filter content bubbles
play

Filter/Content Bubbles Tom Clark Quick History Search engines did - PowerPoint PPT Presentation

Filter/Content Bubbles Tom Clark Quick History Search engines did not personalize information Simply using keywords to find pages 2005: Google implemented a personalized search algorithm for ALL users Social media followed suit


  1. Filter/Content Bubbles Tom Clark

  2. Quick History Search engines did not personalize information ● ○ Simply using keywords to find pages 2005: Google implemented a personalized ● search algorithm for ALL users Social media followed suit after its inception ● ○ Facebook ○ Twitter ○ Reddit (later on) ○ Many more Known as “Deep expert” search engines ● ○ Able to profile individual inquirers ○ “Shallow” just knows specific events Personalization came from a demand for more ● relevant information

  3. What is a “filter bubble”? Term coined by Eli Pariser in 2011 ● A “state of intellectual isolation” ● Your queries are unique ● Personalized “bubble” of information ● ○ “You like this, so you must like this” Almost every website uses personalization ● Synonymous with news echo chambers ●

  4. Filters came before the internet Newspapers usually favor certain political ● sides and cover content relevant to their readers ○ A light filter bubble of content for their readers Magazines only covered content that their ● subscribers wanted. The rise of the internet made it easier to find ● sources that aligned with your views. ○ Made it profitable for them as well.

  5. Companies Utilizing Filters Google ● ○ If you clicked on something, you want more ○ Location history ○ Irregardless if you have an account Facebook ● ○ “Likes” mean you enjoy it ○ Clicks mean you’re interested ○ How fast/slow you scroll ○ Websites you visit (tracking pixels) ○ Facial recognition in tagging = locations Netflix ● ○ User searches show what movies they should add to their service ○ Movie suggestions Twitter ● ○ Suggests people to follow ○ What order to show tweets in

  6. Advantages Increased user happiness ● ○ You see topics you’re interested in ○ Opinions that agree with you Relevant information ● ○ Googling “restaurants nearby” uses location ○ As a CS student, googling “MIPS” should show assembly language content

  7. Disadvantages? Results align with the users’ interests ● ○ Lack of information diversity “Objectivity matters little when you know what you ● are looking for, but its lack is problematic when you do not” - Thomas Simpson The sharing of information is key to the web ● ○ Can’t receive all information if some is hidden

  8. Studies Wharton - Personalized recommendations of music create ● commonality and not fragmentation. ○ “Consumers reportedly use the filters to expand their taste rather than limit it” Oxford/Stanford 2013 - Analyzed browsing histories of 50k users ● ○ Looked at how they voted in the 2012 election versus their history ○ Web searches and social media contributed to ideological segregation ○ Found they were only being shown pages from their side of the spectrum New York University ● ○ Twitter users have access to a wider span of viewpoints directly from political actors or through their friends/relatives.

  9. Google News Originally an “aggregator app” ● ○ Pull from thousands of feeds into one place Re-launched in 2016 as a new app ● Advertised as a heavily personalized news feed ● ○ Buttons to “see less of this” ○ Ability to hide all stories from a specific source Deep neural networks to predict news preference ● ○ Similar to facebook feed ○ Analyzes scrolling speed, location, clicks The longer you spend on the site, the more ● isolated you become.

  10. Political Issues Bubbles lead to the “Us versus Them” mentality ● China’s “Great Firewall” filters content that the ● government deems bad. ○ Gives government control of what ideas are passed through networks ○ Reduces minority opinions US 2016 Presidential Election ● ○ Russians used fake accounts to influence voters through social media ○ Worked to further separate opinions ○ Echo chambers of potentially fake information

  11. Ethical implications Information is hidden without user consent ● ○ Leads to “information blindness” Use of filter bubbles means people are more ● susceptible to confirmation bias ○ “Fake news” effect Cambridge Analytica’s 87 million Facebook ● profiles highlight problems with filter bubbles ○ Christopher Wylie: “...The firm had the ability to develop “psychographic” profiles of those users [to] shape their voting behavior”

  12. Ethical Analysis Companies are personalizing information ● ○ Leads to better ad recommendation and thus more money for companies ○ Increases user happiness and gives valid information. ○ However, people are becoming more isolated in their ideas i.e. information blindness How is it viewed in ethical perspectives? ●

  13. Kantian Analysis Are the companies using users as means to an end? ● ○ Facebook/Cambridge Analytica → Yes! ○ Companies sell user metadata to advertisers But users are getting more valid information… ● ○ If the search engines are giving relevant information, is it bad? Selling user data is a means to gain money ● ○ Invalidates the categorical imperative However, the question is not of selling information, ● but that of filter bubbles being ethically right. Companies are filtering to better recommend ● information, the side effect is the bubble.

  14. Act and Rule Utilitarianism Obligated to take the path of most positivity ● Positives: ● ○ Better recommendations ○ More relevant information ○ Increased user happiness Negatives: ● ○ Information blindness ○ “Us versus them” mentality Positives outweigh the negatives. ●

  15. Virtue Ethics Do they follow the “moral character”? ● It is of good moral to give better ● recommendations. Intellectual isolation is a side effect ● Users should be able to control how much ● recommendation they have. Unethical. ●

  16. Deontological Is it acted in accordance with a set of ● principles and rules? The act of filtering content can result in ● intellectual isolation. They should avoid filtering because it is the ● right thing to do ○ Even though it increases happiness In deontological analysis, it’s about a ● characteristic of the act and not the result, even though it is good.

  17. What does our future look like? Personalization will continue unless users complain. ● There are studies of the benefits and disadvantages ● ○ Scientists are divided on whether or not it is good or bad for users “Neural networks and Virtual Assistants know our preferences ● better than we know our own” - Eli Pariser

  18. Thank you!

Recommend


More recommend