AI explainability statement for customers receiving AI decisions

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-ethics-ai-explainability-statement-for-customers-creators

1. A company wants to include an explanation of an AI decision in a customer notification. Which element should definitely be included?

   1. The training dataset characteristics used to build the model
   2. How the customer can request a human review of the decision
   3. The specific algorithm version number used to make the decision
   4. The exact mathematical formula the model applied

2. Which statement about AI explainability would most likely erode customer trust if included in a customer-facing document?

   1. You may request a human review of this decision within 30 days
   2. The AI considered your payment history and account age in reaching this decision
   3. This decision is 99.7% accurate based on our sophisticated algorithm
   4. The factors most important to this decision were X, Y, and Z

3. Why should an AI explainability statement avoid promising a specific outcome from the appeal process?

   1. Appeals are handled by the same AI system, so outcomes are predetermined
   2. Promising outcomes would violate regulations around disclosure accuracy
   3. The AI cannot know in advance what a human reviewer will decide
   4. Appeals rarely change AI decisions so saying so would discourage customers

4. What is the main risk of including detailed model architecture information in a customer-facing explainability statement?

   1. Customers won't understand technical details
   2. It would make the statement too long to read
   3. It could reveal information that helps people game the system
   4. Customers would lose confidence in automated decisions

5. Which of the following best describes what AI does well when drafting explainability statements?

   1. It can translate technical model details into plain language descriptions of inputs and factors
   2. It can guarantee that the explanation will satisfy regulatory requirements
   3. It can determine exactly how much each factor influenced the specific decision
   4. It can accurately predict what a human reviewer will decide on appeal

6. Why might an explainability statement that claims the company fully understands how its AI made a decision actually backfire?

   1. It would make the company liable for any decision errors
   2. It would bore customers with unnecessary detail
   3. Most AI systems are too complex for any single person to fully understand, so this claim is unbelievable
   4. Customers prefer mysterious decision-making processes

7. An AI explainability statement should describe the data considered in a decision primarily to:

   1. Help customers understand what information influenced the decision about them
   2. Show customers how sophisticated the AI system is
   3. Demonstrate that the AI considered more factors than a human would
   4. Prove that the company follows data protection regulations

8. What should an explainability statement say about what a human reviewer can do?

   1. The human reviewer will always overturn the decision if the customer appeals
   2. The human reviewer will confirm whatever the AI decided
   3. The human reviewer will explain exactly how the AI calculated its decision
   4. The human reviewer has full access to the same information as the AI and can make an independent judgment

9. Which statement best illustrates the concept of 'honest uncertainty' in AI explainability?

   1. Our AI is 100% accurate and never makes mistakes
   2. Our AI uses advanced machine learning that we fully understand
   3. This decision may not be perfectly accurate and you have the right to request a human review
   4. The decision was made automatically based on factors including your payment history

10. A regulatory compliance team should review explainability statements because:

    1. AI-generated text is always inaccurate
    2. Customers prefer communications reviewed by lawyers
    3. AI systems cannot write complete sentences without human review
    4. The statements constitute regulated disclosures that may have legal requirements

11. What is wrong with saying an AI decision was made using 'magic' or 'sophisticated algorithms' in an explainability statement?

    1. The word 'magic' is trademarked
    2. Customers don't care about algorithms
    3. These are metaphors that overstate certainty and explain nothing
    4. The statement would be too short

12. Which scenario best demonstrates the tension between transparency and security in AI explainability?

    1. Deciding whether to tell customers their credit score was a factor
    2. Deciding whether to use a formal or casual tone
    3. Deciding whether to include the company's contact information
    4. Deciding whether to publish the exact neural network architecture

13. An explainability statement includes this line: 'This decision is final and cannot be appealed.' This is problematic because:

    1. It makes the statement too long
    2. It uses the word 'final' which is confusing
    3. AI decisions are always correct anyway
    4. Customers always have the right to some form of review or appeal

14. Why is it important to explain what a human reviewer can do differently from the AI?

    1. Human reviewers don't actually do anything differently
    2. It helps customers understand the value and purpose of the appeal process
    3. It reduces the number of appeals customers will file
    4. It makes the company seem more technologically advanced

15. A good customer-facing explainability statement should focus on:

    1. Criticizing competitors' explainability practices
    2. Technical accuracy over plain language
    3. Mathematical certainty and precise percentages
    4. Plain language descriptions of inputs, factors, and appeal options