AI Associate Ethical Considerations of AI Practice Test 2 — 15 Questions

Question 1

Refer to the exhibit. This JSON snippet is from the Einstein Trust Layer configuration. What is the purpose of this configuration?

Accepted Answer

To detect biased predictions based on gender and race. The Einstein Trust Layer includes bias detection and audit capabilities that surface when model outputs or training data reflect protected attributes such as gender or race. The configuration shown enables monitoring and flagging of biased predictions so they can be reviewed, not blocking or removing the attributes. This supports responsible AI governance and regulatory compliance.

Answer

To block all predictions involving gender or race

Answer

To anonymize gender and race data

Answer

To remove gender and race from the model

Question 2

Refer to the exhibit. A Salesforce developer configures the Einstein Trust Layer as shown. What is the primary purpose of enabling data masking?

Accepted Answer

To anonymize personally identifiable information (PII) in the model output.. Enabling data masking in the Einstein Trust Layer ensures that personally identifiable information (PII) is anonymized before the model output is returned to the user. This protects sensitive data from exposure in AI-generated responses, which is a core requirement for privacy compliance and responsible AI use.

Answer

To improve the accuracy of sentiment analysis.

Answer

To reduce latency of the AI response.

Answer

To comply with Salesforce's service-level agreement.

Question 3

According to Salesforce's AI ethics principles, which three pillars should guide the development of AI applications?

Accepted Answer

Accountability.. Salesforce's AI ethics principles are built around five pillars, and among the options listed, Accountability (A), Transparency (B), and Fairness (E) are three of them. Accountability (A) is correct because Salesforce requires that organizations designate responsible parties who can explain and answer for AI-driven decisions and outcomes. Transparency (B) is correct because Salesforce mandates that AI systems be open about how they work, what data they use, and how they reach decisions, so users can understand and trust them. Fairness (E) is correct because Salesforce commits to designing AI that avoids bias and treats all users equitably, promoting inclusive and unbiased outcomes. Profitability (C) and Scalability (D) are not among Salesforce's AI ethics pillars; while commercially relevant, they are business goals rather than ethical principles guiding responsible AI development.

Answer

Profitability.

Answer

Scalability.

Question 4

Refer to the exhibit. An admin sees this error in the Einstein activity log. What is the most likely cause?

Accepted Answer

Data masking is configured but not applied to the sentiment analysis model.. The error indicates that sentiment analysis is not enabled, but PII is leaking, likely because data masking is not applied to the model generating the output.

Answer

Sentiment analysis is generating PII that data masking cannot hide.

Answer

The model output should not contain any text.

Answer

The Einstein Trust Layer is completely disabled.

Question 5

To comply with Salesforce's AI ethics principles when using Einstein Bots, which two practices should be implemented?

Accepted Answer

Allow users to escalate to a human agent.. Option A is correct because Salesforce's AI ethics principles emphasize transparency and human oversight, so Einstein Bots must provide a clear path for users to escalate to a live human agent rather than trapping them in automated interactions. Option D is correct because users have the right to know they are interacting with an AI system, and disclosing bot identity directly satisfies the transparency and honesty principles. Option B is incorrect because fully autonomous decision-making without human review violates the accountability and human-centered values that Salesforce's ethical AI guidelines require. Option C is incorrect because retaining conversation transcripts indefinitely conflicts with data minimization and privacy-by-design principles, which favor limited retention periods. Option E, while a sound privacy practice, is not one of the two practices specifically called out by Salesforce's AI ethics principles for Einstein Bots in this context.

Answer

Use the bot to make all customer decisions autonomously.

Answer

Store all conversation transcripts indefinitely.

Answer

Minimize data collection to only what is necessary.

Question 6

Refer to the exhibit. A company configures a Prompt Builder policy for Einstein GPT. What is the primary role of the 'checkPromptOutput' flag?

Accepted Answer

To scan the generated text against the banned words list.. The 'checkPromptOutput' flag in a Prompt Builder policy for Einstein GPT is specifically designed to scan the generated text against a banned words list. This ensures that the AI output does not contain prohibited or sensitive terms, aligning with ethical and compliance requirements. It is a content filtering mechanism, not a logging, token-limiting, or human-review function.

Answer

To log all prompts for audit purposes.

Answer

To limit the total number of tokens in the generated response.

Answer

To send the output to a human reviewer before sending.

Question 7

An organization is deploying an AI system for loan decisions. They want to ensure human oversight. Which is the best implementation?

Accepted Answer

The system only flags edge cases for human review.. Option D is correct because it implements human oversight through a human-in-the-loop exception model: the AI handles routine cases autonomously while flagging edge cases for human review, which is the most practical and scalable oversight pattern for loan decisions. This keeps humans meaningfully involved in the highest-risk or ambiguous decisions without creating a bottleneck for every routine application. Option A is wrong because fully autonomous operation with only audit logging provides no real-time human oversight. Option B is wrong because post-hoc review of high-value loans is reactive and does not prevent harmful automated decisions. Option C is wrong because requiring human approval for all decisions is overly burdensome and defeats the efficiency purpose of the AI system.

Answer

The system operates fully autonomously but logs decisions for audit.

Answer

The system makes decisions automatically, with post-hoc review only for high-value loans.

Answer

The system provides recommendations, and a human must approve all decisions.

Question 8

Refer to the exhibit. A Salesforce admin is reviewing an AI model's fairness report. Which action should the admin take?

Accepted Answer

Retrain the model because the equal opportunity score is below threshold.. In Salesforce's AI fairness report, the equal opportunity score measures whether the model's true positive rate is equal across demographic groups. If this score falls below the configured threshold, the model exhibits disparate impact and should be retrained with bias mitigation techniques (e.g., reweighting, resampling, or removing biased features). Retraining is the correct action because it addresses the root cause rather than masking the metric.

Answer

Remove the email_engagement feature to improve fairness.

Answer

Increase the bias threshold to 0.9.

Answer

Deploy the model because all metrics exceed the threshold.

Question 9

Which THREE are core principles in Salesforce's AI ethics framework?

Accepted Answer

Fairness. Salesforce's AI ethics framework is built around a set of core principles, and among the listed options the correct ones are Fairness (B), Transparency (D), and Privacy (E). Fairness (B) is correct because Salesforce's ethical AI guidelines require that AI systems be designed to avoid bias and treat people equitably, which is a foundational commitment of the framework. Transparency (D) is correct because Salesforce emphasizes that AI decisions and data usage should be explainable and open to scrutiny, so users can understand how AI systems operate. Privacy (E) is correct because protecting personal data and giving individuals control over their information is a central pillar of Salesforce's responsible AI approach, reflected in its privacy-by-design commitments. The unmarked options do not belong: Accuracy (A) is a desirable quality attribute of AI models but is not one of the named core ethical principles in Salesforce's framework, and Profitability (C) is a business outcome rather than an ethical principle and is not part of the framework at all.

Answer

Accuracy

Answer

Profitability

Question 10

Refer to the exhibit. Which ethical principle is most directly violated?

Accepted Answer

Privacy. Privacy, because the scenario in the exhibit involves handling or exposing personal data in a way that breaches an individual's right to control their own information, which is the core of the privacy principle. Privacy is most directly violated when confidential or personally identifiable information is accessed, disclosed, or used without proper authorization or consent. Transparency is not the best fit because it concerns openness about data practices rather than the improper exposure of data itself. Accountability relates to taking responsibility for actions, and fairness concerns impartial and non-discriminatory treatment, neither of which is the primary violation shown here.

Answer

Transparency

Answer

Accountability

Answer

Fairness

Question 11

Refer to the exhibit. A company has the Einstein LLM policy shown. What is the primary ethical gap in this policy?

Accepted Answer

It allows LLM for summarization without requiring human review for high-risk summaries. The policy allows LLM use for summarization without requiring human review for high-risk summaries, which is the primary ethical gap because high-stakes summaries (e.g., medical, legal, or safety-related content) can propagate errors or omissions that directly affect people. Ethical LLM governance typically mandates human-in-the-loop review proportional to the risk of the use case, so exempting high-risk summarization from review creates an accountability and harm-mitigation gap. Option A is not the primary gap because the policy's explicit block on employment decisions addresses a specific high-risk domain, and content-generation bias is a broader concern not framed by the exhibit. Option B is too generic; the exhibit's specific flaw is the missing human review for high-risk summaries, not the absence of fairness audits in general. Option C misstates the policy's inconsistency, since the gap is about summarization review, not a clinical-versus-legal review asymmetry.

Answer

It blocks employment decisions but allows content generation which could be biased

Answer

It does not include fairness audits for allowed use cases

Answer

It requires human review for clinical diagnosis but not for legal advice

Question 12

Which THREE strategies can help mitigate bias in an AI model? (Choose three.)

Accepted Answer

Remove protected attributes from training data. Option A is correct because removing protected attributes (such as race, gender, or age) from the training data prevents the model from directly learning and reproducing discriminatory patterns tied to those attributes, reducing disparate impact. Option D is correct because using diverse and representative training data ensures that all relevant subgroups are adequately represented, which prevents the model from overfitting to a dominant group and producing biased predictions. Option E is correct because applying fairness metrics (e.g., demographic parity, equalized odds, or disparate impact ratio) during model evaluation quantitatively detects and helps correct bias before deployment. Option B is incorrect because focusing on majority group data would worsen bias by underrepresenting minority groups and skewing model performance. Option C is incorrect because randomizing model outputs does not address the underlying causes of bias and would degrade model accuracy and reliability without improving fairness.

Answer

Focus training on majority group data for accuracy

Answer

Randomize a portion of model outputs

Question 13

A company is implementing Salesforce Einstein AI for lead scoring. Which TWO actions align with ethical AI practices?

Accepted Answer

Provide clear documentation on how the model makes predictions.. Option D is correct because providing clear documentation on how the model makes predictions supports transparency and explainability, which are core ethical AI principles; stakeholders must be able to understand the factors and logic behind lead scores, especially under regulations like GDPR's right to explanation. Option E is correct because regularly auditing the model for biased outcomes is essential to detect and mitigate algorithmic bias, ensuring fair treatment of leads and compliance with anti-discrimination and responsible AI standards. Option A is not ethical because training on historical data without review can perpetuate existing biases and violate data governance requirements. Option B is not the best fit here because ethical AI is not limited to using only non-sensitive personal data; sensitive data may be used lawfully with proper consent, safeguards, and purpose limitation, and the statement is overly restrictive rather than a complete ethical practice. Option C is not ethical because limiting model access to only senior management reduces transparency, accountability, and oversight, which conflicts with ethical AI governance.

Answer

Use historical data without review to train the model.

Answer

Ensure the model uses only non-sensitive personal data.

Answer

Limit model access to only senior management.

Question 14

Refer to the exhibit. A Salesforce AI Associate is reviewing the AI model evaluation data. Which TWO ethical concerns should the associate identify?

Accepted Answer

The demographic parity difference of 0.12 indicates potential bias against a protected group.. Option A is correct because a demographic parity difference of 0.12 exceeds the commonly used 0.10 (or 0.05) fairness threshold, signaling that outcomes for a protected group differ materially from the reference group and thus indicating potential bias. Option E is correct because an audit trail that records a bias detection without documenting the remediation or mitigation actions leaves the ethical issue unresolved and undermines accountability and traceability. Option B is incorrect because 0.95 accuracy is generally a high performance value and, by itself, does not constitute an ethical concern. Option C is incorrect because a single approver is a governance/process weakness, not one of the two ethical concerns evidenced by the model evaluation data. Option D is incorrect because a disparate impact ratio of 0.85 is above the 0.80 four-fifths threshold, so it does not indicate adverse impact.

Answer

The model accuracy of 0.95 is too low to be deployed in production.

Answer

The model was approved by a single individual, which violates the principle of diversity in AI oversight.

Answer

The disparate impact ratio of 0.85 falls below the acceptable threshold of 0.80, indicating adverse impact.

Question 15

A marketing team uses Salesforce Einstein to generate personalized email content for a campaign. During review, the team notices that the model occasionally produces text that closely resembles copyrighted marketing copy from a competitor's website. The team wants to address this ethical and legal risk. Which action should the team take first?

Accepted Answer

Implement a human review step and use Einstein's content moderation and grounding features to check generated output before publishing.. The Einstein Trust Layer provides content moderation and grounding to detect problematic outputs, and human review adds a final safeguard. Together they address the risk of reproducing copyrighted material without discarding the generative capability. Data masking targets privacy, not intellectual property, and disabling the feature or raising temperature does not provide reliable protection.

Answer

Increase the model temperature setting so the generated text varies more and is less likely to match existing content.

Answer

Disable the Einstein generative AI feature entirely and revert to manually written email copy.

Answer

Enable Einstein Trust Layer data masking on all prompt fields before sending them to the LLM.