Ghost in Machine https://www.pbs.org/wgbh/nova/article/ai-bias/

Fill in the blank: Algorithms may show accuracy rates that vary along age or ____ lines.

• Gender

• Economic

• Educational

• Geographic

Answer explanation

The correct choice is 'Gender' because algorithms can exhibit different accuracy rates based on gender, similar to how they may vary with age. This highlights potential biases in algorithmic performance across demographic lines.

What was the outcome when Joy Buolamwini's robot detected faces of her light-skinned roommates?

• It worked perfectly

• It failed completely

• It was inconsistent

• It was slightly inaccurate.

Answer explanation

Joy Buolamwini's robot successfully detected the faces of her light-skinned roommates, indicating that it worked perfectly in this instance.

What does Erik Learned-Miller emphasize as important for facial recognition vendors?

• Cost efficiency

• Exhaustive evaluations

• Speed

• User-friendly interface

Answer explanation

Erik Learned-Miller emphasizes that exhaustive evaluations are crucial for facial recognition vendors to ensure accuracy and reliability in their systems, making it the most important factor among the options provided.

What is a potential consequence of using biased data in algorithms?

• Increased accuracy

• Fair hiring practices

• Discrimination

• Faster processing

Answer explanation

Using biased data in algorithms can lead to discrimination, as the algorithms may reinforce existing prejudices and unfairly disadvantage certain groups, rather than promoting fairness or accuracy.

COMPAS is a tool used to evaluate ______.

• insurance rates

• credit scores

• criminal recidivism

• educational outcomes

Answer explanation

COMPAS is a risk assessment tool specifically designed to evaluate criminal recidivism, which refers to the likelihood of a convicted criminal reoffending. It does not assess insurance rates, credit scores, or educational outcomes.

According to the text, what is the challenge with current anti-discrimination laws?

• They are too strict

• They are comprehensive

• They have loopholes

• They are well-defined.

Answer explanation

The text highlights that current anti-discrimination laws have loopholes, which can undermine their effectiveness. This indicates that while the laws may exist, they are not foolproof due to these gaps.

What is the main obstacle in assessing algorithmic fairness according to the text?

• Lack of technology

• Multiple potential outcomes

• High cost

• Lack of interest

Answer explanation

The main obstacle in assessing algorithmic fairness is the presence of multiple potential outcomes, which complicates the evaluation of fairness across different scenarios and contexts.

What does Andrew Selbst specialize in according to the text?

• Data analysis

• Legal questions surrounding big data and machine learning

• Algorithm development

• Software engineering.

Answer explanation

Andrew Selbst specializes in legal questions surrounding big data and machine learning, making this the correct choice. The other options do not accurately reflect his area of expertise.

What is the main focus of the text?

• Algorithmic efficiency

• Algorithmic fairness and legal challenges

• Data collection

• Software development

Answer explanation

The text primarily discusses the implications of algorithmic fairness and the associated legal challenges, making this the main focus. Other options like efficiency, data collection, and software development are not the central themes.

What is a concern with automating processes in AI development?

• Increased cost

• Reduced bias

• Reduced human intervention

• Increased speed

Answer explanation

Automating processes in AI development can lead to reduced human intervention, which may result in a lack of oversight and accountability. This can exacerbate issues like bias and errors in AI systems.