NEW
Font size
Worksheets5.6 Algorithmic Bias and Fairness: Crash Course AI #18
Total questions: 8
Worksheet time: 4mins
What is algorithmic bias?
When AI uses only math code.
When AI systems copy or exaggerate real-world biases.
When people purposely make AI unfair.
When AI can't process data.
How can societal biases end up in AI systems?
By only using past data to train AI.
By leaving out certain groups from training data.
When training data contains hidden societal biases.
When AI ignores cultural differences.
What is a common issue with facial recognition AI trained on unbalanced data?
It becomes too efficient at identifying all faces.
It struggles to recognize faces from underrepresented groups.
It prefers certain facial features.
It needs more processing power than balanced data.
How can a positive feedback loop in an AI system like PredPol increase bias?
It ignores new data.
It repeats and amplifies biased patterns.
It becomes less efficient.
It finds new, unbiased data.
Why did Microsoft's Twitter bot Tay start posting offensive content?
It was programmed with offensive language.
It learned from biased data from users.
Its developers made it controversial.
A technical glitch affected its language processing.
Why did John-Green-Bot have a low chance of being hired in the "Hire Me!" AI system?
The system favored unique names.
Previous applicants named John were rejected, causing a negative feedback loop.
John-Green-Bot was not qualified.
All positions were filled.
An AI hiring system suggests hiring younger applicants because it finds older applicants have less tech knowledge. What is a key issue with this recommendation?
It is always illegal to consider age in hiring.
The algorithm might be misinterpreting correlation (two things happening at the same time) as causation (one thing causing another thing) overlooking individual skills.
Younger applicants are inherently more innovative.
Older employees are more expensive to train.
What is the first step in addressing algorithmic bias?
Developing completely unbiased algorithms.
Collecting more personal data from users.
Recognizing that algorithms can be biased and evaluating their recommendations.
Enforcing strict legal penalties for biased AI.
