Message roles (system, user, assistant) in Prompt Engineering / GenAI - Model Metrics & Evaluation

Metrics & Evaluation - Message roles (system, user, assistant)

Which metric matters for Message roles and WHY

When working with message roles like system, user, and assistant in AI chat models, the key metric is accuracy of role classification or correct role assignment. This ensures the model understands who is speaking and responds appropriately. For example, the system role sets instructions, the user role asks questions, and the assistant role replies. Correct role recognition helps the AI behave as expected.

Confusion matrix for role classification

      | Predicted \ Actual | System | User | Assistant |
      |--------------------|--------|------|-----------|
      | System             | 90     | 5    | 5         |
      | User               | 3      | 92   | 5         |
      | Assistant          | 2      | 4    | 94        |

      Total samples = 300

      Precision and recall per role:
      - System Precision = 90 / (90 + 3 + 2) = 90 / 95 = 0.947
      - System Recall = 90 / (90 + 5 + 5) = 90 / 100 = 0.9
      - User Precision = 92 / (5 + 92 + 4) = 92 / 101 = 0.910
      - User Recall = 92 / (3 + 92 + 5) = 92 / 100 = 0.92
      - Assistant Precision = 94 / (5 + 5 + 94) = 94 / 104 = 0.904
      - Assistant Recall = 94 / (2 + 4 + 94) = 94 / 100 = 0.94
    

Precision vs Recall tradeoff with examples

In message role classification, precision means how often the predicted role is correct. Recall means how many actual messages of a role were found.

If precision is low, the model confuses roles often, causing wrong responses. If recall is low, some messages are missed or misclassified, leading to ignored instructions or questions.

Example: If the system role is confused with user role, the AI might ignore important instructions. Here, high recall for system role is critical to catch all instructions.

Example: If user messages are misclassified as assistant, the AI might respond to itself, causing confusion. High precision for user role avoids this.

What good vs bad metric values look like

Good: Precision and recall above 90% for all roles means the model correctly identifies who is speaking most of the time.

Bad: Precision or recall below 70% means many messages are misclassified. For example, if system role recall is 50%, half of instructions are missed, causing poor AI behavior.

Common pitfalls in metrics

• Accuracy paradox: If one role is very common, high accuracy can hide poor performance on rare roles.
• Data leakage: Training data containing future messages can inflate metrics falsely.
• Overfitting: Model memorizes training roles but fails on new conversations.
• Ignoring role context: Metrics without considering conversation flow can mislead about real performance.

Self-check question

Your model has 98% accuracy but only 12% recall on the system role. Is it good for production? Why or why not?

Answer: No, it is not good. Even though overall accuracy is high, the model misses 88% of system messages (instructions). This means important instructions are ignored, causing the AI to behave incorrectly. High recall on system role is critical.

Key Result

High precision and recall for each message role ensure the AI correctly understands and responds to system, user, and assistant messages.

Practice

1. What is the main purpose of the system role in AI chat messages?

easy

A. To store conversation history

B. To provide user input or questions

C. To set instructions and guide the AI's behavior

D. To display the AI's answers

Solution

1. Step 1: Understand the roles in AI chat

   The system role is designed to give the AI instructions or context on how to respond.

2. Step 2: Differentiate from other roles

   The user role is for input, and the assistant role is for AI output, so setting instructions fits the system role.

3. Final Answer:

   To set instructions and guide the AI's behavior -> Option C

4. Quick Check:

   System role = instructions [OK]

Hint: System role sets AI instructions, user inputs, assistant replies [OK]

Common Mistakes:

• Confusing system role with user input
• Thinking assistant role sets instructions
• Assuming system role stores chat history

2. Which of the following is the correct way to represent a user message in a chat API call?

easy

A. {"role": "assistant", "content": "Hello!"}

B. {"role": "user", "content": "What is AI?"}

C. {"role": "system", "content": "You are helpful."}

D. {"role": "bot", "content": "I can help."}

Solution

1. Step 1: Identify the role for user input

   The user messages must have the role set to "user" to indicate input from the person.

2. Step 2: Check each option's role field

   {"role": "user", "content": "What is AI?"} uses "user" correctly, while others use "assistant", "system", or invalid "bot".

3. Final Answer:

   {"role": "user", "content": "What is AI?"} -> Option B

4. Quick Check:

   User role = "user" [OK]

Hint: User messages always have role "user" in JSON [OK]

Common Mistakes:

• Using "bot" instead of "assistant" or "user"
• Mixing system role with user messages
• Forgetting to set the role field

3. Given this message list sent to an AI model:

[{"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Hello!"}]

What role will the AI use to respond?

medium

A. system

B. user

C. moderator

D. assistant

Solution

1. Step 1: Understand message roles in conversation

   The AI responds with messages having the role "assistant" to show its output.

2. Step 2: Identify the role for AI replies

   Since the input messages include system and user roles, the AI's reply will be labeled as "assistant".

3. Final Answer:

   assistant -> Option D

4. Quick Check:

   AI replies use role "assistant" [OK]

Hint: AI responses always have role "assistant" [OK]

Common Mistakes:

• Thinking AI replies use "system" role
• Confusing user input role with AI output
• Assuming unknown roles like "moderator"

4. You wrote this message list for an AI chat:

[{"role": "system", "content": "Be concise."}, {"role": "user", "content": "Explain AI."}, {"role": "user", "content": "Tell me more."}]

Why might the AI ignore the second user message?

medium

A. Because there are two user messages without an assistant reply in between

B. Because the system role is missing

C. Because the assistant role is used twice

D. Because the user messages are empty

Solution

1. Step 1: Check message sequence rules

   AI expects alternating user and assistant messages; two user messages in a row can cause confusion.

2. Step 2: Identify the problem in the message list

   Here, two user messages come one after another without an assistant reply, so the AI may ignore the second user message.

3. Final Answer:

   Because there are two user messages without an assistant reply in between -> Option A

4. Quick Check:

   Messages alternate user-assistant-user [OK]

Hint: User and assistant messages must alternate to avoid ignored inputs [OK]

Common Mistakes:

• Thinking system role is missing
• Assuming assistant role used twice
• Believing empty messages cause ignoring

5. You want the AI to act as a friendly tutor and answer questions clearly. Which message list correctly sets this up before the user asks a question?

hard

A. [{"role": "system", "content": "You are a friendly tutor. Answer clearly."}, {"role": "user", "content": "Explain math."}]

B. [{"role": "user", "content": "Explain math."}, {"role": "system", "content": "Be friendly and clear."}]

C. [{"role": "assistant", "content": "I will be friendly."}, {"role": "user", "content": "Explain math."}]

D. [{"role": "user", "content": "Explain math."}, {"role": "assistant", "content": "I will be friendly."}]

Solution

1. Step 1: Place instructions in the system role

   Instructions for AI behavior must be in the system message before user input.

2. Step 2: Check message order and roles

   [{"role": "system", "content": "You are a friendly tutor. Answer clearly."}, {"role": "user", "content": "Explain math."}] correctly puts the system message first with instructions, then the user question.

3. Final Answer:

   [{"role": "system", "content": "You are a friendly tutor. Answer clearly."}, {"role": "user", "content": "Explain math."}] -> Option A

4. Quick Check:

   System message first with instructions [OK]

Hint: Put AI instructions in system role before user messages [OK]

Common Mistakes:

• Putting system message after user input
• Using assistant role for instructions
• Starting with user message without system context