Vilfredo presented in Milan

[Pietro Speroni di Fenizio]

Vilfredo will be present at the conference: "I codici della Democrazia". On Saturday afternoon I have been given 45 minutes to present the topic. It will be in Italian, and I will probably mostly use previous slides. But as we just had a big experiment with 30 people I think I should add something about it.

Also I am thinking of changing  a bit. I don't just want to explain Vilfredo. In fact it might make sense NOT to enter into the nitty gritty details. And instead give a different message. The conference will be about various platforms, and Vilfredo should try to come out as a specialised platform that operates for small groups and that is willing to interface itself with the others (once you have a particularly hard problem and 5-20 people interested in tackling it you send it to us.

Any thoughts will be appreciated. I would like to post the slides here before I go, but probably they will be ready 10 minutes before I am about to give the talk :-)

[Anthony Zacharzewski]

The thing that strikes me as a good use for Vilfredo is for preparing Bills for Parliament, where different amendments need to be considered on a particular topic. There could then be a series of Vilfredo votes, one on the structure of the Bill (list of clauses, showing the extent of the Bill) then for each individual Clause you could use Vilfredo to get to the best agreed text.

It would certainly be a better system than the serial amendments used in the British Parliament at the moment.

Anthony

[David Bovill]

On 10 September 2013 11:06, Pietro Speroni di Fenizio <pie...@gmail.com> wrote:

The conference will be about various platforms, and Vilfredo should try to come out as a specialised platform that operates for small groups and that is willing to interface itself with the others (once you have a particularly hard problem and 5-20 people interested in tackling it you send it to us.

Totally agree - along the lines we discussed I see other systems being used to select peers / peer juries, and Vilfredo being used by the jury to have quality discussions. This seems strong and  strategically wise, as the more we can do to get different small scale systems within the same ecosphere to synergise - the better chance we have of achieving scale.

I am not sure:

1. I understand fully your doubts about Vilfredo scaling
2. I agree that a Vilfredo like system can scale

However that does not change the fact that I think the strategy of defining the scope of application of Vlifredo to peer juries is not sound. I thin it would help if you took some time to fully spell out the reasons that you feel Vilfredo does not scale, and to do this in some (mathematical detail) along side an analysis that can be used for other systems and their ability to scale. Positioning your thinking clearly on solid thought through technical grounds, and raising the stakes for other systems in this area would be good for everyone.

PS - would like to talk about the Peoples Train on 29th of Septemeber (please give us a call / Skype / Hangout?)

[Pietro Speroni di Fenizio]

Hi Anthony,

the amendment system does not work. It can be proven mathematically. And you just need two amendments: A and B.

Having to Amendments brings out 4 options:

nA, nB (not A, not B)

A, nB, (only A)

B, nA, (only B)

A, B. (A and B)

Now if you ask people to order what do they prefer among those options you can make a graph showing which option beat which other one.

There are cases where the majority prefers A over every other combination. Which means that A is the Condorcet winner. And still by asking the amendments one after the other, you do not reach the A option.

Example if most people prefer

A-->AB--->B-->NULL

Then you start from Null and you ask: 

Shall we have B?

Most say yes. So now you have the B amendment active

Then you ask:

Shall we have A?

Most say yes. So now you have both amendments active.

Now you should ask if you should turn off the B, but the B has already been tested so you do not test it again, and you end up with a suboptimal solution. This is a case where the math is absolutely obvious and easy to see, and yet the political system ignores it.

So why don't we vote for all the options? Because they are two many. If you have 2 amendment, you have 4 options, but if you have 10 amendments you have 1024 options to vote among.

So a mixed system that given the basic amendments, each person can propose the combination he prefers, so we just pick out of all the options the ones we like. And then we try to approach through the generations a better solution COULD work. (And would surely work better, but that means better than something that does not work).

We should try to see if Vilfredo could be used on a piece of text. We would have to add a diff function, to check the differences. And I would suggest to start with something that does not let people work in parallel on too many paragraph. As I have no idea how what you do in one paragraph will affect in another. But maybe you could suggest it to be used in some specific instances.

Thanks for the idea, let's think about it.

Pietro

[Pietro Speroni di Fenizio]

Hi David, I like you ability to think out of the box.

I am very ignorant about juries. How many people are in a Jury? Also, don't Juries just have to give a yes no response? Because usually Vilfredo is used for open questions. You would have to untangle the yes/no answer to find out the why. And then ask that. I don't think that has ever been suggested.

Regarding the problem with scaling, is quite a basic math problem.

Each person can impose a proposal in the Pareto Front simply by voting just for that proposal. Not only that but even if they do not consciously  impose a proposal, often the effect of each person participating is to increase the size of the Pareto Front.

Also consider that the system works only if each person has read (and understood) each proposal. As such it cannot digest too many proposals at the same time. So you have a realistic possibility to have only up to 20-30 proposals. Maybe 40, if they are simple and repeat themselves.

If the number of people who vote are too many you will find that the Pareto Front end up having 20 or 30 proposals. In other words no filter has happened. We have tried (and partially succeeded) in lowering this number by asking people to explain why they vote against a proposal, and so people would vote in more similar ways. And by telling who are key players, so people can convince them to change their mind. But the success was partial. Some people changed their mind, but most didn't. And the PF remained big.

One of the experiment that I want to do, is a small python program to calculate the PF from a question with n proposals and m players that vote at random. And you would see how fast the PF grows as m grows. But the people do not vote at random, but still vote with enough diversity that the PF still grows as m grows. Not as much, but still grows.

Other system can grow much more because they sum the votes, so essentially they do not take into account people point of view. They accept that there will be a majority that will impose its will to the minority. And tough luck for the minority. The difference is fundamental, but since we have all been trained to consider that normal, Vilfredo does not seem so different... until you try it out. You really have to try it out for a few generations. Before seen how it is different.

(1) Said that I am working in an extension of the theory that would permit to people not to read all proposals.This is basically ready and just need to be implemented. But for now it will permit to people to say when they did not understand a proposal. For now this option is equivalent to voting against. But in the full extension it would not be so. 

(For the mathematicians among you, the trivial extension you can think of does not work because it does not maintain transitivity... Ignore this if you have no idea what I am speaking about.)

(2) Also another extension would permit to people to use a traditional voting at the end of a question. So when the situation is stuck people just start vote in the normal way and accept that Vilfredo only could help so far. This has been coded, and is in the test phase.

(3) Another extension considers how you don't really need a full consensus all the time. So the system would signal when some proposals have achieved the majority of supporters.

(4) Another extension asks everybody to rate each proposal in a scale from 1 to 10 (or better to a 100). And then decides where to put the threshold so that the PF is smaller. This is very interesting, and I am sure it has never been tried out before.

(5) Another extension starts to delete extra proposals chosing them from the least voted, but such that even if you delete them, no person finds himself without a proposal representing them.

As you can see this is a major major point of research for Vilfredo.

Cheers,

Pietro.

For the 29 ok. I am busy until the 27 in Germany. Then on the 28 I will probably fly back. So the 29 I should be near internet, probably in Portugal. Can you start another thread, explaining what is going on. Also the time and so on? Maybe other people might also like to join in, if you think they would be welcome :-)