All Flashcards
What is the definition of Bias?
Tendencies or inclinations, especially those that are unfair or prejudicial.
What is demographic parity?
A fairness metric ensuring outcomes are proportional across demographics.
What are fairness metrics?
Measurements used to assess and ensure equitable outcomes in algorithms.
What is meant by a skewed dataset?
A dataset that does not accurately represent the population it is intended to model, leading to biased outcomes.
Define algorithmic bias.
Systematic and repeatable errors in a computer system that create unfair outcomes.
What is the meaning of 'equal opportunity' as a fairness metric?
Ensuring that different groups have an equal chance of achieving positive outcomes.
Define 'unintentional bias'.
Bias that occurs due to unintentional choices in data or algorithm design.
What does 'mitigating bias' mean?
Taking steps to reduce or eliminate bias in algorithms and data.
What is meant by 'representative data'?
Data that accurately reflects the diversity of the population it aims to represent.
Define 'disproportionate flagging'.
When a system unfairly identifies a specific group more often than others.
How is bias manifested in criminal risk assessment tools?
Historical data reflects societal biases, leading to disproportionate flagging of specific groups.
How is bias manifested in facial recognition systems?
Systems trained on datasets lacking diversity may not accurately recognize women or minorities.
How is bias manifested in recruiting algorithms?
Algorithms may learn to prefer candidates similar to past successful candidates, perpetuating existing imbalances.
How can diverse data sets be applied in facial recognition?
Training facial recognition systems with images from various demographic groups to improve accuracy across all users.
How can fairness metrics be applied in loan applications?
Using metrics to ensure loan approval rates are similar across different demographic groups, preventing discrimination.
How can algorithm reviews be applied in AI-powered healthcare?
Regularly checking algorithms used for diagnosis to ensure they don't disproportionately misdiagnose certain groups.
How can addressing human bias be applied in software development?
Training development teams to recognize and mitigate their own biases, leading to more inclusive software designs.
How can increasing tech diversity be applied in product design?
Having diverse teams create products that cater to a wider range of user needs and preferences.
How can bias be identified in a search engine algorithm?
Analyzing search results to see if they disproportionately favor certain viewpoints or demographics.
How can bias in a chatbot be identified?
Testing the chatbot with various prompts to see if it responds differently based on user demographics.
Why do computing innovations often reflect existing biases?
They use data from the world, which is already influenced by human perspectives.
How can diverse datasets help prevent bias?
They reduce the risk of skewed results by representing the entire population.
Why is it important to review algorithms for potential biases?
Helps catch and fix biases early on, ensuring fairer outcomes.
Why is it important to address human bias in tech development?
Human bias can easily creep into the design process, influencing outcomes.
Why is diversity important in the tech industry?
A wider range of perspectives leads to less biased and more inclusive systems.
What is the impact of biased criminal risk assessment tools?
Can lead to unfair judicial decisions, disproportionately affecting certain groups.
What is the impact of biased facial recognition systems?
Can result in misidentification or exclusion, especially for underrepresented groups.
What is the impact of biased recruiting algorithms?
Can perpetuate gender and racial imbalances in the workforce.
Why is it important to use fairness metrics?
Ensures the system doesn't discriminate and promotes equitable outcomes.
What is the role of historical data in creating bias?
Historical data often reflects existing societal biases, which can be perpetuated by algorithms.