1|Page
WGU D685 Practical Applications of Prompt Engineering Study
Guide | Comprehensive Practice Questions, Real-World AI
Prompting Scenarios, Detailed Explanations & Latest Exam
Preparation (2025–2026)
EXAM COVERAGE SUMMARY
This examination covers the fundamental concepts of artificial intelligence including machine
learning, neural networks, deep learning, generative AI, and large language models. You will be
tested on prompt engineering techniques such as zero-shot, few-shot, chain-of-thought, tree-of-
thought, self-consistency, least-to-most, cognitive verifier pattern, and generated knowledge
prompting. The exam addresses AI limitations including fundamental constraints, practical
challenges, and societal concerns like bias, ethics, privacy, and safety. Additional topics include
natural language processing, computer vision, robotics, data types and processing, AI tools and
applications, responsible AI principles, and strategies for mitigating bias and ensuring
transparency in AI systems.
1. A hospital implements an AI system to assist radiologists in detecting abnormalities in medical
images. What fundamental limitation of AI presents the greatest challenge in this scenario?
A) AI systems can process images faster than humans
B) AI systems require minimal training data to function effectively
C) The AI cannot understand the emotional impact of a diagnosis on the patient
D) The AI lacks common sense reasoning about human anatomy beyond training
data (CORRECT ANSWER)
,2|Page
2. When developing an AI-powered recruitment tool, the development team discovers the system
consistently recommends male candidates over equally qualified female candidates. What type of
bias is most likely responsible for this outcome?
A) Measurement bias from poorly designed interview questions
B) Confirmation bias from developers testing only certain scenarios
C) Algorithmic bias from historical hiring data reflecting past discriminatory
practices (CORRECT ANSWER)
D) Sampling bias from collecting data from only one geographic region
3. A user wants to generate creative short stories about space exploration. Which temperature
setting would produce the most diverse and imaginative story outputs?
A) Temperature set to 0.0 for maximum predictability
B) Temperature set to 0.3 for moderate creativity
C) Temperature set to 0.9 for high randomness and creativity (CORRECT ANSWER)
D) Temperature set to 0.1 for conservative responses
4. An AI model incorrectly identifies a non-threatening email as spam and moves it to the junk
folder. What type of error has the AI system made?
A) True negative where the system correctly identified non-spam
B) False positive where the system incorrectly flagged legitimate content (CORRECT
ANSWER)
C) True positive where the system accurately detected spam
D) False negative where the system missed actual spam
5. A researcher wants to help an LLM solve a complex mathematical problem by having it break
down the solution into intermediate steps. Which prompting technique should the researcher
employ?
A) Zero-shot prompting with no examples provided
B) Chain-of-thought prompting to reveal step-by-step reasoning (CORRECT ANSWER)
C) Few-shot prompting with multiple solved examples
D) Generated knowledge prompting for background information
,3|Page
6. What distinguishes structured data from unstructured data in AI applications?
A) Structured data is always more accurate than unstructured data
B) Structured data is organized in databases while unstructured data includes text, images,
and audio (CORRECT ANSWER)
C) Unstructured data requires less processing power than structured data
D) Structured data cannot be used for machine learning applications
7. A customer service chatbot provides a detailed return policy for a product the customer never
purchased. This error is an example of:
A) True positive response confirming customer suspicion
B) False positive where the AI generated irrelevant information (CORRECT ANSWER)
C) True negative where the AI correctly avoided incorrect information
D) Sampling error from insufficient training data
8. Which advanced prompting technique encourages an LLM to explore multiple reasoning paths
before generating a final response?
A) Zero-shot prompting for generalized responses
B) Few-shot prompting with limited examples
C) Chain-of-thought prompting for linear reasoning
D) Tree-of-thought prompting for exploring various possibilities (CORRECT ANSWER)
9. When using generative AI to create marketing content, what is the primary benefit of including
detailed descriptions in the prompt?
A) It reduces the need for human oversight of the output
B) It guarantees the AI will only produce factual information
C) It improves relevance and clarity of the generated content (CORRECT ANSWER)
D) It eliminates the possibility of biased outputs
10. An AI model trained predominantly on data from urban populations fails to accurately predict
health outcomes for rural communities. What type of bias does this represent?
, 4|Page
A) Confirmation bias in the training algorithm
B) Measurement bias in data collection methods
C) Sampling bias from non-representative training data (CORRECT ANSWER)
D) Algorithmic bias from model architecture choices
11. What is the primary function of neural networks in artificial intelligence systems?
A) To store large amounts of structured data efficiently
B) To recognize patterns and make predictions by learning from training data (CORRECT
ANSWER)
C) To replace human decision-making entirely
D) To eliminate all biases from training datasets
12. A user wants to create an AI image of a "peaceful mountain landscape at sunset with glowing
golden light and pine trees." Which element is most critical for generating the desired image?
A) Using only simple keywords without modifiers
B) Including descriptive modifiers for mood, lighting, and composition (CORRECT
ANSWER)
C) Avoiding any reference to color or lighting
D) Keeping the prompt under five words for best results
13. What distinguishes narrow AI from general AI in terms of capabilities?
A) Narrow AI can perform any intellectual task while general AI is limited
B) Narrow AI performs specific tasks while general AI would possess human-level
intelligence (CORRECT ANSWER)
C) Narrow AI requires less data than general AI systems
D) General AI currently exists and is widely available
14. A company wants to implement AI for data cleaning to prepare a large dataset for analysis.
How can generative AI improve the data cleaning process?
A) By automatically deleting all data that contains errors
B) By inferring corrections and detecting anomalies in the dataset (CORRECT ANSWER)
WGU D685 Practical Applications of Prompt Engineering Study
Guide | Comprehensive Practice Questions, Real-World AI
Prompting Scenarios, Detailed Explanations & Latest Exam
Preparation (2025–2026)
EXAM COVERAGE SUMMARY
This examination covers the fundamental concepts of artificial intelligence including machine
learning, neural networks, deep learning, generative AI, and large language models. You will be
tested on prompt engineering techniques such as zero-shot, few-shot, chain-of-thought, tree-of-
thought, self-consistency, least-to-most, cognitive verifier pattern, and generated knowledge
prompting. The exam addresses AI limitations including fundamental constraints, practical
challenges, and societal concerns like bias, ethics, privacy, and safety. Additional topics include
natural language processing, computer vision, robotics, data types and processing, AI tools and
applications, responsible AI principles, and strategies for mitigating bias and ensuring
transparency in AI systems.
1. A hospital implements an AI system to assist radiologists in detecting abnormalities in medical
images. What fundamental limitation of AI presents the greatest challenge in this scenario?
A) AI systems can process images faster than humans
B) AI systems require minimal training data to function effectively
C) The AI cannot understand the emotional impact of a diagnosis on the patient
D) The AI lacks common sense reasoning about human anatomy beyond training
data (CORRECT ANSWER)
,2|Page
2. When developing an AI-powered recruitment tool, the development team discovers the system
consistently recommends male candidates over equally qualified female candidates. What type of
bias is most likely responsible for this outcome?
A) Measurement bias from poorly designed interview questions
B) Confirmation bias from developers testing only certain scenarios
C) Algorithmic bias from historical hiring data reflecting past discriminatory
practices (CORRECT ANSWER)
D) Sampling bias from collecting data from only one geographic region
3. A user wants to generate creative short stories about space exploration. Which temperature
setting would produce the most diverse and imaginative story outputs?
A) Temperature set to 0.0 for maximum predictability
B) Temperature set to 0.3 for moderate creativity
C) Temperature set to 0.9 for high randomness and creativity (CORRECT ANSWER)
D) Temperature set to 0.1 for conservative responses
4. An AI model incorrectly identifies a non-threatening email as spam and moves it to the junk
folder. What type of error has the AI system made?
A) True negative where the system correctly identified non-spam
B) False positive where the system incorrectly flagged legitimate content (CORRECT
ANSWER)
C) True positive where the system accurately detected spam
D) False negative where the system missed actual spam
5. A researcher wants to help an LLM solve a complex mathematical problem by having it break
down the solution into intermediate steps. Which prompting technique should the researcher
employ?
A) Zero-shot prompting with no examples provided
B) Chain-of-thought prompting to reveal step-by-step reasoning (CORRECT ANSWER)
C) Few-shot prompting with multiple solved examples
D) Generated knowledge prompting for background information
,3|Page
6. What distinguishes structured data from unstructured data in AI applications?
A) Structured data is always more accurate than unstructured data
B) Structured data is organized in databases while unstructured data includes text, images,
and audio (CORRECT ANSWER)
C) Unstructured data requires less processing power than structured data
D) Structured data cannot be used for machine learning applications
7. A customer service chatbot provides a detailed return policy for a product the customer never
purchased. This error is an example of:
A) True positive response confirming customer suspicion
B) False positive where the AI generated irrelevant information (CORRECT ANSWER)
C) True negative where the AI correctly avoided incorrect information
D) Sampling error from insufficient training data
8. Which advanced prompting technique encourages an LLM to explore multiple reasoning paths
before generating a final response?
A) Zero-shot prompting for generalized responses
B) Few-shot prompting with limited examples
C) Chain-of-thought prompting for linear reasoning
D) Tree-of-thought prompting for exploring various possibilities (CORRECT ANSWER)
9. When using generative AI to create marketing content, what is the primary benefit of including
detailed descriptions in the prompt?
A) It reduces the need for human oversight of the output
B) It guarantees the AI will only produce factual information
C) It improves relevance and clarity of the generated content (CORRECT ANSWER)
D) It eliminates the possibility of biased outputs
10. An AI model trained predominantly on data from urban populations fails to accurately predict
health outcomes for rural communities. What type of bias does this represent?
, 4|Page
A) Confirmation bias in the training algorithm
B) Measurement bias in data collection methods
C) Sampling bias from non-representative training data (CORRECT ANSWER)
D) Algorithmic bias from model architecture choices
11. What is the primary function of neural networks in artificial intelligence systems?
A) To store large amounts of structured data efficiently
B) To recognize patterns and make predictions by learning from training data (CORRECT
ANSWER)
C) To replace human decision-making entirely
D) To eliminate all biases from training datasets
12. A user wants to create an AI image of a "peaceful mountain landscape at sunset with glowing
golden light and pine trees." Which element is most critical for generating the desired image?
A) Using only simple keywords without modifiers
B) Including descriptive modifiers for mood, lighting, and composition (CORRECT
ANSWER)
C) Avoiding any reference to color or lighting
D) Keeping the prompt under five words for best results
13. What distinguishes narrow AI from general AI in terms of capabilities?
A) Narrow AI can perform any intellectual task while general AI is limited
B) Narrow AI performs specific tasks while general AI would possess human-level
intelligence (CORRECT ANSWER)
C) Narrow AI requires less data than general AI systems
D) General AI currently exists and is widely available
14. A company wants to implement AI for data cleaning to prepare a large dataset for analysis.
How can generative AI improve the data cleaning process?
A) By automatically deleting all data that contains errors
B) By inferring corrections and detecting anomalies in the dataset (CORRECT ANSWER)
