In order for Google to predict the future millions of users would have to be accurately able to predict the future, and then use emoticons to reflect their feelings about the prediction.
Emoticons are often used to express feelings about past and present events. If they express feelings about future events, it is speculation. How would you filter out the noise of the past, present, and erroneous predictions of the future?
Cool concept but in practice you would need more information about who uses the emoticon (location, occupation, age, etc). And even then it still might not be feasible (a bunch of people in the financial sector, aged 25-40, NYC sent frowny faces...now what?)
Reminds me of information (prediction) markets where people purchase positions on future events. Turns out it is a more efficient way to tease information out of a group than polling. The ideas is that when there is money at stake people's predictions magically become more accurate.
You bring up some valid points, but I would postulate that people talk about the recent past (i.e. events that affect the future), the present, and the predicted future, far more than they talk about far-past events... especially so if they're tying emotions into this discussion.
The geo-centricity of the data is definitely a problem. Forest fires in California would certainly incur a lot of grief emoticons from that area, and if the algorithm were not geosensitive it would presume there was a spike in global sadness, which would not be the case.
You guys!!!!!!!!!!!!!!!!! Come on! Surely startups in the present economy would be aware that predicting the future has been exactly the problem that got us into this mess. If you didn't study it yet, get familiar with the history of the Black-Scholes model. The original quants behind Long Term Capital Management, and the quants today use the same math. Yes, it won the Nobel Prize in economics, but nobody said economists were brilliant mathemeticians (quite the contrary actually. In fact, the poor/dishonest reasoning of economists was the whole motivation behind the Post-Autistic Economics Review).
The Black-Scholes is based on a calculus used to by rockets to predict their navigational path through space while seeking their targets. Any rocket scientist can tell you, it doesn't work out all the time. Continuous calculuses don't respond well to unplanned discontinuities. So you might get away with it for a little while, but eventually you'll be wrong. And the consequences will almost certainly be catastrophic.
Using emoticons can give results with an error percentage of around... 99.99% :) and the result maybe just 0.01% right.
Coz, as kwamenum86 said, these icons are used to express personal feelings about an incident or news. My aunt might be in the ICU and when such a news is heard I give a sad emoticon. And that can't be counted when you want to decide if the world will be happy(not the whole world cares about my aunt). And at the same time I might be happy when a guy in my class scored less than me. He would be sad and I would be happy when we express this in emotes. (I'm not this bad).
So if these emotes are to be really considered to count for the world's future happiness measure, then the text along with the emote also has to be parsed using Natural Language Processing to determine if its just a feeling about an incident or news that matters personally or the other way.
Emoticons are less intrusive method of aggregating the public sentiment. Would you want the contents of you emails analyzed to help predict the future?
No, it's not less intrusive. Emoticons are essentially words like any others. Grabbing the emoticon words from writing (public or private) is no different than sampling other classes of words, such as all adjectives or verbs.
And no, I would not volunteer any part of my private communications for a third party to analyze for their own benefit.
It is a little different. The amount of information you garner from ":)" versus "Gee, I think I am going to liquidate now" are completely different. It is a matter of sentiment analysis versus content analysis. But I see why you say "it's not less intrusive". I think sentiment analysis is less intrusive, especially if you are just looking at emoticons, because they reveal less about you. The information encoded in the emoticon has less clarity. It's reasonable that you might find any amount of intrusion unacceptable though.
Doesn't Google look at the contents of your (subject being gmail users) email anyway to serve ads on Gmail right now though? The scary thing is I think people would just get used to it if it were useful.
Of course I think (hope) we are all engaging in this conversation assuming this will never happen.
One crude way for them to measure this right now is looking at the number of requests that are made for the emoticon image. There are some problems there (caching, does emoticon sent or emoticon received, how often do you count a request per ip)...far-fetched...but as long as we are dealing with a hypothetical situation...
Emoticons are often used to express feelings about past and present events. If they express feelings about future events, it is speculation. How would you filter out the noise of the past, present, and erroneous predictions of the future?
Cool concept but in practice you would need more information about who uses the emoticon (location, occupation, age, etc). And even then it still might not be feasible (a bunch of people in the financial sector, aged 25-40, NYC sent frowny faces...now what?)
Reminds me of information (prediction) markets where people purchase positions on future events. Turns out it is a more efficient way to tease information out of a group than polling. The ideas is that when there is money at stake people's predictions magically become more accurate.