Module 2: AI Impacts on People and Responsible AI Principles: Risks and Harms

Question 1

Name some potential individual harms from AI.

Answer 1

• Civil liberties
• Rights
• Physical or psychological safety
• Economic opportunity

Question 2

Implicit Bias

Answer 2

Discrimination or prejudice toward a particular group or individual.

Question 3

Sampling Bias

Answer 3

Data gets skewed toward a subset of the group, so it may favor a subset.

Question 4

Temporal Bias

Answer 4

Model is trained and works well now, but may not work later.

Question 5

Overfitting to Training Data

Answer 5

Model works for training data, but not for new data.

Question 6

Edge Cases and Outliers

Answer 6

Data outside the boundaries of the training dataset.

Question 7

Areas of potential discrimination

Answer 7

• Employment and hiring
• Insurance and social benefits
• Housing
• Education
• Credit
• Differential pricing of goods and services

Question 8

Name some AI privacy concerns.

Answer 8

• Using personal data in training and use of systems
• Easy to recombine data and reidentify individuals
• Appropriation of personal data for model training
• Inferences could identify the wrong individual
• Lack of transparency
• Inaccurate models
• Information about protected classes is often considered sensitive

Question 9

What is an Inference?

Answer 9

An AI system model that makes predictions or decisions.

Question 10

What are some economic risks of AI?

Answer 10

• Job loss
• AI-driven discriminatory hiring practices
• Job opportunities may fail to reach key demographics due to AI driven tools

Question 11

What is the definition of group harms from AI?

Answer 11

Harm to a group such as discrimination against a population subgroup.

Question 12

Name some potential group harms from AI.

Answer 12

• Mass surveillance (especially for marginalized groups)
• Freedom of assembly and protest (due to tracking and profiling)
• Deepening of racial and socio-economic inequities
• Societal harm: Harm to democratic participation and process.
• Spread of disinformation, fostering idealogical bubbles (echo chambers)
• Deepfakes
• Safety issues
• Lack of human oversight

Question 13

What are deepfakes?

Answer 13

Audio, video or images altered to portray a different reality

Question 14

How can AI help the environment?

Answer 14

• Self driving cars (reduce emissions)
• Agriculture (high crop yields)
• Disaster relief (using satellites)
• Weather forecasting

Question 15

What are some environmental harms from AI?

Answer 15

• High carbon emissions
• High energy consumption
• High water usage (lithium batteries)

Question 16

What are some mitigations for the harms from AI?

Answer 16

• Alternative energy sources such as batteries
• Implement appropriate AI governance and processes

Question 17

Name some potential organizational harms from AI.

Answer 17

• Reputational
• Cultural and societal
• Economic (litigation, remediation)
• Acceleration risk
• Legal and regulatory

Question 18

Who are some of the key AI stakeholders?

Answer 18

• AI risk owner (the business unit - whether revenue generating or not)
• Customers or others impacted
• AI risk manager
• AI subject matter expert (data scientist, AI engineer)
• Compliance/Legal/Privacy
• IT leaders (CISO, CTO, CIO)
• Executive business leaders
• Board of Directors

Question 19

What are some steps an organization can take to mitigate the organizational risks of AI?

Answer 19

• Review requirements
• Identify gaps
• Conduct ongoing monitoring