, there hasn't been a spark.  Perhaps, it would grow and then
fizzle or perhaps it would grow and keep growing.  Nobody has tried
it, so it's not possible to be sure.

Anyway, there are 2 major issues to overcome for organisation of a
large group of people.  The first is rational ignorance and the second
is the free rider problem.

Delegable proxy has the potential to overcome both to a certain degree.

Free riding happens less often when there is a small group of people
negotiating and by compressing the large group of people into a small
number of sub-groups, free riding is decreased.

Similarly, the proxy/client relationship helps with rational ignorance
by allowing effective communication channels.  Also, if you can trust
your proxy, you don't need to understand everything.

>> You also need a zero effort way of joining the system.
>
> Yes.  (I'm about to start my usability trials.)

Great.  I think that it is an important issue.  The easier it is to
use, the smaller the initial spark needs to be.

> I just voted for you.  (I used a real email address, else you'd be
> unable to login.  You can blank out the registration later, or I'll
> blank it out for you.)
>
>  http://t.zelea.com:8080/v/w/RoleV/?v=mike%40zelea.com&s=ward-20-councillor
>
> You are currently the only candidate.  Here are the results:
>
>  http://t.zelea.com:8080/v/w/Count/?s=ward-20-councillor

Well, I am clearly not a citizen of 'ward-20' :).  Also, it does
highlight the issue.  There is little benefit to me personally from
participating.

OTOH, I am "wasting" my time posting to this list, so people will
spend some time discussing stuff that is interesting.

Your system is currently a pure vote tabulator, maybe a forum of some
kind would help (in fact even a comments section would be helpful).
Who are the candidates and what do they stand for.

In any case, if it were real, I would probably wait for some others to
get involved before voting and that is not a good sign.  With a forum
at least I could partially participate until I am ready to assign a
vote.  With your system, I don't leave any mark that shows to others
that somebody was looking at the site.

People go to discussion websites because others also go to them ...
again a networking effect.  Don't let users leave (or in our more
freedom inspired times, give them a reason to stay or at least check
the site every few days :) ).

I also notice that you have a doubting system.  Not sure how it works,
but I assume it is for sock puppets.

> As a delegate, you now have the equivalent of 2 votes (mine + yours).
> You can give them to anyone, without restriction.  As a voter, I am
> curious to see who you vote for.  I am expecting you to reveal
> information to me, just as Abd says.

Right, but currently there is no info on the topic at hand.

Ofc, this is true for all web-sites.  It is hard to get people to
actually join yours.  There are so many seeking people's time.

Again, after 'ignition', the proxy system can help with that.  Users
who want more clients could point out to friends about the site and
random users who register to post in a thread would see the interface.

> Do you see how this might be viral?  There are real freedoms and
> opportunities for participation.  I'm especially interested in the
> effect it will have on people when they learn that others have voted
> for them.  I think that's the key.

True, that would be interesting.  However, first, you need a reason
for the first few people to actually join and stay joined.

> So you distribute storage of the votes.  Interesting.  I distribute
> the legislative drafts (each drafter responsible for her own), but I
> store the votes centrally.

No, I was proposing distributing the storage of the proxy tables.
This means that the tree/web is not the property of any one person.

An admin having a bad day cannot just delete the database and break
all the links.

This also means the same tables can be used on multiple forums.

Now assuming there is a good connection between the proxies and the
clients, they can reform somewhere else.  However, they would probably
still lose lots of people.

Having it distributed means that this cannot happen and any breaks
will often just be one proxy 'disappearing', who can be routed around.

> Your idea is naturally a pull architecture, vs. push.  The talliers
> harvest the votes from the periphery of the network by polling it.
> Polling will lead to problems (traffic overhead and timing issues)
> especially for continuous elections.  The solutions will complicate
> the system, so it's probably not a good way to start.  But it's an
> interesting idea, maybe for later...

The votes would be held on the forum in question.  It could also
handle the tallying.

However, it would publish the list of usernames that have voted each way.

It would probably check the proxy tables only once every day and only
if they are actually in use.

> Central storage of votes is simpler.  I do distribute the storage
> sites (electoral servers) one per city or region.  Otherwise it would
> be hard to scale, and offensive to local sensibilities.

The problem is that (assuming it actually works :) ), once the system
is in place it is hard to change.

If you give power to certain groups of people, then they will resist
changing it to give more freedom.

However, you can add these rules in at the start, and there is no
interest group to stop you.

Also, you could implement the first version with everything handled
centrally, but still that the standard/rules requires that everything
is made public.

> I don't quite understand proxy lists.  There are no PR elections in my
> area.  Maybe you are suggesting that, with different strategies for
> tallying results (different voting mechanisms and eligibility
> criteria), it would be informative to compare the results
> side-by-side.  Yes, so ideally the system is open enough so people can
> define their own strategies and "roll" their own counts - all from the
> same raw electoral data.

Sorry, what I meant was that there can be issues with sock puppets.
This is in fact one of the main issues raised with the delegable proxy
system.  Someone could create lots of accounts and then have them all
direct votes to him.

You need some way to protect against it.

The solution I was thinking was that you could point at a proxy and in
effect say "I will trust anyone that person trusts as being a real
person".

That person would have done the same thing.  This creates links of
trust leading to top-level verifiers.  These top-level verifiers might
personally check or just randomly test lower level verifiers's
verifications.

In effect, to get a vote, you have to convince someone who is broadly
trusted that you are real.

This wouldn't need to be a yes/no type thing.  I might give a person a
trust level of 0.8.  That would mean that people verified by that
person get a vote of 0.8 as he tends to be reasonable in his
verification.  If you are verified by 2 people I have designated as
trusted, then you get a vote weight that it the higher of the 2.

This creates an inventive to keep your lists current, or you will lose
the trust of the community.

It could also be used to determine the 'quality' of the user.  For
example, if it was an activist organisation, a proxy might give 2
lists.  One list would be real residents of the area and the other
would be a list of people who have actually put the effort into
canvassing on some issue.  This also creates the incentive for people
to get involved, you 'canvasser rating' would increase if you spend a
few hours canvassing on some issue.

This would give multiple poll results.  One would total all the
usernames (some people might turn it off), another would total real
people and the final one would total canvassers.

I think that sock puppets aren't always binary.  If the group is
actually going to do something you might have a few levels of
participant.

1) fake accounts, i.e. pure sock puppets
2) vaguely interested people -> they might vote for the issue and
participate in the online discussion
3) interested people -> they might donate a few dollars
4) 'core' group -> these people will put in a lot of effort

It might be worth being able to split the users out into those kind of sections.

>
> For sock puppets, I have a trust network to authenticate the voter
> lists.  It effectively ties online identity to physical street
> addresses.  Any vote you cast in Toronto's elections won't count once
> the trust network kicks in.  (Meantime I've housed you in the
> University library.)

You can also use your delegable proxy network to do that.

You might for example be a high level trust proxy and go to the
trouble of verifying street addresses, then lots of people would link
to you.

You might handle say 20% of the verifications, but others could too.

> They had unlisted telephone numbers in the old days, but most people
> were "in the book".  Every couple of years they posted the voter lists
> around the neighbourhood.  They included everyone.  Nobody complained,
> except maybe if they were excluded.  I'm hoping that people haven't
> fundamentally changed.

True, but those lists weren't internet subject to internet data mining.