Just how much does google know about you?

[GreyICE]

The problem is that you're both hitting the same woo-woo roadblock: You think you're a beautiful and unique snowflake.

You're not. Your behaviors can be predicted by statistics, because you are similar to people who are similar to you. Give me a little information about who you are, and a bit of information on how people behave in a situation, and I have a poor idea of what you will do, represented statistically. Refine the detail I have on you and the detail I have on the general population, and I have a better picture. As the detail further refines, I receive an even better idea of what you'll do.

Is it perfect? Nah. Sometimes, someone does do something spontaneous. Maybe even 5-10% of the time. But having 95% information 95% of the time is really quite good enough.

Your entire argument against this is that people behave unpredictably. Yet Fiona has already agreed they're predictable, and Ludwig doesn't seem to have a problem with that particular concept either.

So what are we left with? Our limitations are not significantly the unpredictability of humanity. Nor are they processing power. Nor are they limits on information gathering.

What are they, programmer stupidity?

 

[Ladewig]

The problem is that you're both hitting the same woo-woo roadblock:

Such name calling is unnecessary, please stop.

You're not. Your behaviors can be predicted by statistics, because you are similar to people who are similar to you. Give me a little information about who you are, and a bit of information on how people behave in a situation, and I have a poor idea of what you will do, represented statistically. Refine the detail I have on you and the detail I have on the general population, and I have a better picture. As the detail further refines, I receive an even better idea of what you'll do.

Is it perfect? Nah. Sometimes, someone does do something spontaneous. Maybe even 5-10% of the time. But having 95% information 95% of the time is really quite good enough.

Your entire argument against this is that people behave unpredictably. Yet Fiona has already agreed they're predictable, and Ludwig doesn't seem to have a problem with that particular concept either.

The name is Ladewig. Actually, I do have a problem with it, but I may be able to overcome my problem if you answer the question I asked in my last post:


Give me an example of what actionable information this computer can generate on these 3000 individuals. What exactly is the candidate going to do to change these people from the likely-to-vote category into the unlikely-to-vote category?

 

[Ambrosia]

The problem is that you're both hitting the same woo-woo roadblock: You think you're a beautiful and unique snowflake.

You're not. Your behaviors can be predicted by statistics, because you are similar to people who are similar to you.

Asimov figured this out in 1942.

I *AM* a unique and beautiful snowflake, there is noone else like me and there never will be. Your analogy breaks down a little because despite all snowflakes being unique and pulchritudinous they all do the same thing. Namely fall from the sky and land on stuff.

Supermarkets dont give away 1% of their takings for no good reason either, they are buying a detailed record of all of your shopping habits to better target their advertising and offers.

I agree that information is power and we are fast reaching the point where less scrupulous people might well use said information about us for nefarious purpose.

I will be very very disappointed if in 2025 we all still vote by trekking to a polling station and scratching an X on a piece of paper. Surely we will be voting over the internet by then.

 

[GreyICE]

If you think I am calling you a name, please report me to the moderators as breaching rule 12. If I am not breaching rule 12, please do not refer to me as doing so. I think even Locknar would have problems saying I was doing that here.

The behavior I am describing remains as I described it - woo-woo. You have an idea that human motivations are irreducibly complex, that we must analyze each individual as unique and different from each other individual, sharing only negligible similarities in our own unique journey.

We're not, we're a product of statistics.


ETA: I'm quoting 'Fight Club' on the snowflake line, as that's the movie that kicked that notion right in the crotch.

 

[Fiona]

The problem is that you're both hitting the same woo-woo roadblock: You think you're a beautiful and unique snowflake.

Er....no.

You're not. Your behaviors can be predicted by statistics, because you are similar to people who are similar to you. Give me a little information about who you are, and a bit of information on how people behave in a situation, and I have a poor idea of what you will do, represented statistically. Refine the detail I have on you and the detail I have on the general population, and I have a better picture. As the detail further refines, I receive an even better idea of what you'll do.

Er....no.

Statistics exists because it is not possible to predict individual behaviour. That is the basic flaw in your argument. Think of it this way: epidemiology can tell us that a person is more likely to get cancer if they smoke. It can tell us that you are more likely to get cancer if you work with asbestos. etc. We can give general advice on that basis: and we can say the Mr X, who smokes and works with asbestos is more likely to get cancer than Mr Y who does neither of these things. When Mr Y goes to the doctor I do not expect the doctor to rely on epidemiology and tell him he does not have cancer: because in fact Mr Y does have cancer. And so it goes. 95% of the information is not good enough and I do not think it ever will be. I could be wrong about that but I don't see any evidence to suggest I am so far

Your entire argument against this is that people behave unpredictably. Yet Fiona has already agreed they're predictable, and Ludwig doesn't seem to have a problem with that particular concept either.

No, it doesn't. There is a lot of variability within the general predictability and that is what statistics smooths out: but it cannot tell us what the individual will do, even in principle: and it is particularly subject to the influence of the investigation itself, too.

You want to believe otherwise I cannot stop you: but unless you can show me that human variability is not less than you imagine I believe it to be but is in fact non-existent then I will stick with a doctor who diagnoses ME and not my demographic

 

[Ambrosia]

ETA: I'm quoting 'Fight Club' on the snowflake line, as that's the movie that kicked that notion right in the crotch.

A fictional account of a mentally ill office worker whos alter ego is a means to escape the drudge his life has become. Who starts his own cult of similarly disenfranchised rat racers, and uses woo and dogmatic ideology to inspire them to cause anarchy.

hmmmm.

I'm still very beautiful and completely unique.

 

[GreyICE]

Er....no.



Er....no.

Statistics exists because it is not possible to predict individual behaviour. That is the basic flaw in your argument. Think of it this way: epidemiology can tell us that a person is more likely to get cancer if they smoke. It can tell us that you are more likely to get cancer if you work with asbestos. etc. We can give general advice on that basis: and we can say the Mr X, who smokes and works with asbestos is more likely to get cancer than Mr Y who does neither of these things. When Mr Y goes to the doctor I do not expect the doctor to rely on epidemiology and tell him he does not have cancer: because in fact Mr Y does have cancer. And so it goes. 95% of the information is not good enough and I do not think it ever will be. I could be wrong about that but I don't see any evidence to suggest I am so far

To quote you, 'Um... no.'

The flaw in your argument is that our knowledge of cancer is at best laughable. We have a couple of things that sometimes cause cancer, and we know them, we call them risk factors. But the thing is that even when we have a lung cancer sitting on a table right in front of us, having been torn out of the patient's body only seconds before, not a single thing we can do would tell us why it happened.

In contrast, when someone votes for McCain or Obama, we can develop a reasonably good idea of why they voted the way they did.

In terms of biology, if you handed an expert on knee replacement surgery information that the person was a runner, that's one data point. That the person regularly ran marathons, another. A schedule of the person's running career and practice times, the person's weight at all points during their life (and their related activity level), a 15 year old x-ray of their kneecaps and reports on complaints about knee pain, and the doctor could not only give you a damn good idea if the person will need replacement knees or not, they probably also can tell you what sort of conditions the person's knee will be in.

No, it doesn't. There is a lot of variability within the general predictability and that is what statistics smooths out: but it cannot tell us what the individual will do, even in principle: and it is particularly subject to the influence of the investigation itself, too.

You want to believe otherwise I cannot stop you: but unless you can show me that human variability is not less than you imagine I believe it to be but is in fact non-existent then I will stick with a doctor who diagnoses ME and not my demographic

As Ambrosia noted, companies do not give you 1% discounts on everything you buy because they are sweethearts who really like you. Google did not obtain the fortune they did by having an algorithm that a kid playing 'pin the tail on the donkey' could replicate.

The thing is, we are already deeply involved in modeling human decision making and cognitive processes, to suddenly state that the entire field is impossible is quite a stretch.

It is, of course, not entirely comfortable. One may note 'comfortable' and 'possible' are two different concepts.

 

[Ladewig]

If you think I am calling you a name, please report me to the moderators as breaching rule 12. If I am not breaching rule 12, please do not refer to me as doing so. I think even Locknar would have problems saying I was doing that here.

The behavior I am describing remains as I described it - woo-woo. You have an idea that human motivations are irreducibly complex, that we must analyze each individual as unique and different from each other individual, sharing only negligible similarities in our own unique journey.

My point is that motivations for voting really are irreducibly complex when all you have is one data point every four years. But rather than argue about that, I'll ask my question a third time:


Give me an example of what actionable information this computer can generate on these 3000 individuals. How exactly is the candidate's party going to spend X-thousand dollars to change these people from the likely-to-vote category into the unlikely-to-vote category?

 

[GreyICE]

My point is that motivations for voting really are irreducibly complex when all you have is one data point every four years. But rather than argue about that, I'll ask my question a third time:

And, finally, we are done. The argument stands here - you do not believe your logic for voting is in any way similar to anyone else's - so your actions will not in any way resemble theirs, even if you share motives and beliefs. You are a beautiful and unique snowflake, we must use no others when considering your beautiful uniqueness. Only your data point may reflect on your data point.

Give me an example of what actionable information this computer can generate on these 3000 individuals. How exactly is the candidate's party going to spend X-thousand dollars to change these people from the likely-to-vote category into the unlikely-to-vote category?

Here, let me give you a real world example:
http://archive.salon.com/politics/feature/2000/12/04/voter_file/

Using the most painfully sloppy of demographics (skin color), and the most heavy-handed tool (voter registration denial), this was already done. We're not talking about science fiction here, we're talking about last decade.

Of course refinements to the information available produce refinements to the methodology.

 

[Ladewig]

And, finally, we are done. The argument stands here - you do not believe your logic for voting is in any way similar to anyone else's - so your actions will not in any way resemble theirs, even if you share motives and beliefs. You are a beautiful and unique snowflake, we must use no others when considering your beautiful uniqueness. Only your data point may reflect on your data point.

I want to admit I made several mistakes. I did not understand the snowflake argument you were making until last night. I have NEVER suggested that in a country of approximately 210,000,000 voters there could possibly be 210,000,000 different motivations. The largest number that appears in any of my posts is 1100 different motivations (in post #37). My first mistake was not making it more clear that I did not hold the position that every voter held a unique voting motivation.

My second mistake was overestimating the number that I actually gave. Eleven hundred, I now believe, is too big. I would say that the vast, vast majority of U.S. voters hold one or more of maybe 100 different motivations for voting or not voting in any particular election. My third mistake was dividing that number into two groups: trackable and untrackable. I now believe that these motivations can be placed on a spectrum where one end represents easily trackable motivations and the other represents more difficult to track motivations. In the hypothetical in post #23, GreyICE says that "information gathering, [allows] the political opinion of most of the inhabitants of the city, [to be known with] 95% certainty." That claim sounds reasonable to me so I will not debate it at all (in fact, if "most" were changed to "almost all" I still might not challenge it). However, it is important to note that it is not 100% of the electorate that can be predicted to a 100% level. Even 15 years from now, there will still be some variability in both the total number of voters classified and the confidence level associated with the prediction. I also want to note that at this point we are talking about the political opinions (e.g. I like moderate Democrats) of specific members of electorate, not the voting motivations (e.g I will turn out only if the pre-election polls show a difference of less than 5 points) of specific members of the electorate.

I believe that predicting these voting motivations is much harder than predicting political opinions. I think that in 2025, in a city the size of Harrisburg, in a mayoral election, one might get the clear majority of the electorate correctly predicted at a very fine level (e.g. approximately 76 of the 120 Democratic voters between 20 and 30 who earn less than $30,000 will turn out given the current conditions of the political race). But I also believe predicting exactly which 76 voters will show up (at a 100% confidence level) will be beyond the technology available in 2025. Maybe 50-80% of them can be correctly predicted, but I cannot believe that all 100% of them can be identified before the election. I base this belief on what I mentioned in the previous paragraph - some traits associated with voting motivations are very hard to track. I could go into greater detail here, but will only do so if asked.

My next mistake. I asked how 3000 voters could be moved from likely-to-vote to unlikely-to-vote by spending money. You responded by giving me an example in a presidential election. I will heartily concede the point that dropping one one-hundredth of one percent of the actual voters can be done in a federal election.¹ But I am more interested in the hypothetical. In the 2025 mayoral race in Harrisburg, the numbers you cited indicated that preventing as few as 1050 voters (7%) would change the results of the election. Specifically:

I can absolutely locate the EXACT PEOPLE in the 15,000 voters necessary to swing the election, and carefully demographically target the opposition voters with various information-related measures to disrupt their lives so they can't vote.

So I will ask that question now. How will you use the computer-generated results to prevent these specific 1050 voters from voting in this election? Obviously, you cannot falsely accuse them all of being felons. Without breaking the law, how will you "disrupt their lives so that can't vote."

So that's my question. In the hypothetical that you laid out, how will you do it?
_ _ _ _ _ _ _ _ _ _

Here, let me give you a real world example:
http://archive.salon.com/politics/feature/2000/12/04/voter_file/

Using the most painfully sloppy of demographics (skin color), and the most heavy-handed tool (voter registration denial), this was already done. We're not talking about science fiction here, we're talking about last decade.

Of course refinements to the information available produce refinements to the methodology.

Lastly, I want to take issue with your final claim. How will future technology that will allow for better tracking of people allow unscrupulous registrars to maliciously pretend that voters are more untrackable to the point that they can be easily confused?



. . . . . . . . .
1) the example works only in one direction. Republicans used the technique to drop Democrats from the rolls. I doubt it could be used in the other direction.