Trustworthy AI

Question 1

Why Trustworthy AI

Answer 1

Is it Accurate
Is it Fair
is it easy to understand
is it accountable
did anyone tamper with it

Question 2

Procedural Justice

Answer 2

Perceived fairness of the rules and processes to determine outcomes

Question 3

Distributive Justice

Answer 3

perceived fairness of the outcomes or resource allocations themselves

Question 4

AI Fairness considerations

Answer 4

1) The impact of outcomes
2) Nature and scope of decisions
3) Operational complexity and limits to scale
4) Compliance and governance capabilities

Question 5

Fairness: Allocation

Answer 5

Allocation: used as classification problem

Question 6

Fairness: Group Fairness
Individual Fairness

Answer 6

Groups defined by protected attributes receiving similar treatments
individual: similar individuals receiving similar outcomes

Question 7

Fairness Metrics

Answer 7

Statistical Parity Difference
Equal Opportunity Difference

Question 8

Pillars of Trustworthy AI

Answer 8

• *Explainability** is the ability of the AI model to explain how and why it arrived at a particular decision
• *Fairness** is the ability of the AI model to be free of bias in its decisions and to avoid unfair treatment of certain groups
• *Robustness** is the ability of the AI model to be safe and secure and not be vulnerable to any tampering or compromising the data they are trained on.
• *Transparency** is the ability to disclose information to increase the understanding of how an AI model or service was created and deployed
• *Governance** is the ability to direct, manage and control the AI activities throughout the AI lifecycle.

Answer 9

test, evaluation, verification, and validation (TEVV)

Question 10

AI Risks and Trustworthiness (NIST)

Answer 10

• valid and reliable
• safe
• secure and resilient
• accountable and trans- parent
• explainable and interpretable
• privacy-enhanced
• fair with harmful bias managed

Question 11

Validation

Answer 11

Validation is the “confirmation, through the provision of objective evidence, that the re- quirements for a specific intended use or application have been fulfilled” (Source: ISO 9000:2015).

Question 12

Reliability

Reliability

Answer 12

Reliability is defined in the same standard as the “ability of an item to perform as required, without failure, for a given time interval, under given conditions” (Source: ISO/IEC TS 5723:2022). Reliability is a goal for overall correctness of AI system operation under the conditions of expected use and over a given period of time, including the entire lifetime of the system.

Question 13

Accuracy

Answer 13

Accuracy is defined by ISO/IEC TS 5723:2022 as “closeness of results of observations, computations, or estimates to the true values or the values accepted as being true.”

Question 14

Robustness or generalizability

Answer 14

Robustness or generalizability is defined as the “ability of a system to maintain its level of performance under a variety of circumstances”

Answer 15

AI systems, as well as the ecosystems in which they are deployed, may be said to be re- silient if they can withstand unexpected adverse events or unexpected changes in their envi- ronment or use – or if they can maintain their functions and structure in the face of internal and external change and degrade safely and gracefully when this is necessary

Question 16

what are: Common security concerns?

Answer 16

Common security concerns relate to adversarial examples, data poisoning, and the exfiltration of models, training data, or other intellectual property through AI system endpoints. AI systems that can maintain confidentiality, integrity, and availability through protection mechanisms that prevent unauthorized access and use may be said to be secure

Question 17

Security and resilience

Answer 17

Security and resilience are related but distinct characteristics. While resilience is the abil- ity to return to normal function after an unexpected adverse event, security includes re- silience but also encompasses protocols to avoid, protect against, respond to, or recover from attacks. Resilience relates to robustness and goes beyond the provenance of the data to encompass unexpected or adversarial use (or abuse or misuse) of the model or data.

Question 18

Transparency, explainability, and interpretability

Answer 18

• Transparency: “what happened” in the system.
• Explainability: “how” a decision was made in the system.
• Interpretability: “why” a decision was made by the system and its meaning or context to the user.

Question 19

Bias

Answer 19

• systemic
• computational and statistical
• and human-cognitive.