|
Facebook is being creepy (pg. 3)
|
View this Thread in Original format
| planetaryplayer |
| i am the chosen one |
|
|
| Lira |
| quote: | Originally posted by Jon_Snow
Oh Lira |
Hello there, sir :gsmile: |
|
|
| Psyshell |
| I'm suprised no one has mentioned this yet but apparently people you may know is quite often people who have recently been looking you up on facebook. She's probably found you somehow (through friends of friends maybe) and has been stalking you. |
|
|
| SYSTEM-J |
| Lira has already said that. And one of these girls never knew my full name and I never knew hers. |
|
|
| JEO |
I'm quite sure data giants like Facebook use machine learning to great extents for this stuff. For example a well trained neural network (what they use at least for showing the right ads to the right people) can find patterns in a huge set of data, even when for a human everything in that data set would seem completely random. Probably these things are responsible for that "creepiness", since it's kind of a hallmark of neural networks in some way; noticing something we don't. In other words, there's something other than the fact you dated connecting you, you just don't know what :p neither do the developers, necessarily.
I personally wait for the day Facebook starts friend-suggesting me people I met physically and hung around with for some time (given I let Facebook access my location at all times), but had no connection to on Facebook. Probably even now that's just one semi-complex SQL query away for them.
Or they just looked you up, like everyone else said. You must be some hot stuff, man. |
|
|
| Jon_Snow |
| quote: | Originally posted by Lira
Hello there, sir :gsmile: |

At least you're not using drums or smoke signals to save money. |
|
|
| Chimney |
| quote: | Originally posted by Psyshell
I'm suprised no one has mentioned this yet but apparently people you may know is quite often people who have recently been looking you up on facebook. She's probably found you somehow (through friends of friends maybe) and has been stalking you. |
Really? Had no idea of this. |
|
|
| Lira |
Who said I'm not using drums to communicate, IGK?! :p
| quote: | Originally posted by SYSTEM-J
Lira has already said that. And one of these girls never knew my full name and I never knew hers. |
Can't have been me, I'm saying this was a coincidence all along!
Seriously, let's do some maths here. Suppose you have 200 Facebook friends (Facebook's most common friend number) and you've interacted with 10000 people your entire adult life. There are 60 million Brits britting in Britain. That's like looking for needles in a haystack, right? Not really.
If your friends have in average, 200 friends as well, you'd have 40000 people with mutual friends to choose from (200 x 200) - if everyone you've interacted with fell into this group, you'd know 1 out of every 4 people Facebook recommends (40000 / 10000). That's what Facebook would do if it only recommended people you've got a mutual friend with. That's a terrible way of looking at things because you can't know all of your friends' friends.
If we take this one step further, we now have 8 million people. That's an incredibly large sum, right? No way you could pick these 8 million people at random and find someone you know. However, they're not evenly distributed, because not everyone knows the same amount of people - some are more popular (I'm going to call them nodes) and some aren't. These nodes attract larger groups, and you've got to have incredibly popular nodes around you.
This makes sorting your friends' friends a lot easier (people you no longer share mutual friends with), because with few database queries Facebook can analyse which friends of yours interact with the same nodes... And I'm not even taking into account similar interests, which would make this a lot easier. So, out of the 8 million friends' friends, if these nodes have 800 friends in average (they're 4 times more popular than you), Facebook would sort between groups of 640 thousand people (Just 8% of that population of 8 million people). That's a lot, sure, but then Facebook would have to pick only those nodes that interact with you, which would make that percentage spike (I'm not going to take this into account, we've seen enough numbers here). The thing is, you're more likely to interact with them because they're more popular than you, and Facebook is more likely to use their help to filter its results.
If Facebook suggests 20 people to you every single day, that's 7300 people a year (or 11% of those 640 thousand people from the previous paragraph). Odds are you'd see one familiar face out of every 100 people if nothing else were relevant (or 1 every 5 days). Of course there are more variables and you'd expect the system to do better than that (my numbers were educated guesses anyway, so I'm not going for accuracy here), but the system should do better than that. Mind you, it works better with some people (e.g. you) than with others (e.g. me).
Cor version: You don't need any black magic to spot your romantic interests after a while. It helps, sure, but it really is a small world after all. |
|
|
| SYSTEM-J |
| quote: | Originally posted by Lira
Can't have been me, I'm saying this was a coincidence all along!
|
It was in the article you linked.
You don't need to give me the maths lesson, because I'm familiar with the idea you're talking about. We forget all the everyday events because there's nothing unusual about them and we notice the anomalies so they seem significant, even though statistically their probability is likely to be very normal given the sample size. Statistics very often cluster rather than being evenly distributed so a chain of seemingly significant events can occur in short order, strengthening the false impression. Meanwhile, billions of unmemorable non-coincidences are happening every single day in the background and are ignored by the mind.
I don't really buy into it because my Facebook account simply isn't churning through the suggestions as you're saying it does. There isn't the sample size for these things to crop up. Of course, given one billion Facebook users it's possible that I'm the one with the incredibly unlikely coincidence and it had to happen to someone. But I'm going into the SIM card explanation, especially as Lews reports similar findings. I remember Facebook in the past clearly suggesting friends based on the contacts in my email account I'd signed up with. |
|
|
| Mr.Mystery |
| I was just watching Louis CK on YouTube and now Facebook's saying I should add him as a friend. |
|
|
|
|