Skip to Main Content
It looks like you're using Internet Explorer 11 or older. This website works best with modern browsers such as the latest versions of Chrome, Firefox, Safari, and Edge. If you continue with this browser, you may see unexpected results.

Algorithm Bias & Gaming

workshop materials (fall 2020)

What do we find?

How Search Algorithms Work (from Google)

  • Google ranking systems - uses a series of algorithms
  • employ Search Quality Raters worldwide

Guidelines:

  • meaning of your query (intent, parsing synonyms correctly)
  • webpage relevance (keyword matching, interaction data)
  • content quality (prioritize reliable sources; use signals such as PageRank e.g. other reputable websites linking to the content)
  • webpage usability (works in a variety of browsers, designed for different devices, page loading time)
  • context & settings (your location, search history, search settings e.g. language or SafeSearch)

image of Google Search

image by 200 Degrees at Pixabay

Google's Influence

Survey Results:

What do you think? Would individuals respond that way today?

Question: Are search engines a fair and unbiased source of information?

Google Image Search

Demo: librarian

  1. What do you see? What don’t you see?
  2. Is this what you expected to see?
  3. Is a certain population represented, under-represented, mis-represented?
  4. From these images, how would you describe X person?
  5. Might this information influence your view of X person?

Other Terms:

  • Business leader
  • Rich person
  • Smart person
  • Financial Analyst
  • Professors
  • Computer Engineer

Question: Are search engines a fair and unbiased source of information?

Demo

  • vaccinations cause ...
  • artists are ...

Questions

  1. What did you notice about the auto-fill suggestions?
  2. What was missing that you found surprising?  What was included that you found surprising?
  3. Do you think that the Google / Yahoo / Duck-Duck-Go search predictions are biased?  Why or why not?
  4. Do some of these results seem sanitized or inflammatory?

Sample Searches:

  • climate change is ...
  • college students are ...
  • chemists are ...
  • voting
  • Trump
  • Biden (compare Google vs. Yahoo)

Questions:

  1. What sparked this researcher's interest in the topic of algorithm bias?
  2. How does she define an algorithm?
  3. In her research, who has been misrepresented in Google searches?
  4. What is one solution she is proposing?

Video: Algorithms of Oppression (3:43) Safiya Umoja Noble

Grind, Kirsten, et al. "How Google Interferes with its Search Algorithms and Changes Your Results; the Internet Giant Uses Blacklists, Algorithm Tweaks and an Army of Contractors to Shape what You See." Wall Street Journal (Online), Nov 15, 2019. ProQuest.

  • article based on 100 interviews and testing of Google search results
  • findings / claims
    • Google favors big business over smaller businesses; boosts favored websites such as Amazon and Facebook
    • Google engineers make adjustments behind the scenes; employees can ask for revisions to search results; low-paid contractors give feedback on the quality of Google's algorithms
    • has not always been agreement about how much interference there should be in search results
    • certain sites are prevented from surfacing in search results
    • some auto-fill suggestions are removed when the subject is controversial
    • Google wants to keep people on the results page where there are ads.

"Far from being autonomous computer programs oblivious to outside pressure, Google's algorithms are subject to regular tinkering from executives and engineers who are trying to deliver relevant search results, while also pleasing a wide variety of powerful interests and driving its parent company's more than $30 billion in annual profit ."


Heilweil, Rebecca. “Why Algorithms Can Be Racist and Sexist.” Vox, 18 Feb. 2020.

  • notes the challenges of algorithms that are not transparent, limited information on the design, data used, or how it really works
  • serious consequences can be the result of judgements or predictions from data
  • Are the systems accurate across multiple demographics? What is the data used to train the system is biased? There is a need for representative training data.
  • The intent is also very important. Is a creditworthy determination really intended to maximize profits?
  • Also concerns that the AI field may be dominated by men and with over representation by those who are white.
  • Algorithms, even if accurate, don't address issues of fairness or ethics.

Video: The Moral Bias Behind Your Search Results (9:11) Andreas Ekström, TEDxOslo (2015)

Questions:

  1. According to the speaker, why do individuals use Google?
  2. For what types of searches is Google good or less optimal? 
  3. What information might make something more findable on the web?
  4. Complete this statement: Behind every algorithm is always a _____________.

 


Video: Machine Intelligence Makes Human Morals More Important (17:11) Zeynep Tufekci, TEDSummit (2016)

Questions:

  1. Tufekci notes that software is getting more ______________  but less __________________.
  2. What is machine learning different from?
  3. What does she say computer systems (algorithms) could be amplifying?
  4. Describe how Facebook decides what news is highlighted? (Remember this is 2015.)
  5. What does Tufekci warn against outsourcing?