Computing Bias
Implicit Data
- Reinforces bias based on past behavior
Explicit Data
- More accurate, but bias can occur if user input is limited
Types of Bias
- Algorithmic Bias: Faulty computer system
- Data Bias: Data has errors or bad info
- Cognitive Bias: Person unintentionally shows bias (subconscious)
Intentional vs Unintentional Bias
- Intentional: Deliberate introduction of prejudice/unfairness for a specific outcome
- Unintentional: Trained on flawed or incomplete data
Mitigation Strategies
- Prevent computing bias by gathering more representative data
- Pre-processing phase: Prevent algorithmic bias, model planning/preparation
- In-process phase: Algorithm development/validation
- Post-process phase: Deployment and usage
Popcorn Hack 1
- Explicit Data: B) Provide name, age, preferences when making a Netflix account
Popcorn Hack 2
- Data Bias: B) System trained on a faulty dataset
Popcorn Hack 3
- Example of Unintentional Bias: Voice assistant only recognizes female voices since it was trained largely on female voices
Homework Hack
Implicit data and explicit data are basically two ways information gets collected, but one is more direct than the other. Explicit data is the stuff people willingly provide, like when you fill out a form with your name, age, or favorite type of music. Implicit data, on the other hand, is collected behind the scenes based on actions. An example of this is how Netflix figures out what shows you might like based on what you’ve already watched. For example, if you rate a movie 5 stars, that’s explicit data because you’re directly giving feedback, but if you keep watching horror movies without saying anything, Netflix assumes you love horror, which is implicit data.
Completed MCQ:
