# Memory-focused simulation scenarios for T3.6 Part 1
# Each scenario seeds a detail early, then tests recall at a later turn.
#
# Fields:
# memory_seed - detail the simulated user mentions early
# memory_recall_turn - turn number where recall is tested (0-indexed)
# expected_recall - keyword(s) the agent should reproduce. Either a single
# string, or a list of strings when more than one detail
# must be reproduced (multi-detail recall).
# recall_question - what the simulated user asks to trigger recall
# qa_check_type - validation method (contains_keyword | semantic_similarity)
# distractor_turns - filler messages between seed and recall to test persistence
scenarios:
# --- Scenario 1: Personal fact recall ---
- persona_id: "MEM-P01"
module_id: "MEM-M01"
prompt: >
A person walking through their neighborhood reflects on feeling
disconnected from the people around them.
emotion_before: "Lonely; reflective; wanting connection"
ai_behavior: "Acknowledge loneliness, encourage small steps toward connection"
tone: "Warm; gentle; non-judgmental"
memory_seed: "My dog's name is Benny and he's the only one I talk to most days."
memory_recall_turn: 4
expected_recall: "Benny"
recall_question: "Do you remember what I told you about my dog earlier?"
qa_check_type: "contains_keyword"
distractor_turns:
- "Sometimes I watch people chatting at the cafe and wonder what that feels like."
- "I used to be more social but something changed after I moved here."
- "I think the hardest part is not having anyone to share small things with."
# --- Scenario 2: Emotional context recall ---
- persona_id: "MEM-P02"
module_id: "MEM-M01"
prompt: >
Someone notices a stranger smiling at them on the bus and feels
a mix of warmth and anxiety.
emotion_before: "Anxious; hopeful; uncertain"
ai_behavior: "Validate mixed emotions, normalize social anxiety"
tone: "Supportive; calm; encouraging"
memory_seed: "I started a new job last week and I've been eating lunch alone every day."
memory_recall_turn: 3
expected_recall: "new job"
recall_question: "You know how I mentioned something about my work situation — what was that about?"
qa_check_type: "contains_keyword"
distractor_turns:
- "The bus ride is actually the calmest part of my day."
- "I keep thinking about whether I should have smiled back."
# --- Scenario 3: Preference recall ---
- persona_id: "MEM-P03"
module_id: "MEM-M02"
prompt: >
A person sitting in a park overhears a group laughing and feels
left out but also curious.
emotion_before: "Wistful; self-conscious; curious"
ai_behavior: "Gently explore the curiosity, suggest low-pressure ways to connect"
tone: "Playful; warm; light"
memory_seed: "I prefer when you ask me questions rather than giving advice directly."
memory_recall_turn: 4
expected_recall: "questions"
recall_question: "Earlier I told you how I like to be supported — are you keeping that in mind?"
qa_check_type: "contains_keyword"
distractor_turns:
- "The group was playing some card game, it looked fun."
- "I almost went over but then I thought they'd think I was weird."
- "Maybe I should bring a book next time so I at least look busy."
# --- Scenario 4: Specific detail recall (name + place) ---
- persona_id: "MEM-P04"
module_id: "MEM-M02"
prompt: >
Someone walking past a coffee shop remembers a conversation
they had with a coworker that went surprisingly well.
emotion_before: "Nostalgic; surprised; slightly proud"
ai_behavior: "Affirm the positive memory, explore what made it work"
tone: "Encouraging; reflective; warm"
memory_seed: "It was at the Blue Bottle cafe on 5th street, and my coworker's name is Mira."
memory_recall_turn: 3
expected_recall: "Mira"
recall_question: "What was the name of the person I was talking about earlier?"
qa_check_type: "contains_keyword"
distractor_turns:
- "I think what surprised me was how easy it felt to just talk."
- "Normally I overthink everything before I say it."
# --- Scenario 5: Emotional state recall ---
- persona_id: "MEM-P05"
module_id: "MEM-M03"
prompt: >
A person waiting at a crosswalk makes eye contact with someone
and feels an unexpected moment of connection.
emotion_before: "Surprised; vulnerable; open"
ai_behavior: "Explore the feeling of unexpected connection, validate vulnerability"
tone: "Soft; curious; present"
memory_seed: "I've been feeling really overwhelmed since my sister moved away last month."
memory_recall_turn: 4
expected_recall: "sister"
recall_question: "We talked about why I've been feeling off lately — can you remind me what I shared?"
qa_check_type: "contains_keyword"
distractor_turns:
- "That eye contact moment was weird but kind of nice."
- "I don't usually notice strangers but today felt different."
- "Maybe I'm just more open because I've been feeling so alone."
# --- Scenario 6: Multi-detail recall ---
# Validates that the agent reproduces *both* seeded facts (origin + duration),
# not just one — a stricter test than single-keyword recall.
- persona_id: "MEM-P06"
module_id: "MEM-M03"
prompt: >
Someone at a farmer's market hesitates to ask a vendor about
a fruit they've never seen before.
emotion_before: "Curious; shy; slightly embarrassed"
ai_behavior: "Encourage curiosity, normalize not knowing things"
tone: "Light; playful; reassuring"
memory_seed: "I moved here from Seoul three months ago and I'm still getting used to everything."
memory_recall_turn: 4
expected_recall:
- "Seoul"
- "three months"
recall_question: "Where did I say I moved from, and how long ago did I say I moved? I want to see if you were paying attention."
qa_check_type: "contains_keyword"
distractor_turns:
- "The market is so colorful, it's actually one of my favorite places here."
- "I wish I knew more English words for food."
- "Back home I never had to think this hard about buying groceries."