Not Sure How You Feel Today? Ask the Facebook National Happiness Index
Posted 10/12/09 at 06:37:09 PM by Jason Barry
Comments
Print
Email
Adam Kramer has been working on a project at Facebook aggregating 100 million users’ status updates into a database and parsing it for positive and negative words. When you map this data over a timeline spanning a couple years, what do you have? The Facebook United States Gross National Happiness Index.
They have taken precautions so no one’s privacy is in trouble, but they tally a score each day based upon the status updates’ positive and negative emotion words. Some of the conclusions are obvious and expected: people are much happier (9.7% happier) on Friday than Monday—the saddest day of the week. Further, according to the study, two of the saddest days of the year were the days when Heath Ledger and Michael Jackson died.
The other fairly common spikes fell around major U.S. holidays: Thanksgiving, Christmas, and the fourth of July.
You can check out the index yourself over on the Facebook site. How accurate do you think this type of “polling” can be, and do you think its findings are credible?
app deleted.
Submitted by bathtbgin on Mon, 10/12/2009 - 7:39pm
Thanks for the article, if it were not for MPC I would not have been aware that this app had been added to my profile without my knowledge.
Did you read the article or
Submitted by Ntldr on Tue, 10/13/2009 - 6:02am
Did you read the article or just comment. They didn't say anything about an app being added, they are looking at status messages and having a database program doing a function checking for specific things.
Submitted by bathtbgin on Tue, 10/13/2009 - 9:32am
I read the article, checked out the link and then went and had another look at the app privacy settings on my FB account. Thats when i noticed the app for this "study" had been added to my account without my knowledge.
THIS is science?
Submitted by 1337Goose on Mon, 10/12/2009 - 3:56pm
"I'm not very happy today"
That had the word "Happy" in it, does it register as being a happy day for me?
This can't be accurate enough to actually mean anything.
~Goose
Submitted by aldude505 on Mon, 10/12/2009 - 9:17pm
I would say that since you have 'not' and 'happy' they would count that as a negative... and this really isn't that complex of code either you would just have to parse the string (your status) looking for a space as a delimiter and then looking for key words... and either way its just cool to see that around holidays and big events that this algorithm produces happier results.. makes me believe there is some truth to it
Submitted by 1337Goose on Tue, 10/13/2009 - 3:45pm
Okay, so I gave the most trivial example, but natural language processing is a heavily researched yet surprisingly enigmatic area of artificial intelligence study.
"I am not happy today" may be easy to process using a simple conditional, but I think it's being a little naive to assume that all your data will be in simplistic English. In fact, I think that people try to be as witty as possible with their status updates, referencing pop culture and current events where possible.
Consider the following example: "I'm not as happy as a kid in a candy store, but I sure am close!"
How would the computer interpret that? How happy is a kid in a candy store? How close is "sure am close"? And lastly, should this be interpreted as "not happy"? Am I nitpicking? Perhaps, but I thought it'd be a good topic for potential discussion...
~Goose
Submitted by JCBarry on Mon, 10/12/2009 - 4:10pm
I'd say that it's more specifically psychology and statistics. In both fields you can massage data into telling the story you want it to. However, I think it is interesting what people are doing with all of this data we've never had access to so freely.
Jay Barry
MaxPC Profile | Jason Barry
Submitted by 1337Goose on Mon, 10/12/2009 - 4:26pm
Okay fine, I guess it is kind of interesting to see this data, but I don't really think it has any meaning. There are too many potential flaws.
If I'd had designed this experiment, I would have chose twitter as a platform, and I would have checked for all the happy and sad words associated with specific topics.
You know? I type in "Windows 7" and see how many happy words come out versus how many sad words come out. That could almost qualify as market research. This facebook thing almost seems like a trivial pursuit to me.
On a side note, I thought bloggers could delete their double posts? From what I hear, you guys can edit and delete posts left right and center (and post images in comments?!) Also, I think I remember Paul Lilly saying something about getting private jets?
~Goose
Submitted by Muerte on Tue, 10/13/2009 - 6:11am
The people who legitimately do these things for study already think of these things. Word parsers have been around since word processors were developed. There have programs for years that mimic a person conversing with the person at the keyboard most convincingly. and this was 20 years ago.
The same kind of word parser that can do these things would be very effective in parsing phrases and tell a happy statement from a sad statement.
So if done properly they can eliminate most if not all of the false positives/negatives.
Submitted by 1337Goose on Tue, 10/13/2009 - 4:00pm
I understand where you're coming from, and the technology is quite good, but it's also far from perfect.
You mentioned that computers have been able to mimic human conversations over a terminal "most convincingly" although it should be noted that most these machines were given simplistic hypothetical situations (such as a patient speaking with a psychiatrist).
Also, I doubt that the organizers of this experiment thought to use an industrial strength natural language processor to analyze their results, but of course I'm speculating because I haven't seen the procedure of this experiment. However, I'm basing my speculations on the fact that the experiment is designed to determine that "National Happiness Index" based on a sample from Facebook.
Also, I posted a comment above describing how references to pop culture and figurative language could also trip up the machine. It's key to remember that some of the original AI systems designed to pass the Turing test would be confused by statements like "I love mother nature". These original machines would assume that mother nature was some type of maternal figure. How big of an influence could that even have on the results? Well, considering that most people do make references to pop culture on their Facebook pages, I think it would be significant.
To conclude, I think I'd have to stick by my original point, this experiment wasn't rigorous science, it wasn't even science in general. It was mostly just a "Hey guys, wouldn't it be cool if we could get a gist of when people are the happiest and when people are the saddest?" And yes, it did give us a gist, and it was amusing to read the results, but it's not bulletproof.
~Goose
Submitted by JCBarry on Mon, 10/12/2009 - 5:42pm
Yeah, I hear what you're saying.
I think you would run into the same potential flaws using Twitter as a data feed. However, I think you're right that there could be more specific, and interesting applications of the data.
Hmm...I'll have to check into the deletion thing. Lilly supposedly ordered my jet a few weeks ago, I'll have to check on that too. ;)
Jay Barry
MaxPC Profile | Jason Barry
Submitted by JCBarry on Mon, 10/12/2009 - 4:08pm
-- double post --
Must Read Articles
Feature
Review
Feature
Feature
Feature
Most Popular Articles
This Month's Issue
FEATURE How to Get FREE Programs, Services, Software & MoreFEATURE Digital Photo Printer RoundupHOW TOBuild a 3D CameraFEATUREDIY Arcade PCWHITE PAPERHow TRIM Works
Technology News
Computer Cooling Fans
Computer Cases
PC Game Controllers
PC Games
Computer Hardware
Headphones
MP3 Players
Stream Video
Computer Mouse
Monitors
Motherboards
NAS Storage
Networking
Laptop Computers
DVD Burner
Digital Cameras
Portable Storage
Computer Accessories
Smartphone
Antivirus Software
Sound Cards
Speakers
Computer Systems
Thumb Drives
Video Cameras
Video Card Reviews
Water Cooling
Gadgets
- Keyboards
© 2009 Future US, Inc. All Rights Reserved.