AI Associate Ethical AI and Data Privacy Practice Question
A sales team uses Einstein Lead Scoring and notices that the model gives disproportionately low scores to leads from a certain demographic group. The team suspects historical bias in the training data. Which THREE steps should they take to address this bias?
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Investigate the historical data to identify and correct labeling errors or sampling bias
To address bias, the team should audit the model for disparate impact, investigate and correct the historical data, and consider adding features that reduce reliance on biased proxies. Removing the model entirely is unnecessary, and manually adjusting scores is not a systemic fix.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Investigate the historical data to identify and correct labeling errors or sampling bias
Why this is correct
Correcting the root cause in data is essential for a sustainable fix.
- ✓
Audit the model for disparate impact on the affected demographic
Why this is correct
Auditing quantifies the bias and is the first step to understanding the problem.
- ✗
Manually increase scores for the affected group by a fixed percentage
Why it's wrong here
Manual adjustments are not transparent and may introduce new biases; a data-driven approach is better.
- ✗
Remove the model and rely on manual lead scoring
Why it's wrong here
Manual scoring may also be biased and is not scalable; the model can be improved instead of discarded.
- ✓
Add new features that are not correlated with the protected attribute
Why this is correct
Adding non-correlated features can help the model learn more accurate patterns and reduce reliance on biased proxies.
Go deeper
Related to this question
About these practice questions
Courseiva writes every AI Associate question from scratch — 753 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
Same concept, more angles
3 more ways this is tested on AI Associate
These questions test the same concept from different angles. Work through them to make sure you can recognise it however the exam phrases it.
Variation 1. A sales operations manager wants to use Einstein Lead Scoring to prioritize leads. They have historical data showing that leads from a certain postal code have a low conversion rate. However, they suspect the low conversion is due to a past marketing campaign that was poorly targeted, not the demographics of that area. What is the BEST way to ensure the AI model does not unfairly penalize leads from that postal code?
medium- A.Manually increase the lead scores for all leads from that postal code
- ✓ B.Audit the model for disparate impact on that postal code and retrain with updated labels that reflect the true conversion potential
- C.Remove the postal code field from the model training data entirely
- D.Use the model as-is because Salesforce AI is certified to be fair
Why B: Bias in historical data can lead to unfair predictions. The best approach is to audit the model for bias and retrain with corrected labels or additional features that capture the true drivers of conversion, rather than simply omitting the feature or adjusting scores manually.
Variation 2. A sales operations team notices that an Einstein Lead Scoring model assigns lower scores to leads from a particular geographic region, even though those leads have historically converted at a higher rate. What is the most likely cause of this discrepancy?
medium- A.The model is using outdated lead source fields
- B.The model's threshold for lead conversion is set too high
- ✓ C.The model has been trained on biased historical data that underrepresents or undervalues leads from that region
- D.The model is overfitted to noise in the data
Why C: The model was trained on historical data that may have reflected past biases (e.g., underinvestment in that region). The biased training data leads to unfair predictions. The correct action is to audit the model for bias.
Variation 3. A sales team is using Einstein Lead Scoring and notices that leads from a certain geographic region are consistently scored lower, even when the lead's profile matches high-performing customers from other regions. What is the most likely cause and recommended first step?
medium- A.The model is accurate; the region is genuinely underperforming. The team should accept the scores.
- ✓ B.The model may be biased due to underrepresented data from that region in the training set. Audit the model for bias and review training data demographics.
- C.The region has lower quality leads, so the scores are correct. The team should focus on other regions.
- D.Retrain the model with more data from that region, even if it means duplicating records.
Why B: The low scores for a specific region likely indicate bias in the model due to historical data imbalances. The first step should be to audit the model for bias using tools like Fairness in AI or reviewing training data distribution.
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This AI Associate practice question is part of Courseiva's free Salesforce certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI Associate exam.