The Myth of Unbiased Search Results

This week we have an assigned academic blog post. The assignment is to critique “The moral bias behind your search results”, a TED Talk given by Andreas Ekström.

Ekström’s primary point is that while Google search may be quite useful for isolated facts, it does not and in fact can not provide unbiased results for more complicated queries. He supports the first half of this point with a simple example and provides more extensive support for the second half.

The support for the first half of Ekström’s point is straightforward and relies on our common sense understanding of the world. He points out that when looking for an isolated fact, there is not much or any disagreement as to what the answer should be. He notes that there are no groups working on proving that London is the capital of France or that a water molecule is actual H30. I both agree with his point and find the way he makes it eloquent. For finding simple facts with no reasonable dispute, Google search is a great source.

The more important second portion of the point Ekström makes is also well supported although I do have one quibble. He starts out by pointing out that a more complicated search or a search for knowledge rather than isolated fact introduces value judgements. Value judgements are filtered through our own personal experience and the communities in which we participate. While I agree with this point and find his statement and description of it well said, the example he uses to drive it home fails in one area.

For an example, Ekström brings up a search of images of Michelle Obama, first lady of the United States, and then later a search for images of Anders Behring Breivik, a terrorist. He shows the audience the image searches for these two individuals while telling the two stories of how people manipulated Google images specifically for these people at different times in the past.

Michelle Obama’s Google images results were manipulated by racists who made modified images to distort her face so that it looked like a monkey’s face and used their knowledge of how Google image searches work to put those images at the top of any search for her images.

After his terrorist attack, Breivik had a similar campaign launched against him by a search engine optimization expert where pictures of dog poop were made to appear at the top of his image results on Google.

For Michelle Obama, Google did not wait for the corrective features of their algorithm to kick in. They wrote new code to fix the problem quickly. For the terrorist, they did not make the same fix. Ekström points out the value judgement inherent in these decisions. He notes that being able to make these decisions for how their search engine behaves is power.

I agree with his point. My quibble is with his demonstration. He pulls up the image searches for the two individuals in question. Since it is now years later, the image search results come up with good results of accurate pictures for each individual. This is not a good demonstration of his point.

Ekström concludes his talk with the important reminder that behind every computer algorithm is a human being. Search engines were created by people and it is important to identify and be aware of what bias you are working with when using one. I agree with Ekström that taking responsibility for understanding your bias and the bias of the information you work with is vital.

People have an unfortunate tendency to believe that the computer algorithms behind search engines will work in an unbiased fashion. Computer algorithms are created by human beings. They don't spontaneously spring into existence from a cloud of electrons. Even if they did, the electrons would probably have some sort of bias that we simple humans just don’t understand.