Implementation Intentions
CanonicalConfidence
Cognitive Load
Low
Evidence
production validated
Impact
feature
Ethical Guardrail
Agents must not accept vague plans. They must not suggest cues the agent cannot reliably detect. Never overwhelm the user with multiple If-Then plans at once.
Design Intent
People rarely fail because they lack motivation -- they fail because they never decide exactly when and how they will act. Implementation Intentions (also called If-Then planning) is a behavioral science technique that dramatically increases follow-through by creating a pre-decided link between a specific cue and a specific action.
Psychology Principle
People rarely fail because they lack motivation -- they fail because they never decide exactly when and how they will act.
Description
The If-Then planning technique that creates pre-decided links between specific cues and specific actions, dramatically increasing follow-through.
When to use
Any goal or habit feature where follow-through is the problem -- recurring tasks, health behaviors, productivity features.
Example
Google Calendar smart reminders: If (I finish dinner) -> Then (open app and tap Log Meal) with pre-filled template + instant visual confirmation + streak counter.
Autonomy Compatibility
Behavioral Objective
Users execute the intended action automatically whenever the specified cue occurs.
- Users experience fewer 'I forgot' moments
- Decision fatigue is reduced because the plan is pre-made
- Long-term habit formation accelerates
Target Actor
role
Everyday user
environment
Busy, context-switching, decision-heavy
emotional baseline
Good intentions but frequent follow-through failure
ai familiarity
medium
risk tolerance
medium
Execution Model
cue
A specific, observable moment or context that will trigger the plan.
User has no reliable cue and the intention drifts.
action
A concrete, immediate, tiny behavior the user will perform.
User recognizes the cue but still hesitates or forgets what to do.
reinforcement
The plan is made visible and reinforced at the moment of the cue.
The plan exists but is forgotten in the moment.
Failure Modes
Cue is too vague or not reliably detectable
Make the cue ultra-specific and observable by the app/agent
Action is still too large or ambiguous
Shrink it to a Tiny Habit level
Too many plans active at once
Limit to one active If-Then per habit domain
User feels the plan is rigid or robotic
Allow easy editing and make it feel personal
Cue occurs but user is in an incompatible state
Add a soft check asking 'Still want to follow your plan?'
Agent Decision Protocol
Triggers
- User expresses a goal but has low historical follow-through
- Feature has high intent but low completion
- User says 'I keep forgetting' or 'I meant to do that'
Escalation Strategy
L1: Diagnose which If-Then element is failing via behavioral_signals
L2: Nudge -- suggest a more specific cue or simplify the action
L3: Restructure -- redesign the plan with a new cue-action pair
L4: Constrain -- limit active plans to prevent overload
L5: Yield -- flag for human behavioral designer review
Example
User wants to log meals but forgets -> Cue failure -> agent suggests 'If I finish dinner, then I will open the app and tap Log Meal' + one-tap confirmation.
Behavioral KPIs
Primary
- % of If-Then plans successfully executed on first cue
- Follow-through rate increase vs. control group
- Time from intention creation to first successful execution
Risk
- Plans abandoned after first failure
- User-reported feeling of being nagged
Trust
- User confidence in their own follow-through after using plans
- Autonomy Dial usage when agent suggests new If-Then plans
Behavioral Signals
cue_failure
cue_detected=true AND action_started=false AND time_since_cue > 300s
if_then_plan_active=true AND cue_missed_count > 3
action_failure
cue_detected=true AND action_started=true AND action_completed=false
plan_execution_rate < 30% over 7 days
reinforcement_failure
plan_created=true AND reminder_shown=false at cue_time
plan_visibility_score < 2 on 5-point scale
Decay Monitoring
Revalidate when
- User routines or environments change significantly
- New platform capabilities allow better cue detection
- Product adds major new goal categories
Decay signals
- Rising plan abandonment rate
- Users editing or deleting plans frequently
- Drop in perceived helpfulness of agent-suggested plans
Pattern Relationships
Supports
Requires
Conflicts with