Ethical AI and Data Privacy practice questions

Empathy

Why wrong: Empathy is about designing for human benefit, but the core issue here is erroneous scoring.

Safety

Why wrong: Safety focuses on preventing harmful outputs, which is not directly the issue here.

Transparency

Why wrong: Transparency is about explaining AI decisions, not about correctness of predictions across segments.

Allow the AI to send emails automatically if confidence is high

Why wrong: Automated sending without human review may violate GDPR's right to human intervention.

Disable the AI composer entirely to avoid GDPR risk

Why wrong: Overly restrictive; human oversight can mitigate risk without disabling the feature.

Anonymize customer data before drafting

Why wrong: Anonymization is a privacy measure but does not address the right to human intervention for automated decisions.

To increase the speed of AI responses

Why wrong: Toxicity detection adds a processing step and may slightly reduce speed.

To monitor user interactions for compliance with data protection laws

Why wrong: Compliance monitoring is broader; toxicity detection is specifically about harmful language.

To identify and mask PII in user prompts

Why wrong: PII masking is a separate component of the Trust Layer.

Keep both features to maximize information

Why wrong: Keeping both can inflate variance and make the model less interpretable.

Apply principal component analysis to combine them

Why wrong: PCA creates new components that are harder to interpret in the business context.

Increase the regularization strength

Why wrong: Regularization shrinks coefficients but does not remove the correlation between the two features, so multicollinearity-driven coefficient instability remains. Removing one of the correlated features eliminates the redundancy directly.

PII masking

Why wrong: PII masking protects sensitive data but does not prevent data retention.

Grounding

Why wrong: Grounding connects AI to CRM data but does not address model training.

Toxicity detection

Why wrong: Toxicity detection filters harmful content; it is unrelated to data retention.

Use only historical data from the past six months

Why wrong: Short time windows may not capture enough data for robust model training and could introduce seasonality bias.

Include as many features as possible to maximize predictive power

Why wrong: Including too many features, especially sensitive ones, can introduce bias and reduce interpretability.

Delete the lead from the CRM to avoid influencing future predictions

Why wrong: Deleting data removes evidence and does not address the underlying model bias.

Manually override the score to 50 and move on

Why wrong: Manual override without investigation does not fix the root cause and may reintroduce bias.

Contact Salesforce support to disable the AI model

Why wrong: Disabling the model is drastic; the rep should first investigate and escalate.

Automatic model retraining

Why wrong: Automatic retraining would use customer data, violating privacy commitments.

Data masking for all PII

Why wrong: Data masking is a feature, but zero data retention is the key privacy protection that prevents model training on customer data.

Open-source model deployment

Why wrong: Open-source deployment does not directly address data privacy.

Share the full training dataset for transparency

Why wrong: Sharing the dataset raises privacy concerns and is not an explanation.

Allow stakeholders to retrain the model themselves

Why wrong: Retraining does not explain existing predictions.

Provide the raw model weights and coefficients

Why wrong: Raw weights are too technical for non-technical audiences.

The model's algorithm is inherently biased against certain regions

Why wrong: Algorithms are not inherently biased; bias comes from data or feature selection.

The model overfits to the training data

Why wrong: Overfitting would cause poor generalization but not necessarily bias against specific postal codes.

The model was not trained long enough

Why wrong: Training duration does not cause systematic bias against specific groups.

Store all customer chat transcripts permanently

Why wrong: Storing transcripts may violate data minimization; it's not required for explanation.

Use only rule-based bots to avoid AI complications

Why wrong: Rule-based bots may not need explanation, but the goal is to use AI compliantly, not avoid it.

Allow customers to opt out of all AI interactions

Why wrong: Opt-out is a separate right (right to object), not directly the right to explanation.

Sharing Rules

Why wrong: Sharing rules govern record visibility, not data masking.

Permission Sets

Why wrong: Permission sets control access, not data masking.

Data Export Service

Why wrong: Data export is for backup, not masking.

The emails are sent automatically to maximize efficiency

Why wrong: Sending automatically without review violates the empathy principle.

The emails are sent only to customers who opted in to AI interactions

Why wrong: Opt-in is needed, but human review is still required per empathy.

The AI model is retrained daily on new customer data

Why wrong: Retraining does not replace human oversight.

To limit the AI's responses to pre-approved templates

Why wrong: Grounding does not restrict to templates; it uses live data.

To encrypt all data sent to the AI model

Why wrong: Encryption is separate; grounding is about data relevance.

To ensure the AI only uses data from Salesforce, not third-party sources

Why wrong: Grounding uses CRM data, but the key is relevance, not source exclusivity.

Include all available fields to maximize model accuracy

Why wrong: Including unnecessary fields violates data minimization.

Obtain explicit consent for every field used

Why wrong: Consent is required, but minimization is separate; consent does not justify using irrelevant fields.

Anonymize all data before training, including purchase amounts

Why wrong: Anonymization is good, but purchase amounts may be necessary; excluding them could harm accuracy unnecessarily.

Safety

Why wrong: Safety is about preventing harmful outputs, but empathy is broader about human benefit.

Accuracy

Why wrong: Accuracy is about correctness, not human benefit.

Honesty

Why wrong: Honesty is about transparency, not benefit.

Store all AI-generated emails indefinitely for auditing

Why wrong: Storing indefinitely violates data minimization.

Obtain consent only once at account creation

Why wrong: Consent must be ongoing and specific; one-time consent may not suffice.

It lists all AI features available in Salesforce

Why wrong: Feature lists are in product documentation, not the DPA.

It provides a template for customers to build their own AI models

Why wrong: The DPA is not a model building template.

It is a user guide for configuring AI privacy settings

Why wrong: The DPA is a legal document, not a user guide.

Adjust the lead scoring formula manually to increase scores for that region

Why wrong: Manual adjustments are not a sustainable or transparent fix; they may mask underlying data issues.

Remove the region field from the model

Why wrong: Removing the field may eliminate the bias but also removes potentially useful information; data augmentation is preferable.

Accept the scores as-is because the model is certified

Why wrong: No model is immune to bias; ongoing evaluation and improvement are necessary.

Provide a chatbot that can explain the recommendations

Why wrong: While helpful, this is not a substitute for upfront disclosure. Honesty requires proactive communication, not just reactive explanation.

Use a generic label like 'Recommended for you'

Why wrong: Vague labels do not inform users that AI is involved; Honesty requires explicit mention of AI.

Do not mention AI; just show the recommendations

Why wrong: Omitting disclosure misleads users into thinking recommendations are human-curated, which is dishonest.