Searching for Turkey (not turkey): A Thanksgiving themed experiment

Ah, Thanksgiving. My favorite time of the year to conduct slap-shod experiments! This year I decided to put the filter bubble to the test. The filter bubble is the selective presentation of search results, news, and other information by a search engine or social media due to the use of an algorithm using personal data (search history, browsing history) to personalize your search results or in the case of Facebook, your News feed. Eli Parisier spoke about filter bubbles in 2011.

The influence of filter bubbles has received more attention recently as a contributing factor to the widening liberal-conservative divide in the wake of the 2016 Presidential election. Using data from a 2015 Facebook study, The Wall Street Journal created a graphic showing mocked-up liberal and conservative Facebook feeds side-by-side.

While much attention has been paid recently to the filter bubble created via social media platforms, I decided to focus my attention on the filter bubble caused by search engines, namely the search behemoth, Google.

I LOVE Google! I LOVE that it is run by magical information seeking gnomes living in Google, California. I LOVE that the Google gnomes know exactly what news will bring me joy because it won’t conflict with any of my preconceptions of the world. Those gnomes are FANTASTIC!

But seriously…the “gnomes” are really factors plugged into an algorithm that ranks the results of my searches based on what Google knows about me. Two primary factors include my location and previous browsing history.

asda_fb_gnomes_wide

Asda giant gnomes

Clearly, it was time to devise an experiment! Since it is close to Thanksgiving, turkeys are on my mind. Turkeys are also non-political. I have conducted very few (if any) previous searches on turkeys or Turkey. Looks like the word turkey is a prime candidate for my experiment.

I decided to try to cause Google to erase all mentions of “turkey” the bird from my first page of search results.

turkish-au-pairs

Turkey, the country

file0002043695191

Turkey, the bird

 

 

 

Method

Using two Chrome browser windows, one opened to Google, the other opened to DuckDuckGo (a search engine that does not track your searches or browsing), the following steps were taken:

  1. The word “turkey” was typed into the search box of each search engine.
  2. Three links to sites about “Turkey,” the country, not the bird, were clicked on in each window.
  3. In between clicks the back button was used to get to return to the original results.
  4. After the third link was clicked, the browser window was closed.
  5. New browser windows were opened to Google and DuckDuckGo.
  6. The pattern was repeated until links referring to turkey birds dropped off the Google results list.

Results

It took three fresh searches for “turkey” before the first page of results from Google no longer contained links to sites about turkey birds. That translates to 3 searches and 9 clicks on links to information on the country. The Wikipedia article on turkey (birds) was the exception. Even after a 4th round of link clicking the Wikipedia article persisted.

How did DuckDuckGo fare? The 4th search for “turkey” in Duck Duck Go yielded the same mix of results (equal bird/turkey references) as had been retrieved after the initial search.

google-turkey

Google results after 3 searches

duck-turkey

DuckDuckGo results after 3 searches

It only took 3 searches, clicking on 9 links in Google related to Turkey the country, instead of turkey, the tasty Thanksgiving Day mascot, to have Google stop presenting results related to the turkey bird option in the top 10 results.

Clicking on three links in between refreshing the browser windows was an arbitrary decision made to save time. If this were a real experiment a better estimation of the number of clicks could be determined with a higher refresh rate. The experiment lacked a control group. For instance, a separate user could have logged into Google and DuckDuckGo and repeatedly refreshed the search without clicking on links. This would have provided a baseline for comparison. After all, what if Turkey (country) or turkey (birds) were trending in the news during the experiment?

In the future, comparison of search engines other than Google and DuckDuckGo would be interesting.

Breaking Through the Bubble

Now just imagine how your search results for something like gun control, the death penalty, or immigration might be skewed toward a particular opinion based on your browsing history.

Some tools have been developed recently to help people break out of their news filter bubbles: AllSides is a web site that shows the liberal, center, and conservative side of daily news stories. The site also has a handy news media bias ranking scaleEscape your bubble is an add-on that daily injects a news article about an opposing viewpoint into your Facebook feed. Unfiltered.news is a data visualization tool that shows what news is least reported is least reported in a given location.

Here are a few things you may do to break out of your filter bubble:

  1. Use private browsing – You can change the settings on your browser to private (Incognito mode on Chrome). You will lose most of the personalized goodness brought to you by the magical algorithm gnomes, but don’t worry, you can get it back after you’re done.
  2. Use a search engine that doesn’t track you. When looking for news and information use a search engine like DuckDuckGo or StartPage.
  3. Sometimes click on things you don’t find interesting. Remember it only took clicking on nine links to make turkey birds disappear.

About the Author

Katherine Gerwig is an Information Commons Specialist at Metropolitan State University Library

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s