Get Help
Skip to Main Content

Digital Citizenship Workshop Series

Rise Against the Machines: Understanding Algorithmic Bias

Algorithms are everywhere in our daily life. Amazon makes recommendations for books. Netflix recommends movies. News and trending stories appear on our Twitter and Facebook feeds. Algorithms determine what information we see, and in what order. This includes search engines like Google, Google Scholar, and library databases.  A 2020 study found that college students placed trust in Google as "the arbiter of truth." But is this trust misplaced? Algorithms are designed by humans and therefore reflect the assumptions and biases of their designers. Algorithms are not neutral, but this does not mean they are not useful tools for research or everyday life. It helps to know their limitations and biases.  In this workshop, you will learn how algorithms can perpetuate bias and discrimination, and you will identify the potential causes of algorithmic bias as well as some preventive strategies. This event is open to all.

Key Ideas- Algorithmic Bias

Algorithm: The set of logical rules used to organize and act on a body of data to solve a problem or accomplish a goal that is usually carried out by a machine 

Example Algorithm:
Program that encodes a person's options for navigating the library.
Goals = food or research
If goal = food, then location = Starbucks.
If goal = research: if on campus, then location = Information Desk; if off campus, then location = online chat

Algorithm Infrastructure- Example:

Graphic adapted from: Long, D., Moon, J., & Magerko, B. (2021). Introducing AI worksheet (model activity). In EEAI-2021: The Eleventh Symposium on Educational Advances in Artificial Intelligence (p. 15706). Association for the Advancement of Artificial Intelligence.

Types of Algorithms:

  1. Rule-based algorithms (instructions are constructed by a human and are direct & unambiguous) 
  2. Machine learning algorithms (fits under the broad umbrella of artificial intelligence; you give the machine data, a goal and feedback when it's on the right track and leave it to work out the best way of achieving the end)

Algorithmic Bias: Occurs when a computer system reflects the implicit values of the humans who are involved in collecting, selecting, or using data

Causes of Bias:

  1. Historical human biases in training datasets
  2. Incomplete or unrepresentative training data
  3. Proxies for sensitive attributes become feedback loops
  4. Algorithmic objectives that minimize prediction errors and benefit majority groups


Personalization Resources

Personalization: Process (used in targeted digital advertising) of displaying search results or modifying the behavior of an online platform to match an individual’s expressed or presumed preferences, established through creating digital profiles and using that data to predict whether and how an individual will act on algorithmically selected information 

Preventive Strategies

  1. Think twice before downloading free apps - you are paying with your data!
  2. Provide feedback when you see biased or inappropriate search results
  3. Request more transparency from corporations who use algorithms
  4. Get involved with advocacy and educational groups