5 Other Topics

5.1 Recommendation Systems

5.1.1 Discussion Activity

Pick one of the following questions. Read one or more of the linked articles, or find a different article on the topic. Write a response of a paragraph or two where you:

• Summarize the main point of the article, for the sake of others who didn’t read it,
• Argue why the question is (or isn’t) important or interesting, and
• State your own response: a counter-point, an additional example, a question it raises for you, how it might shape your own life, etc.

Here are the list of questions:

• Should social media platforms be legally responsible for the content they recommend to you?
  • When Curation Becomes Creation (original link), By Liu Leqi, Dylan Hadfield-Menell, Zachary C. Lipton.
  • the facebook files
• How do social media platforms determine what content to recommend?
  • Read rsweeny21’s comment from a developer of Netflix’s autoplay.
  • Do you use Pinterest? Here’s how Pintrest’s recommender system works, at least a few years ago.
  • YouTube’s Recommender System uses neural nets to learn from watch times. This paper is a few years old but still insightful.
• Do recommendation systems “radicalize” people?
  • search for “YouTube radicalize” or similar. You’ll probably find articles asserting that it happens.
  • But you’ll hopefully find some counterpoints, like Examining the consumption of radical content on YouTube | PNAS.
  • Designing Recommender Systems to Depolarize | Abstract
• What do recommendation systems do to human creativity?
  • Predictive Text Encourages Predictable Writing
• Ads are often shown alongside other recommended content. Can ads be harmful?
  • Discrimination in Online Ad Delivery – classic paper by Harvard professor Latanya Sweeney

5.2 Text Mining (and bias)

• Tidy Text Mining (link is to a specific interesting section)
• How to make a racist AI without really trying and a follow-up in R

5.3 Resources

ACM Selects:

• Data Science

5.4 Relational Databases

Contrasting data storage formats: * CSVs are useful to store and exchange small to medium amounts of “static” data * relational databases are useful when data is growing / changing (especially from multiple sources) * relational databases are useful when multiple systems or stakeholders need the same data * relational databases are useful when data is too big to fit in memory, since RDBMS’s can often automatically figure out how to compute things in a “distributed” or “streaming” way * So-called NoSQL systems, like MongoDB and Firebase, have been popular in recent years, but SQL remains pervasive in industry because of its robustness, consistency, and performance.

SQL tutorials:

• Codecademy
• w3schools

SQL Syntax reference

• I kinda like the “railway diagrams” in the sqlite syntax documentation (e.g., for the SELECT statement)
• Codecademy has a reference
• TutorialsPoint

sqlfiddle