CorrelatingPolls

The basic idea is, do a lot of polls. On various things, almost anything can be useful - who you'd vote for in next election, what's your personality type, if you like science fiction, how often do you watch TV, etc. But then, you check for correlation between the results. That means you have a common identifier between different polls.

This is a statistician or data miner's dream, but until the internet it seemed impossible. Oh, you can do polls to try to correlate say, political allegeance and judgement on current events, but you need to start anew each time.

But with the net, things change. How would this be done ? I imagine something like FOAF - people have profiles on their page, that can record things about them. Like what books they buy at amazon, what reviews they give to movies, books, websites, politicians, songs or games. What results they get on those "What kind of movie robot are you ?" or "which spice girl are you ?" online tests. What blogs you read. Anything, they don't even have to be honest. The point is simply to have things together in a machine-readable format.

This is a lot like collaborative movie rating system, only extended to everything. The more data, the more correlations to check for.

Things to poll for

Benefits

I see mainly two :

This could also be a form of DebateTool - MeaningMap? maps debate not through arguments but simply by trying to put in a SubjectSpace various opinions. It helps situate the debate.

problems for people being polled

Privacy, mainly. But then, like blogging, this is voluntary disclosure of private stuff - so it's up to people to decide what they want their employers to be able to find out.

One possibility is to use a pseudonym that only you know, that can be entirely seperate from your blog or anything that relates to you.

problems for people analyzing the data

Trust. How do you prevent a movie company from making 10 000 false profiles ? (An analysis of algorithms used by collaborative movie rating system shows them to be pretty vulnerable to this - though manual fixing can work in their case). Having it linked to your blog / homepage is a pretty good indicator that your profile is for real.

problems for both

how to handle simple mistakes ? How do we even discover that mistakes are being made ?

implementation

I'm not really advocating implementing this, but I think something like this will happen, under one form or another. Distributed or centralized ?

Would this be plugged int blogs ? SocialSoftware (things like Orkut) ? Wikis ?

See also RatingAsContent - this page tries to show how rating can be extremely valuable content, since it scales much more easily than comments and reviews.

Discussion

Anyway, that's how I saw this idea, I'm sure it's floating around out there.

I think so. Saw the poll I created on php-wiki/demo: poll-person-contribution? It's not too difficult. Just a little ;).

Hopefully now, about a year after the slider-soaked "idea for a better use of human intelligence" (didn't know about the existence of wikis when I wrote it in July 03) wild-polling will become the new [WikiFashion? wiki-fashion]. Jouw!

This is a great idea- I can easily envision poll files, and then accompanying encodings of how you answered the poll.

You go to a website, answer a poll, and it spits out a file for you, or offers to host it for you.

The web of trust could be tied to the web of trust that FOAF will have. (Or, wherever it ends up appearing first.)

You would edit your FOAF file, and say: "Here's a poll I answered," pointing to the file.

I don't think that just poll's in the wild would have any value, once it becomes slightly popular. You'd need Internet Bonded (InternetBonding) authentication, eventually.

I added of lists of things I think would be interesting. Actually I'd like to see a graph of those for any online community I participate in :)

Having these kind of things explicit may increase partisanship and division, but I believe it's more likely to make one acknowledge the immense diversity of people. Plus, correlations between things you chose (stand on issues, political affiliation) and things you don't (country, personality type, age) may make it appear that choices that seem to be the product of a well-thought weighting of possibilities are mere products of one's environment.

Yes, I would find the results of such a survey mildly interesting. But I worry that people will use this information for nefarious purposes.

Or (just as bad) well-intentioned people will do bad things because of a mis-interpretation of the data. If we discover that 100% of convicted car thieves drank Mountain Dew less than 12 hours before of the crime, doesn't it seem reasonable to handcuff all Mountain Dew drinkers for 12 hours every time they consume the beverage ?

BookShelved:DavidBrin (in TransparentSociety) has convinced me that massive data collection is inevitable. Trying to stop it is futile. But it would be good to try to forsee the worst problems, and try to find work-arounds for them. And also try to make sure the benefits outweight the problems.

After reading SubjectSpace, I wonder: Perhaps we shouldn't treating people as being "at" a fixed place in this map, with various polls only refining the precise location. Maybe it would be more interesting to see how people drift/teleport from one location to another. Religious conversions and political re-affiliations and marriage seem to involve many, many changes in a persons outlook – "he is a new man". Other things – going to college, the first trip to a different country, learning a new language, etc. all seem to "change a person". If I'm going to give gifts to my future self, maybe it would be good to find out a little bit about what sorts of things he likes – which might follow a common pattern of changes from the things I like.

I don't think you really need to identify individual people and force them to reveal every scrap of information about themselves. If you only want to know if A is correlated with B or C, then you only need to ask about A, B, and C. If you find that they are un-correlated, and then a different survey (possibly with completely different people) finds out that D is very strongly correlated with A, then you can draw the conclusion that D is (probably) not correlated with B or C. No need to ask about A, B, C, D, … Z all on one survey.

DavidCary

There is complete relativity to poll results. They do in fact not tell you anything. There is no way to say where is up and where is down, where is left and where is right. What only tells you something is to see how a result, a parameter changed over time. Comparing the maps over time tells you something about the hive-mind. That means play serious, pretend serious even when you know it can not be serious for the moment. Pretend serious as you know when comparing the results over time you can learn to understand how the hive-mind ticks. I don't say its easy. It's higher hive-mind psychology, a still rather young science.
More: it is us.
I think we are worth it.

To get into experimenting we need a plugin for pages like php-wiki/demo: poll/ person - contribution into odd-muse like it exists for crao-wiki: poll/ person - contibution (for some reason I have to reload this page in firefox to see it with php-wiki/demo embedded).

Already been done?


See SubjectSpace, Wiki:SlightlySpookyExperienceWithAmazon, Deme .

Define external redirect: WikiFashion MeaningMap

EditNearLinks: RatingAsContent

Languages: