From what I was able to gather, they only ask a participant a single question in exchange to be able to access some premium content. Because it's just one question, they don't ask you demographic information. That what those "inferred age/gender" things are. Google gives a best guess as to what your demographics are based on other information they've collected about you.
The fact that its google's guess about demographics + only data from people actively seeking to avoid paying for premium content biases the results somewhat. I'm not entirely clear what the premium content is, but if the content consistently appeals to one type of person over another, it would definitely be a source of bias. Google's guess about your demographics probably just increases the error a little bit, but I'd be surprised if it consistently skewed the results in one direction or another.
So, not a perfect scientific survey, but miles above most web based surveys out there.
But intellectual concerns about accuracy aside, my initial emotional reaction is one of terror at google's apparent omniscience. I mean, I already knew they know a lot about us, but this is pretty visceral reminder, complete with very pretty charts.
The use that's presented here is fine. Data in aggregate is fine. But there are questions that advertisers could ask that would be much more valuable to them if they knew who answered which question what way. Google could make a ton of money passing on the emails of people who answered questions in a way that makes clear that they'd be likely potential customers. They'd make more money selling other information along side the emails. Information about things like political leanings, sexual orientation, health concerns, if not already in google's database couldn't be too far away. Target already has a system for determining if a woman is pregnant: http://www.nytimes.com/2012/02/19/magazine/shopping-habits.h...
Can google resist the temptation of selling that oh so valuable information? I really hope so.
It's pretty cool that this data can be so easily collected and shared. Some results:
- Gender: Women were less impressed by Ryan's speech
- Age: the 35-44 and 65+ segments really liked Ryan's speech; 18-34 were unimpressed.
- Geography: slightly better response in the South and West
- Urban density: Rural areas reacted much more positively than urban/suburban areas.
- Income: the poorest respondents (0-25K) actively disliked the speech, 75K-100K were unimpressed, and 25K-50K liked it.
Most of these results are not statistically significant according to the error bars, but the trends largely agree with my intuition. This kind of information would be well worth $20 (if it was $0.10 per response).
The survey asks the same wrong question as everyone else does: "What do you think other people's reaction to X will be?"
If it doesn't specifically ask the person of their own opinion, it's not going to be a useful survey.
However, we would not know about the shortcoming of the survey, if Google did not provide the exact phrase of the question, which I think is a huge advantage over many other surveys.
I don't really know or care about the relevance of these results, and I don't know why it's at all relevant to Hacker News, but damn, Google consumer surveys look really nice.
From what I was able to gather, they only ask a participant a single question in exchange to be able to access some premium content. Because it's just one question, they don't ask you demographic information. That what those "inferred age/gender" things are. Google gives a best guess as to what your demographics are based on other information they've collected about you.
The fact that its google's guess about demographics + only data from people actively seeking to avoid paying for premium content biases the results somewhat. I'm not entirely clear what the premium content is, but if the content consistently appeals to one type of person over another, it would definitely be a source of bias. Google's guess about your demographics probably just increases the error a little bit, but I'd be surprised if it consistently skewed the results in one direction or another.
So, not a perfect scientific survey, but miles above most web based surveys out there.
But intellectual concerns about accuracy aside, my initial emotional reaction is one of terror at google's apparent omniscience. I mean, I already knew they know a lot about us, but this is pretty visceral reminder, complete with very pretty charts.
The use that's presented here is fine. Data in aggregate is fine. But there are questions that advertisers could ask that would be much more valuable to them if they knew who answered which question what way. Google could make a ton of money passing on the emails of people who answered questions in a way that makes clear that they'd be likely potential customers. They'd make more money selling other information along side the emails. Information about things like political leanings, sexual orientation, health concerns, if not already in google's database couldn't be too far away. Target already has a system for determining if a woman is pregnant: http://www.nytimes.com/2012/02/19/magazine/shopping-habits.h...
Can google resist the temptation of selling that oh so valuable information? I really hope so.