Yesterday I happened to hear a feature on CBC Radio 1’s ‘Spark‘ an episode called Algorithm awareness. Working in the technology sector I am fairly aware of the algorithms around me, and little came to me as a surprise, but it served as a great reminder that the bulk of the population are relatively unaware of the significant role of algorithms our in society, or even their existence.

Algorithms are everywhere now-days, taking over tasks that used to take humans huge amounts of time, often more than anyone could handle. As a result things are possible now that never were in the past. OK, big surprise right? But really, did you ever stop to think about how a spam filter works? A spelling or grammar checker? How Netflix or Amazon are able to suggest a movie that you might be interested in? How a scanner can read a printed page and convert it into editable characters, or how Google was able to develop a car that could drive itself across the United States? The algorithms used for each of these applications are remarkably similar.

In 1959 Arthur Samuel defined Machine Learning as the field of study that gives computers the ability to learn without being explicitly programmed. 1959, that’s right. Machine learning is not a new concept, but it is only recently that computers have become powerful enough to make these algorithms practical.

While most computer programs are based on a series of logical arguments (If true, elseif, else etc.) from which a certain input would deliver a concrete answer, Machine Learning algorithms keep re-asking the questions with different inputs to predict what the best answer might be when no explicit direction is defined. In fact Machine Learning is pretty much computer-aided statistics.

.jpg)

Backpropagation algorithm thanks to Dr. James McCaffrey of Microsoft MSDN Magazine

Ever try to make sense of an equation like this? Some poor bloke in the Machine Learning department has been staring at something like this all day.

So how does this affect Facebook, and why does it apply to me? Well, having read this so far, it is probably of little surprise that Facebook engages Machine Learning algorithms in some capacity. You may have heard that Facebook drew a lot of negative media attention earlier this year for their mood manipulation experiment. Well in fact, Facebook uses Machine Learning algorithms to control just about everything you see.

Why not? After all, we want to see what is of interest to us right? Facebook is trying to engage us. Why show us updates from a bunch of ‘friends’ that we never seem to read or click on? Well, chances are, they don’t.

I always kind of assumed that if my friends scrolled down a bit they would find a post or two of mine. There was some sort of understanding that things were in something that resembled reverse chronological order. In fact, it is possible to do this. Took me about 5 minutes to figure out even after I was told it could be done. In the top-left corner under your picture, next to “News Feed” there is a tiny little arrow that lets you switch between “Top Stories” and “Most Recent”. Change this to “Most Recent” and I am sure you will be shocked to see friends you didn’t know you had.

Jonny, Carl, Jenn, where have you been? My guess is that, while I may have read your posts, I never clicked ‘like’, or clicked through to the post itself, so Facebook had no idea that I may have actually been interested.

According to the CBC Spark report Twitter had considered such an algorithm as well, but realized that its role in providing unbiased news feed was too important, and decided against it. Facebook, on the other hand, relies heavily on advertisers for ad revenue. Its whole business model is based on the ability for a Machine Learning algorithm to analyze the HUGE supply of data it stores and gear ads towards people that are more likely to buy.

Facebook is trying to do you a favor by showing you updates that are likely of interest to you, particularly it seems if there is some advertising revenue in it for them. If your posts have been popular in the past, they are more likely to be shown. Spark went on to say that mentioning an advertiser such as BUD LIGHT in your post then increased your chances of having your post seen, while Carl may have had devastating news that he needed to share, but because no-one ‘liked’ his last piece of bad news no one ever even saw this one.

Now take Google. Google uses a similar method to show you search results that are geared to you. This becomes evident when trying to help a friend find something that is the first result returned to you, and doesn’t show up on their first screen of results. Often this is a help, but other times I find myself asking my wife to find something for me knowing how Google biases results, and recognizing that it is more likely to tell her the answer than it is to tell me. You can disable this feature by using an incognito window in Chrome. (Google it. It’s the first thing on the list. At least, it is for me).

Google is another master of tracking personal information. It uses your location, your G+ activity, your searches, browser cookies, and the rich information that it attaches your G-mail account if you are logged in. The good news is that much of the time we can get the information we are looking for faster. The bad news is that our behaviour biases the results we see, and we wind up with a skewed perception of what is really going on in the world.

My suggestion to you, if you are a Facebook user: Try switching to ‘Most Recent’ once in a while, just to see who is out there. You will probably be surprised. Remember that if you only ‘like’ or comment on wedding and baby pictures, Facebook may decide that that is the only thing you really want to see, and oh, say ‘hi’ to Jonny, Carl, and Jenn if you see them, and tell them that I have missed them.