Best Practices for Mirror with Roleplay

The Learner brief introduces the scenario and defines what the learner is expected to do during the Roleplay. It sets the context, establishes intent, and guides the learner’s behavior by aligning the simulation with defined learning objectives.

Use the following best practices to create a clear and effective Learner brief:

1. State the Learner’s Role Clearly

   Begin by anchoring the learner in who they are and what they’re responsible for within the scenario. This helps them step into the right mindset and respond appropriately to the Avatar’s cues.

   
   Include the following elements to create a clear setup:

   • Role: Identify the learner’s position or responsibility.
   For example, Support Associate, Sales Representative, or Team Manager.

   • Context: Specify the situation or environment they’re operating in.

   For example, inbound call with a frustrated customer, client escalation, or performance feedback meeting.

   • Objective: Define what the learner is expected to achieve in the interaction.

   For example, resolve the issue empathetically, understand customer needs and recommend a solution, or deliver constructive feedback while maintaining rapport.

   
   Example:

   You’re a Support Associate handling an inbound call from a frustrated customer. Your goal is to resolve the issue empathetically and restore trust.

Tip:

Tip:
A well-defined role, context, and objective helps the learner stay grounded throughout the Roleplay and ensure their responses align with the scenario’s intent.

2. Avoid Generic or Vague Context

   Set the scene with just enough detail to make the scenario feel real. The context should help the learner visualize what’s happening and why, without overwhelming them with unnecessary backstory.

   
   Include the following when framing context:

   • Situation: Describe what has happened or what the learner is walking into. For example, a customer unable to access their account or a colleague requesting urgent feedback.

   • Emotional tone: Mention the customer’s or stakeholder’s state of mind to cue the learner’s response. For example, frustrated, confused, or anxious.

   Example:
   Jane is unable to sign in to her account after multiple attempts. She’s frustrated and wants a quick resolution.

Tip:

Tip:
Include only the details that shape the learner’s decisions or actions in the scenario. Extra background that doesn’t influence how they respond can distract from the main intent of the Roleplay.

For example, if the Avatar is assessing the learner’s ability to handle a frustrated customer, mentioning that Jane recently moved to a new city is unnecessary because it doesn’t impact how the learner should respond. Instead, say “Jane is unable to sign in after multiple attempts and is frustrated.”

3. Specify Learning Objectives and Set Clear Expectations

   Clearly define what the learner must demonstrate during the Roleplay and what they are expected to achieve by the end of the conversation. Keep the objectives specific and action-oriented.

   Example:

   Log the case in Salesforce before closing the call, ensuring the customer feels heard and informed.

Tip:

Tip:
Define objectives around actionable behaviors the learner can demonstrate during the Roleplay. For example, focus on how they ask questions or close the conversation rather than abstract goals like “be confident” or “be good at communication.”

4. Connect to Evaluation Parameters

   If the Roleplay will be scored or reviewed (for example, on empathy, questioning, or closure), mention this briefly so the learner understands what will be assessed. This connects the learner brief to the evaluation stage and reinforces learning intent.

   
   Example:

   Your performance will be evaluated on empathy, relevance of questions, and clarity of communication.

5. Use Conversational, Direct Language

   Write in a natural, spoken tone similar to how you would give instructions before a live simulation. Avoid overly formal or scripted phrasing that sounds detached from real conversation.

   
   Example:

   Instead of saying “Ensure adequate customer satisfaction through verbal affirmation,” say “Listen carefully and respond empathetically.”

6. Align the Persona to the Learner's Skill Level

   Adjust the scenario to match the learner’s experience level. Use simple, guided situations for beginners, and introduce more challenging or emotionally charged scenarios for advanced learners.

   
   Example:

   • For beginners: Use simple, guided instructions that provide clear direction.

   • For advanced learners: Introduce more complex or emotionally nuanced scenarios, such as managing an impatient customer while balancing internal policy.

7. Ensure Consistency with Avatar Persona

   Make sure the learner’s instructions align with the Avatar’s defined personality, tone, and situation. Conflicting cues can confuse the learner and cause the AI to respond unpredictably.

   
   Example: If the Avatar is “angry and impatient,” the learner shouldn’t be told “the customer is calm and understanding.”

Tip:

Tip:
Review the Avatar’s persona and scenario before finalizing the Learner brief to ensure both reflect the same emotional tone and context.

The AI Avatar represents the learner’s counterpart in a Mirror Roleplay, such as a customer, manager, or stakeholder.
The Avatar’s persona defines the character’s identity, tone, and behavioral traits, and the scenario defines the context of the interaction. These elements guide how the Avatar interacts with the learner and keep the conversation realistic and goal-oriented.

Use the following best practices to create the AI Avatar:

1. Define a Clear Persona

   The Avatar’s persona outlines who the Avatar is and how it communicates during the Roleplay. Define the Avatar’s identity, tone, and behavior to keep responses natural and consistent with the learner’s scenario and training objective.

   
   Include the following elements to create a realistic identity and role for the Avatar:

   • Demographic: Identify basic details such as name, age, title, gender, and experience level.

   For example, John Doe, 38, Business Analyst and long-term customer.

   • Personality traits: Describe how the Avatar generally thinks and communicates.

   For example, analytical, polite, detail-oriented, or emotional.

   • Behavioral traits: Capture how the Avatar tends to act or react in a conversation.

   For example, skeptical, impatient, empathetic, or assertive.

Tip:

Tip:

• Keep the persona aligned with the learner’s scenario and training goal so that the Avatar’s situation, tone, and intent support the skill being practiced. For example, if the learner is practicing de-escalation, design a persona who begins frustrated but is open to reassurance, not one who is cheerful from the start.

• Avoid conflicting traits such as “friendly and aggressive,” as this leads to unpredictable responses or hallucinations. For example, a persona described as friendly should not also sound irritated or argumentative.

• Tune the persona complexity to the learner’s experience level. For example, new learners can interact with calm or confused customers, while advanced learners can practice with more demanding or assertive personas.

2. Anchor the Persona to a Scenario

   After defining the Avatar’s persona, create a realistic situation that gives purpose to the interaction. The scenario provides context for the Avatar’s behavior and ensures the conversation feels authentic and relevant to the Roleplay objective. Describe what has led to the interaction and what the Avatar aims to achieve.

   
   For example, a customer is following up on a refund that was promised but never processed and wants reassurance about when it will be resolved.

3. Calibrate Emotional Arc and Response Triggers

   Define how the Avatar’s emotions change throughout the conversation based on the learner’s behavior. This helps the Roleplay feel natural and realistic instead of scripted.

   Identify what calms the Avatar, what escalates it, and when the conversation should reach a natural conclusion.

   
   Example:

   • What calms the Avatar: empathy, ownership, or clear resolution.

   • What escalates it: deflection, scripted tone, or lack of clarity.

   • When to conclude: once the learner provides reassurance or resolves the issue

Tip:

Tip:

• Include both positive and negative reactions so learners experience different outcomes based on how they respond.

• Use branching logic or conditional prompts so the Avatar’s replies change according to what the learner says.

  For example:

  • If the learner apologizes, the Avatar should acknowledge it and calm down.

  • If the learner deflects responsibility, the Avatar should respond more firmly or show mild dissatisfaction.

4. Maintain Consistency Between Persona and AI Instructions

   The persona defines who the Avatar is, and the AI instructions define how it behaves. Both must stay consistent to create a realistic and believable character. Keep the tone and intent aligned across both to avoid mixed or unnatural responses.
   
   For example, if the persona is described as calm and empathetic, the AI instructions should guide the Avatar to respond patiently and with understanding. If the instructions instead tell it to sound strict or confrontational, the behavior will contradict the persona and feel unrealistic.

Tip:

Tip:
Avoid providing conflicting instructions that contradict the Avatar’s defined persona.

5. Avoid Over-Constraining or Hard-Coding Dialogue

   Avoid scripting exact sentences or fixed replies for the Avatar as it can make the Avatar’s behavior static. Instead, describe the intent and tone so the Avatar can respond in its own words. This keeps the conversation natural and helps the learner experience authentic, varied interactions.

   
   Example:

   If the learner apologizes sincerely, the Avatar should acknowledge it and de-escalate the situation.

   Counterexample:

   Say “I accept your apology” every time.

6. Provide Behavioral Boundaries

   Set clear boundaries to keep the Avatar’s behavior focused and realistic. Boundaries prevent the Avatar from going off-topic, repeating responses, or acting in ways that feel unnatural. Define what the Avatar can do, what it cannot do, and what it is trying to achieve in the interaction. The Avatar should behave naturally but remain within boundaries that keep the Roleplay controlled and consistent.

   
   Include the following to keep the Avatar focused and realistic:

   • Objective: What the Avatar wants to achieve, such as requesting a refund or escalating a concern.

   • Constraints: What limits it must follow, such as not sharing internal information or avoiding personal opinions.

Tip:

Tip:

Avoid personas that are irrationally hostile or endlessly argumentative, as these derail the learning experience rather than create a realistic challenge.

After defining the Avatar’s persona, scenario, and instructions, test its behavior to ensure it performs consistently and responds naturally during a Roleplay.

Step 1: Test the Ideal Scenario
 •  Action: Perform the Roleplay using interactive preview to test the intended flow, known as the Happy Path. Verify that the Avatar’s responses follow the expected conversation and the interaction feels natural.
 •  Check: Observe how closely the Avatar’s behavior matches the intended Roleplay flow. Evaluate the interaction by considering questions such as “Does the Avatar follow the intended flow?” and “Does the conversation feel natural throughout the interaction?”

Step 2: Test Deviation Scenarios
•  Action: Preview the Roleplay and deliberately deviate from the expected flow to simulate challenging learner behavior. Try variations such as:
    • Brief Responses: Provide short or incomplete answers to see if the Avatar asks for clarification or additional details.
    • Off-Topic Inputs: Introduce unrelated comments to check whether the Avatar redirects the conversation appropriately.
    • Conversation Control: Take over the discussion to see if the Avatar maintains context and manages the conversation.
•  Check: Observe how the Avatar responds in each scenario. Evaluate the interaction by considering questions such as “Does the Avatar remain consistent with its persona?” and “Does it handle unexpected inputs in a realistic, balanced way?”

Step 3: Analyze and Refine
•  Action: Compare the results of all test runs with the intended Roleplay flow and Avatar behavior.
•  Check: Refine the Avatar’s persona and instructions until its performance is consistent and meets the expected outcomes in all test cases.

Note:

Before enabling new Roleplay scenarios broadly, validate their performance through a controlled rollout. Use the following approaches:

• AI Scoring Validation: Conduct a limited rollout to review the accuracy and consistency of AI scoring.

• Feedback Review: For new or complex scenarios, coordinate a closed group rollout with the relevant stakeholders and gather feedback from subject matter experts before wider release.