Monday, September 12, 2016

Confirmation Bias with Search Engines


A short preface: the rest of this post is not mean't as a political statement. While I am a US citizen, I do not have many positive feelings about any of our presidential hopefuls - they all have their faults.

This morning I got a bit of a disappointing message. An acquaintance of mine posted a photo of three search engines (Google, Bing, and Yahoo) filling in the statement "Hillary Clinton is ". Both Yahoo and Bing came out with some very negative comments against Clinton, which is not too surprising. Then there was the Google search. Google was purported to have only good things to say. This individual then took this to indicate that Google may be in the pockets of HRC, or are just trying to bias their own results in favor of her. I simply cannot let this slide.

I responded with a decently long comment thread with images of each of these search engines responding to the same phrase, but I also replaced Hillary Clinton with the names of each of the other three main contenders for office: Donald Trump, Gary Johnson, and Jill Stein. The results were hardly shocking.

I'll start with Google. To be fair, I was using Google Chrome web browser which maybe was giving me different results, but I have my doubts. In each case the results were mixed. Clinton was indeed called 'awesome', and is suspected of winning according to Google. However at the same time she was referred to as a 'robot' and suspected of losing. The last result is just a statement of fact, because she is indeed running for president. A similar ratio is there for Trump, 2 positive ('awesome' and 'going to win'), 1 negative ('dead'), and 1 statement of face ('orange'). I will not comment on the 'not a conservative' because that could be seen as positive or negative depending on your personal leanings. Of the four, only Gary Johnson has fillings that are somewhat policy related, and then poor Jill Stein doesn't even get a full fill-in ('Jill Stein is from'). In my opinion, Google's search seems to actually force a neutrality bias because as we will soon get to, the other two engines have very clear impressions of our presidential candidates.

On to Yahoo, the little engine that could. Yahoo seems to favor giving far more results than Google which gives us an opportunity to get a larger perspective on the situation. As it turns out, my acquaintances post is more or less accurate. Most accusations appear, just in a different order. So is Yahoo in the pockets of Trump? Well clearly no. Heck Yahoo called him Hitler, falling for Godwin's law. None of those fill-ins sound positive to me either. Gary Johnson and Jill Stein get similar treatments, but more of the questioning variety. Is Johnson a libertarian, liberal, or republican - taking all bets. At least Stein has some facts associated with her: she is indeed an option for president. So perhaps this means that Yahoo is in the pockets of the candidate that has the least support nation wide and has the least money - that definitely makes sense.

Finally we get to Bing, the Google wanna-be. Again Hillary gets no love as noted in the picture above. My search didn't come out with 'murderer', but hey being a witch is close enough, right? Oh, but what's this. A little Trump love from Bing? Trump is apparently 'going to win' and is 'America's last hope'. Sounds to me like Trump has Bing's number... Oh wait, there's Hitler references again. Good ol' Godwin strikes back. Johnson, again, brings up actual policy debate (vaccines and borders). Finally, Stein just get's hit by the synonym monster twice ('not anti-vaccination' and 'not an anti-vaxxer'; 'wacko' and 'loon'). With more positive findings from Trump and Stein, it looks like Bing is clearly against Johnson and Clinton.

But really no. The problem with basing assertions on what search engines fill in for you is that you completely misunderstand human psychology and search engines. First of all, humans love to complain, we just do. Any student that has ever used RateMyProfessor.com knows this. You could have the best professor, but there are just too many people that hold grudges because they couldn't figure out a simple due date and they want you to know about it. The point being, it's far easier and more entertaining to write something negative than it is to write something positive. Negativity just makes us more 'colorful' with our words and is ironically fun to express. As a result, a majority of articles (especially when dealing with politics) are going to be negative. Hell, this entire post is based off my own disdain towards a topic and I wrote it in about 15 minutes. Secondly, with so much negativity going around search engines are going to pick up on that because search engines are predominately based on what is the most frequently found and what is searched. Having actually written an 'n-gram' code in the past, I'm well aware of this limitation. Lastly, as mentioned before, people like to read negativity, especially when it agrees with them. Thus, more searches are likely going to be negative searches. This is called 'confirmation bias'. You want something negative, so you search for something negative that agrees with you. It's no wonder that these searches are predominately negative then.

As my last point, I just want to refer back to the original image. The Google search shown only brought up 2 fill-in's for Clinton, but mine had 5. The only one's shown were the same two positives that I had gotten as well. Simple answer: it's the fucking internet, don't believe everything it spouts. In 5 minutes with Powerpoint I was able to make Bing a total Trump whore. It's a bit sloppy, but I hope it gets the point across: don't trust everything you read. I encourage you, as potential voters, go and do these same searches for yourselves, because for all you know, I'm lying too!