CHECKLIST

Use this checklist to evaluate existing interfaces or plan new ones.

1. Memory & Context Awareness

Current State Assessment:

• Not done: System remembers user preferences (basic: theme, language)
• Not done: System tracks user behavior patterns
• Not done: System maintains context across sessions
• Not done: System recognizes user’s emotional state (frustration, satisfaction)
• Not done: System understands temporal patterns (time of day, day of week)
• Not done: System learns from user’s actual behavior (not just stated preferences)
• Not done: System maintains awareness of user’s ongoing goals
• Not done: System provides cross-device context continuity

Gap Analysis:

• Not done: What user context is currently lost between sessions?
• Not done: What behavioral patterns would be valuable to track?
• Not done: What temporal patterns affect user behavior?
• Not done: What emotional states impact user experience?

Implementation Priority:

• Not done: High: Essential context that users complain about losing
• Not done: Medium: Patterns that would improve experience noticeably
• Not done: Low: Nice-to-have personalization

2. Trust Evolution Architecture

Current State Assessment:

• Not done: System explains its reasoning for suggestions/decisions
• Not done: System shows confidence levels
• Not done: System allows users to adjust autonomy levels
• Not done: System has different trust stages (transparency → selective → autonomous)
• Not done: System provides undo/correction mechanisms
• Not done: System learns from user corrections
• Not done: System has trust recovery protocols for mistakes
• Not done: System escalates uncertain decisions appropriately

Gap Analysis:

• Not done: How transparent is system reasoning currently?
• Not done: Can users control autonomy levels?
• Not done: What happens when system makes mistakes?
• Not done: How does trust evolve over time (or does it)?

Trust Stage Design:

• Not done: Define what should always be transparent (high-stakes, uncertain)
• Not done: Define what can become autonomous (routine, high-confidence)
• Not done: Design transition criteria between stages
• Not done: Create user controls for trust progression

3. Relationship-Centric Metrics

Current Metrics Assessment:

• Not done: We measure: session duration, page views, conversion rates (traditional)
• Not done: We measure: relationship quality indicators
• Not done: We measure: compounding value over time
• Not done: We measure: context accuracy
• Not done: We measure: democratic alignment / ethical boundaries
• Not done: We track metrics longitudinally (weeks/months, not just sessions)
• Not done: We compare Month 1 vs. Month 6 experience quality

Gap Analysis:

• Not done: What traditional metrics are misleading for our use case?
• Not done: What relationship metrics would better indicate success?
• Not done: How do we measure improvement over time?
• Not done: What longitudinal tracking do we need?

New Metrics to Implement:

• Not done: Relationship Quality: Trust scores, delegation comfort
• Not done: Compounding Value: Time-to-success improvement, capability expansion
• Not done: Context Accuracy: Intent prediction, preference matching
• Not done: Democratic Alignment: Value alignment, boundary respect

4. Collaborative Planning Patterns

Current State Assessment:

• Not done: System understands user’s ongoing goals
• Not done: System provides proactive suggestions (not just reactive)
• Not done: System and user co-create plans together
• Not done: System adapts interface based on usage patterns
• Not done: System learns from user’s path choices
• Not done: System generates alternative paths dynamically
• Not done: System recognizes when user is stuck/frustrated
• Not done: System offers help at appropriate moments (not intrusive)

Gap Analysis:

• Not done: How well does system understand user goals?
• Not done: Is system proactive or only reactive?
• Not done: Do users and system collaborate on planning?
• Not done: Does interface adapt to individual users?

Collaborative Features to Add:

• Not done: Goal capture and tracking interface
• Not done: Proactive suggestion engine
• Not done: Co-creation workspace (human + AI contributions)
• Not done: Adaptive UI elements
• Not done: Learning feedback mechanisms

5. Privacy & Control

Current State Assessment:

• Not done: Users can see what system remembers about them
• Not done: Users can control what gets remembered
• Not done: Users can forget/delete specific memories
• Not done: Users can export their data
• Not done: System has clear retention policies
• Not done: System scrubs PII appropriately
• Not done: System respects user boundaries explicitly
• Not done: System explains data usage transparently

Gap Analysis:

• Not done: What memory controls do users currently have?
• Not done: Can users see and manage what’s remembered?
• Not done: Are privacy policies clear and actionable?
• Not done: Do users understand data retention?

Privacy Controls to Implement:

• Not done: Memory visualization interface
• Not done: Granular forgetting controls
• Not done: Data export functionality
• Not done: Clear retention policy UI
• Not done: Boundary setting interface

Memory & Data Contracts Sprint Guide

A structured 3-day sprint to design your memory architecture.

Day 1: Discovery & Mapping

Morning: Behavioral Inventory (3 hours)

1. List all user interactions (30 min)

   • What actions can users take?
   • What choices do users make?
   • What paths do users follow?

2. Identify valuable patterns (60 min)

   • Which patterns indicate user goals?
   • Which patterns indicate frustration?
   • Which patterns indicate satisfaction?
   • Which patterns evolve over time?

3. Map temporal dimensions (30 min)

   • What changes by time of day?
   • What changes by day of week?
   • What changes by season/context?

4. Define context signals (60 min)

   • What indicates user’s current state?
   • What indicates user’s goals?
   • What indicates user’s constraints?

Afternoon: Privacy & Retention Design (3 hours)

1. Categorize memory types (45 min)

   • Behavioral patterns
   • Explicit preferences
   • Ongoing goals
   • Historical trends
   • Sensitive information

2. Define retention policies (45 min)

   • What must be kept?
   • What should expire?
   • What should users control?
   • What requires special handling?

3. Design user controls (90 min)

   • Memory visualization interface
   • Forgetting controls
   • Export functionality
   • Boundary settings

Day 1 Deliverable: Memory inventory and privacy framework

Day 2: Architecture Design

Morning: Data Structures (3 hours)

1. Design event schema (60 min)

   Copy

   What does a behavioral event look like?
   - Timestamp
   - Event type
   - Context (user state, environment)
   - Outcome

2. Design pattern schema (60 min)

   Copy

   What does a detected pattern look like?
   - Pattern type
   - Confidence level
   - Supporting evidence
   - Temporal scope

3. Design memory graph (60 min)

   Copy

   How do memories connect?
   - User identity
   - Behavioral patterns
   - Ongoing goals
   - Trust level
   - Relationship timeline

Afternoon: Implementation Planning (3 hours)

1. Storage strategy (45 min)

   • Hot memory (recent, always loaded)
   • Warm memory (patterns, load on demand)
   • Cold memory (historical, analysis only)

2. Pattern detection algorithms (90 min)

   • Frustration indicators
   • Success patterns
   • Temporal patterns
   • Preference evolution

3. Privacy implementation (45 min)

   • PII scrubbing
   • Encryption
   • Access controls
   • Audit logging

Day 2 Deliverable: Complete memory architecture specification

Day 3: Metrics & Testing

Morning: Metrics Design (3 hours)

1. Select relationship metrics (45 min) Choose 2-3 from each category:

   • Relationship Quality
   • Compounding Value
   • Context Accuracy
   • Democratic Alignment

2. Define measurement methods (90 min) For each metric:

   • How to calculate?
   • What data needed?
   • What indicates success?
   • What indicates problems?

3. Design metric visualization (45 min)

   • Dashboard layout
   • Trend displays
   • Alert thresholds
   • User-facing vs. internal metrics

Afternoon: Testing Strategy (3 hours)

1. Longitudinal test plan (60 min)

   • Week 1 tests (onboarding, transparency)
   • Month 1 tests (pattern learning, trust evolution)
   • Month 3 tests (compounding value, autonomy)
   • Month 6 tests (relationship maturity)

2. Success criteria (60 min)

   • Week 1: Can system explain reasoning? Do users understand?
   • Month 1: Are patterns being detected? Is trust evolving?
   • Month 3: Is experience improving? Are metrics trending positive?
   • Month 6: Is compounding value evident? Are users delegating comfortably?

3. Risk mitigation (60 min)

   • Privacy risks and mitigations
   • Trust violation risks and recovery protocols
   • Metric degradation and intervention triggers

Day 3 Deliverable: Metrics specification and longitudinal test plan

Trust Evolution Design Worksheet

Use this worksheet to design trust evolution for your specific domain.

Step 1: Define Transparency Requirements

High-stakes decisions (ALWAYS transparent):

• Decision: _______________________________________________
• Why high-stakes: ________________________________________
• What to explain: _________________________________________

Uncertain decisions (ALWAYS show confidence):

• Decision: _______________________________________________
• Uncertainty source: ______________________________________
• Confidence threshold: ____________________________________

Routine decisions (CAN become autonomous):

• Decision: _______________________________________________
• Why routine: ____________________________________________
• Autonomy criteria: _______________________________________

Step 2: Map Trust Stages

Stage 1: Transparency Phase (Weeks 1-X)

User needs to:

• Not done: Understand system reasoning
• Not done: See confidence levels
• Not done: Learn system capabilities
• Not done: Build initial trust

System should:

• Not done: Explain all suggestions
• Not done: Show data sources
• Not done: Display alternatives considered
• Not done: Highlight uncertainties

Exit criteria:

• Not done: User accepts suggestions >60%
• Not done: User requests explanations <40%
• Not done: User indicates comfort with system

Stage 2: Selective Disclosure Phase (Weeks X-Y)

User needs to:

• Not done: See reasoning for important decisions
• Not done: Trust system for routine actions
• Not done: Understand when system is uncertain

System should:

• Not done: Full transparency for high-stakes decisions
• Not done: Confidence indicators for medium-stakes
• Not done: Quiet execution for routine tasks
• Not done: Clear escalation for uncertainties

Exit criteria:

• Not done: User accepts suggestions >75%
• Not done: User comfortable with routine autonomy
• Not done: User delegates specific categories

Stage 3: Autonomous Action Phase (Month Y+)

User needs to:

• Not done: Trust system to act independently
• Not done: Easy correction mechanisms
• Not done: Transparency on demand
• Not done: Control over autonomy levels

System should:

• Not done: Act autonomously for delegated categories
• Not done: Subtle notifications for actions taken
• Not done: Escalate uncertainties appropriately
• Not done: Learn from corrections

Maintain trust:

• Not done: Consistent with user values
• Not done: Clear undo mechanisms
• Not done: Explain on demand
• Not done: Adjust autonomy based on feedback

Step 3: Design Trust Indicators

Visual trust indicators:

• Confidence meter: How to display? ____________________________
• Reasoning toggle: Where to place? ____________________________
• Autonomy controls: How to adjust? ____________________________
• Trust score: Show to user? ____________________________

Trust evolution feedback:

• How does user know trust is evolving? ________________________
• How does user see autonomy increasing? ______________________
• How does user control progression? __________________________

Step 4: Trust Recovery Protocol

When system makes mistake:

1. Acknowledge (within X hours/days): ______________________ Template: “I made a suboptimal decision about [X] because [Y]”

2. Explain what went wrong: ________________________________

   • Wrong assumption: _______________________________________
   • What system learned: ____________________________________

3. Offer corrections:

   • Option A: _______________________________________________
   • Option B: _______________________________________________
   • User’s choice: __________________________________________

4. Adjust trust level:

   • Reduce autonomy in category: ____________________________
   • For duration: ___________________________________________
   • Re-evaluation criteria: _________________________________

5. Update model:

   • Pattern learned: ________________________________________
   • New guardrails: _________________________________________

Relationship Metrics Implementation Worksheet

Select Your Metrics (Choose 2-3 per category)

Relationship Quality Metrics:

Option 1: Trust Score

• Not done: Calculation method: _____________________________________
• Not done: Data sources: ___________________________________________
• Not done: Success threshold: ______________________________________

Option 2: Delegation Comfort

• Not done: Measurement: ____________________________________________
• Not done: Categories to track: ____________________________________
• Not done: Target: _________________________________________________

Option 3: Override Rate

• Not done: What counts as override: ________________________________
• Not done: Acceptable rate: ________________________________________
• Not done: Alert threshold: ________________________________________

Compounding Value Metrics:

Option 1: Time-to-Success Improvement

• Not done: Baseline measurement: ___________________________________
• Not done: Current measurement: ____________________________________
• Not done: Improvement rate: _______________________________________

Option 2: Capability Expansion

• Not done: Features used at baseline: _____________________________
• Not done: Features used currently: ________________________________
• Not done: Adoption rate: __________________________________________

Option 3: Outcome Quality

• Not done: Quality measurement: ____________________________________
• Not done: Baseline vs. current: ___________________________________
• Not done: Improvement trend: ______________________________________

Context Accuracy Metrics:

Option 1: Intent Prediction Accuracy

• Not done: How to measure intent: __________________________________
• Not done: Prediction vs. actual: __________________________________
• Not done: Target accuracy: ________________________________________

Option 2: Preference Matching

• Not done: Recommendation acceptance rate: _________________________
• Not done: Top choice hit rate: ____________________________________
• Not done: Target: _________________________________________________

Option 3: Timing Accuracy

• Not done: Suggestion timing evaluation: __________________________
• Not done: User feedback on timing: ________________________________
• Not done: Target: _________________________________________________

Democratic Alignment Metrics:

Option 1: Value Alignment Score

• Not done: Constitution/values defined: ___________________________
• Not done: Violation detection: ____________________________________
• Not done: Target: _________________________________________________

Option 2: Boundary Respect

• Not done: Boundaries defined: _____________________________________
• Not done: Violation tracking: _____________________________________
• Not done: Target: _________________________________________________

Option 3: Fairness Metric

• Not done: Segments to compare: ____________________________________
• Not done: Fairness calculation: ___________________________________
• Not done: Target: _________________________________________________

Longitudinal Tracking Plan

Week 1 Baseline:

• Metrics to capture: _________________________________________
• Measurement method: __________________________________________
• Baseline targets: ___________________________________________

Month 1 Check-in:

• Expected improvements: ______________________________________
• Red flags to watch: _________________________________________
• Intervention triggers: ______________________________________

Month 3 Assessment:

• Compounding value emerging: _________________________________
• Trust evolution complete: ___________________________________
• Feature adoption: ___________________________________________

Month 6 Maturity:

• Relationship quality: _______________________________________
• Compounding factor: _________________________________________
• Success indicators: _________________________________________

Quick Start: Minimum Viable Relationship (MVR)

Can’t do everything at once? Start here.

Week 1: Basic Memory

Implement:

• Not done: Store last 7 days of user interactions
• Not done: Detect 2-3 simple behavioral patterns (e.g., repeat searches, abandoned tasks)
• Not done: Show “Last time you…” context on relevant screens

Measure:

• Not done: Do users notice the context?
• Not done: Do users find it helpful?

Week 2-4: Trust Indicators

Implement:

• Not done: Add confidence indicators to suggestions
• Not done: “Explain this” button for system decisions
• Not done: Simple reasoning display

Measure:

• Not done: How often do users click “Explain this”?
• Not done: Does explanation increase acceptance?

Month 2: Longitudinal Metrics

Implement:

• Not done: Track 1 relationship quality metric
• Not done: Track 1 compounding value metric
• Not done: Compare Week 1 vs. Week 8

Measure:

• Not done: Is experience improving over time?
• Not done: Are metrics trending positive?

Month 3: Basic Autonomy

Implement:

• Not done: Identify 1-2 routine tasks
• Not done: Offer autonomous execution (with easy undo)
• Not done: Track delegation comfort

Measure:

• Not done: Do users accept autonomous actions?
• Not done: Is trust evolving naturally?

Month 6: Full Relationship Architecture

Implement:

• Not done: Complete memory architecture
• Not done: Full trust evolution (transparency → selective → autonomous)
• Not done: Comprehensive relationship metrics
• Not done: Collaborative planning features

Measure:

• Not done: Relationship quality scores
• Not done: Compounding value evident
• Not done: Context accuracy high
• Not done: User satisfaction

Red Flags: When Relationship Design Is Going Wrong

Memory Issues

Red Flag: Users complain “system doesn’t remember”

• Not done: Check: Are patterns being detected?
• Not done: Fix: Improve pattern detection algorithms

Red Flag: Users complain “system remembers too much”

• Not done: Check: Are privacy controls clear and accessible?
• Not done: Fix: Add memory visualization and forgetting controls

Red Flag: Context is wrong

• Not done: Check: Is pattern detection accuracy measured?
• Not done: Fix: Improve context signals and learning algorithms

Trust Issues

Red Flag: Users never progress beyond transparency phase

• Not done: Check: Is system reasoning clear?
• Not done: Fix: Improve explanation quality

Red Flag: Users don’t use autonomous features

• Not done: Check: Is delegation comfortable?
• Not done: Fix: Reduce autonomy scope, improve trust recovery

Red Flag: Trust violations

• Not done: Check: Do we have recovery protocols?
• Not done: Fix: Implement transparent recovery, adjust trust level

Metric Issues

Red Flag: Traditional metrics up, relationship metrics down

• Not done: Check: Are we optimizing for wrong things?
• Not done: Fix: Align incentives with relationship health

Red Flag: No compounding value

• Not done: Check: Is system learning from user behavior?
• Not done: Fix: Improve learning algorithms, pattern detection

Red Flag: Context accuracy declining

• Not done: Check: Is user behavior changing?
• Not done: Fix: Adapt model, update pattern detection

Summary: Relationship Design Readiness

Use this final checklist to assess readiness:

Foundation

• Not done: We understand the difference between screens and relationships
• Not done: We’ve identified user’s ongoing goals (not just immediate tasks)
• Not done: We’ve mapped behavioral patterns that matter
• Not done: We’ve designed for privacy and user control

Memory Architecture

• Not done: Event streaming or behavioral tracking implemented
• Not done: Pattern detection algorithms defined
• Not done: Privacy controls and PII handling designed
• Not done: Memory visualization planned

Trust Evolution

• Not done: Three trust stages designed for our domain
• Not done: Transparency requirements defined
• Not done: Autonomy criteria established
• Not done: Trust recovery protocols created

Relationship Metrics

• Not done: Selected 2-3 metrics per category
• Not done: Longitudinal tracking plan (Week 1, Month 1, 3, 6)
• Not done: Success thresholds defined
• Not done: Metric visualization designed

Collaborative Planning

• Not done: Goal capture interface designed
• Not done: Proactive suggestion logic defined
• Not done: Adaptive UI patterns planned
• Not done: Learning feedback mechanisms created

Ready to build?

• Not done: Team understands relationship-centric paradigm
• Not done: Product roadmap includes longitudinal success criteria
• Not done: Testing plan includes relationship development phases
• Not done: Privacy and ethics frameworks established

If all checked: You’re ready to build agentic, relationship-centric experiences!

See EXAMPLES.md for domain-specific implementations. See REFERENCE.md for technical implementation details.