Comments on: How a Computer Program Can Learn All About You From Just Your Facebook Likes http://blogs.smithsonianmag.com/science/2013/03/how-a-computer-program-can-learn-all-about-you-from-just-your-facebook-likes/ Ideas, innovations and discoveries from the world of science Mon, 17 Jun 2013 16:53:16 +0000 hourly 1 http://wordpress.org/?v=3.4 By: Rowan Gonzalez http://blogs.smithsonianmag.com/science/2013/03/how-a-computer-program-can-learn-all-about-you-from-just-your-facebook-likes/comment-page-1/#comment-10330 Rowan Gonzalez Tue, 02 Apr 2013 08:56:55 +0000 http://blogs.smithsonianmag.com/science/?p=16452#comment-10330 I guess we are what we like. I guess we are what we like.

]]> By: JohnD http://blogs.smithsonianmag.com/science/2013/03/how-a-computer-program-can-learn-all-about-you-from-just-your-facebook-likes/comment-page-1/#comment-9969 JohnD Tue, 12 Mar 2013 17:06:24 +0000 http://blogs.smithsonianmag.com/science/?p=16452#comment-9969 Fernando, the particulars of what will indicate various things will vary by group, but the basic math behind the algorithm will always work. This is merely an applied version of a Bayesian routine, and is common in many fields of science. The number of people that were used is enough to allow them to derive fairly good results. Indeed, you could get results almost as good with just 100 people as a starting population. Having more people allows you to refine the routine and reduce the uncertainties, but won't change the final output that much. Fernando, the particulars of what will indicate various things will vary by group, but the basic math behind the algorithm will always work. This is merely an applied version of a Bayesian routine, and is common in many fields of science.

The number of people that were used is enough to allow them to derive fairly good results. Indeed, you could get results almost as good with just 100 people as a starting population. Having more people allows you to refine the routine and reduce the uncertainties, but won’t change the final output that much.

]]> By: Fernando http://blogs.smithsonianmag.com/science/2013/03/how-a-computer-program-can-learn-all-about-you-from-just-your-facebook-likes/comment-page-1/#comment-9958 Fernando Tue, 12 Mar 2013 08:54:26 +0000 http://blogs.smithsonianmag.com/science/?p=16452#comment-9958 Well, first, Facebook has more than 1 billion users, so the size of your sample is %0.000058 I'm not sure that this is representative, also the majority of sample likes in the article are very "american", I don't not even what the colbert report is, so the conclusion should be "You could infer this information from the americans". I'm sure that this information should change by country and even by time. The way we appreciate something, some company, etc. change over time. And even by demographic region. Probably you could estimate the global average, would be interesting to calculate the margin of error, especially for persons that deviate from the "Norm". Well, first, Facebook has more than 1 billion users, so the size of your sample is %0.000058 I’m not sure that this is representative, also the majority of sample likes in the article are very “american”, I don’t not even what the colbert report is, so the conclusion should be “You could infer this information from the americans”. I’m sure that this information should change by country and even by time. The way we appreciate something, some company, etc. change over time. And even by demographic region. Probably you could estimate the global average, would be interesting to calculate the margin of error, especially for persons that deviate from the “Norm”.

]]>