Milan Event Report

[Pietro Speroni aka Vilfredo]

We organized a small pilot in Milan, trying to see if Vilfredo could
be used for political organisation among activists.
The event was developed through the MoVimento 5 stelle. The local
group from Milan that has been inspired by the italian comedian and
top blogger Beppe Grillo. The question was selected using Vilfredo
itself, and it was (the english translation follows):
"In molti hanno la sensazione che la democrazia settecentesca, nella
quale si eleggono rappresentanti che ti rappresentano per quattro o
cinque anni, non sia più sufficiente. La gente oggi ha i mezzi per
essere informata, per esprimere la propria opinione, per contare di
più. Usando gli strumenti che già sono stati sviluppati, come si puo'
realizzare concretamente l'idea della democrazia consentendo la
partecipazione dei cittadini senza cadere nel caos ?"
Translates as:
"in many have the sensation that democracy as it has been conceived in
the 18century, in which we elect representatives for 4 or 5 years is
not going to be enough anymore. People today have the tools to be
informed, to express their opinion, to be more relevant. Using
instruments that have already been developed, how can we concretely
realize the idea of Democracy, letting in citizen participations,
without falling into caos?"

About 15 people came to the event. Milan was under a huge rain, and
many people sent mails saying that they could not come due to the
weather, and how blocked Milan traffic was. Also there were some other
problems, from the internet that was quite unstable, and also some
bugs of Vilfredo that came up just in that occasion. Part of this was
because we worked hard to extend the funtionalities of Vilfredo to let
multiple people vote and participate anonymously (so they could share
a computer), and made an Italian version of the website.

Focusing more on the question, I would say many lessons were learned.
What became obvious in the course of the evening was a series of
weakness of Vilfredo, as well as specific things on which Vilfredo was
really good, as opposed to others.

The first thing that was clear was that there are two different kind
of brainstorming that are involved depending if we want to brainstorm
a single solutions to a specific question, or multiple solutions that
all can apply. When a question is very vague, or involves something
that requires many distinct actions, we are more in the second case.
When we need to find a single solution to the problem, we are more in
the first case. Vilfredo supposedly works well in this first case.
This is because people are invited to integrate different solutions.
Also the Pareto Front filter tends to eliminate some of the proposals,
and although they are dominated, the dominating really is logically
better only if we want a single solution. Suppose A dominates B, then
we would supposedly take off B because everubody who voted for B are
also happy with A. But in the multiple solution model, many of the
people who voted for B would also be happy with A+B (or A and B). In a
sense we would have to add the solution A-and-B and let that run by
itself.

In theory when someone who did not vote for either A or B was happy to
have A and B, then A would also dominate a proposal that includes both
A and B (A-and-B). In practice when in a generation two proposals made
it into the pareto front, and both A and B made it to the next
generation, then A-and-B would be presented along side A and B. In
those cases there often were people who would vote for that instead of
voting for either A or B. With the idea that the combined effect of
the two was better than a single one. But when proposals are
independent, the filtering out pareto front system does not take this
into account, with the result that proposals are filtered out too
soon.

To do it in another way we would havee to go through all the couple of
proposals, and see which can be simply connected. Then let the system
build all the options and let people vote them. This is impossible due
to the combinatorial explosion of the possibilities. If we have 5
options, we would have (2^5)-1=31 options to vote on, and with only 10
options the voting possibilities would grow to a staggering 1023. It
might be done using a more complex voting system, but then it would
become very hard for the voters to use it, indeed.

So maybe in the multiple proposal situation a different kind of
algorithm should be applied. Something that does not filter out the
proposals, but ranks all of them in a fair way. And then the best ones
are done.

On the other hand if the problem is a on the construction of a single
solution, that can be made of several parts, but the parts are not
independent, then Vilfredo works very well. The Wall-of-text case is
the perfect example, and now it is becoming clear why it went so well.
Yes there were multiple ideas, but it is also not trivial the way they
needed to be integrated. So people needed to decide the elements in
the recepy of the solution, and then decide how to put them together.
Note also how all the metaquestions. The question about a question:
"what question should we ask…", all fall in this category. And in fact
they always worked well.

And then there is also the case where we need a single solution, and
no great combination of basic components. Those situation is not
really an open question, and if it is possible the best way is to use
some form of condorcet voting. Maybe after having permitted everybody
to suggest their idea. You still can work with a mixed system that
first selects the pareto front, and then let everybody order the
proposals in the pareto front. A two layer system would work fine
here.

The second element that was really important was the realization that
it is not enough that people read all the proposals. They need to do
it with enough time that they can really understand them. Then they
really vote based on their value. And that is when the theorem that
states that the Pareto Front would not explode works. If people do not
have the time to understand them all, but the proposals are all easy
enough to understand, then people will randomy pick and chose some of
them. As the voting becomes random, the pareto front grows.

A different case is if proposals are not all easy to understand. In
this case giving more time to people will still help, because people
will not vote for the "too complex to understand" proposals. And since
this will tend to be consistent, it will happen across the board. And
the proposals that are too hard to understand will be filtered out.
And so the pareto front in that case will not grow.

So an event where people are pushed to propose and then vote all in
little time does not naturally bring few proposals in the pareto
front. Because the filtering did not work out well. But then everybody
will have to work much more in trying to integrate and find compromise
among a much wider rose of proposals. So it pays to give more time to
the voting process, or at least as much time in the voting process
than in the proposing process. This showed that Vilfredo really works
better in an online setting, with people having some time to read
through the various options.

Another thing that was really good were the key players. The fact of
saying to some people what we wished them to work on, really helped
focus the system. This was even too good, to the point that people who
were not key players started wondering, "hey, I am not a key player, I
feel left our, what should _I_ work on?".

A troubling thought is also the size of the group. It looks like 10-15
people already start to reach the limit of the system. If this is the
case this would not pose a significance difference respect to the fact
that 10-15 people around the table can in general brainstorm an idea.
Of course the big difference is that the participation around a table
is far from fair, and some people (often males) tend to dominate the
discussion.

Comments received: "there is not enough time to think the proposals
through"
"this would work very well among a group of experts"
"can we have thematic rooms to discuss each theme?" (yes!)

Everybody also agreed that the system need to work on the design, but
before that on the "user interface". Also there was a general need for
some explenations, so much that I had to go from person to person to
make sure that everybody has understood what they were supposed to be
doing at each phase.

Note: Maybe imposing smaller size to the proposals, would have worked,
because then people would not have the space to beat around the bush
and would just write what they think. But then it would make it hard
for people to integrate different ideas.

All together the experiment was successful in the sense that it
presented the tool to several people, we learned a lot about the
possible uses, pro and vices of it. But it might be still some time
before Vilfredo takes over as the strategy for discussions around the
world :-).