Facebook Data Team: Relationships and Happiness(facebook.com)
facebook.com
Facebook Data Team: Relationships and Happiness
http://www.facebook.com/note.php?note_id=304457453858
12 comments
"We already have methodology for measuring the happiness of Facebook users: by considering how many positive words people use in their status updates (see the USA Gross National Happiness Index). This method allows us to see whether a person's Facebook relationship status affects how positive and negative they are." (from post, emphasis mine)
No it doesn't. Correlation vs. Causation.
Counter-example: assume these two men (simplifying to single and non-single categories) are equally happy in real life:
Singles' non-relationship status updates: "Jeff is psyched for his Super Bowl party!" vs. Non-singles' non-relationship status updates: "Tony is psyched for his Super Bowl party!" Comparison by Fb methodology: equal happiness level
Single's relationship status updates: [non-existent] vs. Non-singles' relationship status updates: "Tony loves his super awesome girlfriend Melissa!!! Sweetheart, you're the bestest gf evarrrr!!!1!" Comparison by Fb methodology: Tony seems happier.
Conclusion: obvious nonsense. Somebody tell these guys to take a stat class.
No it doesn't. Correlation vs. Causation.
Counter-example: assume these two men (simplifying to single and non-single categories) are equally happy in real life:
Singles' non-relationship status updates: "Jeff is psyched for his Super Bowl party!" vs. Non-singles' non-relationship status updates: "Tony is psyched for his Super Bowl party!" Comparison by Fb methodology: equal happiness level
Single's relationship status updates: [non-existent] vs. Non-singles' relationship status updates: "Tony loves his super awesome girlfriend Melissa!!! Sweetheart, you're the bestest gf evarrrr!!!1!" Comparison by Fb methodology: Tony seems happier.
Conclusion: obvious nonsense. Somebody tell these guys to take a stat class.
She probably is, she's just an intern.
So, to optimize for positivity, be in a relationship but don't get married.
Doesn't the article contradict this strategy? FTA: People who are in a relationship seem less happy compared to married folks
With the massive amount of relational data being created in social networks, there is a huge opportunity in the field of analytics. This particular min-study shows trends in happiness vs relationship status, which could be used for a dating site/app spin-off. In a similar manner, analytics can be used to improve media consumption (video, sound, image, text)-- content aggregation is no longer feasible because too much information is being created everyday, so we must find a way to reduce SNR and provide the user with what he/she needs to see.
I am really excited to see many analytics-oriented startups in 2010.
-Ivan
I am really excited to see many analytics-oriented startups in 2010.
-Ivan
Agreed. I would recommend friends pursuing an MBA go for a masters degree or PhD in behavioral economics instead. In addition, I'd like to see more analytics oriented around the individual's own delta - how many times did they switch from "complicated" to a relationship in a 5 year period? How many friends did they add? How many people did they block? A more realistic measure of personal happiness might be obtained by more robust analysis of how a person's self reporting changes over time, rather than a static snapshot of how they report at a simple point compared with how many other people report that same thing in stasis.
One thing I'd like to see more on these kinds of data-mining analyses are some sort of error bars, variances, etc. They tend to just report group averages, like A=5, B=2, but there's a big difference between (using, say, 1 stddev as the bracket), A=5 +/- 1, B = 2 +/- 1, in which case they really do seem to differ, and A=5 +/- 15, B = 2 +/- 25, in which case you've got mostly noise.
I recall reading a study about happiness somewhere years ago [1] that showed people's happiness generally doesn't change that much. It reacts temporarily to positive/negative effects but then tends to go back where it was.
[1] I think it was probably a critique of this article: http://www.nber.org/papers/w10499
[1] I think it was probably a critique of this article: http://www.nber.org/papers/w10499
It's worth pointing out that the 4 graphs on the page have different normalizations, which makes naive comparisons between them misleading.
Did Facebook ask me if I would be happy if they did this kind of data-mining on my status updates?
That type of question is probably covered in their Terms of Service.
"... Did Facebook ask me if I would be happy if they did this kind of data-mining on my status updates? ..."
Good question. You may not have realised you gave Fb the rights to "infer" from their server logs & interactions when you created your account. However there is a solution. Get back your right to say who can peek into your server logs by owning your own servers again. Listen to Eben Moglen and his "Freedom in the Cloud" talk to understand how ~ http://www.isoc-ny.org/?p=1338
Good question. You may not have realised you gave Fb the rights to "infer" from their server logs & interactions when you created your account. However there is a solution. Get back your right to say who can peek into your server logs by owning your own servers again. Listen to Eben Moglen and his "Freedom in the Cloud" talk to understand how ~ http://www.isoc-ny.org/?p=1338
They don't have to ask you since you already gave them permission.
References for that, please?
From the Privacy Policy § How We Share Information:
> To help improve or promote our service. Sometimes we share aggregated information with third parties to help improve or promote our service. But we only do so in such a way that no individual user can be identified or linked to any specific action or information.
> To help improve or promote our service. Sometimes we share aggregated information with third parties to help improve or promote our service. But we only do so in such a way that no individual user can be identified or linked to any specific action or information.
data mining seems like one of the least offensive things they could do with my data. as is clear in the report, since it's all aggregated information, there's no information tied to any person. i might be concerned that someone at facebook could read my status updates, but lets be honest, many employees at facebook probably have the ability to read my status updates anyways.
that's a rather odd way of reporting numbers. "negative 5% more positivity than average." wtf?
There's definitely a problem with the readability of the graphs in particular, and with some of the sentences in the summary.
People in an open relationship are less happy. Why?
Because most people who declare themselves to be in an open relationship on Fb aren't in an open relationship. It's most often done just for the sake of being silly or telling everyone who your "BFF" is. The latter is more likely the closer you are to being a 16-year-old girl.
But why would that correlate with "negativity" across a subsection of people?
No, people in open relationships post fewer status updates with "happy" words. They also post fewer status updates with "unhappy" words. So, we might be lead to conclude that people in open relationships choose, for whatever reason, to expose their emotions in status updates less often than others. That doesn't particularly mean they have fewer emotions, positive or negative than anyone else.
You also see this strongly in the "widowed" category, who have dramatically less happy or unhappy postings.
You also see this strongly in the "widowed" category, who have dramatically less happy or unhappy postings.
Probably jealousy + lack of emotional gratification.
Facebook has no shame. This is a complete ripoff of okcupid analysis.
Yeah, great example of how facebook does not understand privacy. Why should I care if its anonymous or not, I dont want them to be even able to find out that kind of stuff about me!? It's like they dont even understand how far reaching this is; their finding out about the happiness of millions of people and they disguise it at as some kind of "V-Day Trivia". This shit creeps me out.
It's fairly easy to deactivate or delete your account. Account > Account Settings > Deactivate Account. To preempt any objection that you want to be able to use Facebook, you just _also_ don't want them to analyze your status updates: that's nice, but they are not willing to offer you service on those terms. I guess it's fine to complain about it in hopes that something will change, but the winds are not blowing in that direction, Mr. Quixote.
"... It's fairly easy to deactivate or delete your account. Account > Account Settings > Deactivate Account. ..."
Is all the data deleted? cf: Eben Moglen ‘Freedom in the Cloud’~ http://www.isoc-ny.org/?p=1338
Is all the data deleted? cf: Eben Moglen ‘Freedom in the Cloud’~ http://www.isoc-ny.org/?p=1338
I think they actually make an attempt to do so when you delete (not deactivate) your account. However, this goes beyond the promises they make to you in their Terms of Service and beyond, in my opinion, any moral duty they have to their users.
"... I think they actually make an attempt to do so when you delete (not deactivate) your account. ..."
Facebook has repeatedly removed services [0] that allowed you to remove all traces of your account via the API ~ http://www.google.com/search?q=facebook+suicide
[0] Cnet, Rafe Needleman, "Facebook cuts off Suicide Machine access" http://news.cnet.com/8301-19882_3-10424683-250.html
Facebook has repeatedly removed services [0] that allowed you to remove all traces of your account via the API ~ http://www.google.com/search?q=facebook+suicide
[0] Cnet, Rafe Needleman, "Facebook cuts off Suicide Machine access" http://news.cnet.com/8301-19882_3-10424683-250.html
I'm not against analytics or facebook in general. It's just that this company and this article in particular gives me the impression, that they dont give two shits about privacy and just do it because the last few times they tried to cross the line they got caught. I'm not expecting them to be nice, no company is, but they should at least try to show some respect for our privacy - and checking whether I'm happy or not is not doing that.
I would have liked to see time domain analysis -- so when a person goes from 'single' to 'in a relationship' does one see a pop in happiness? Is it sustained?