Design behavior users can reliably anticipate.
Ensure system actions align with established expectations.
Contents 9. AI-Mediated Interaction 3. Predictability
Why stability matters more than surprise.
First, ask yourself
“Can users of your system reliably predict its behavior?”
Mission statement
Ensure users anticipate predictable outcomes by stabilizing system behavior and minimizing unexpected results.
Key heuristics Predictability
- System actions must align with established user expectations.
- Variation across comparable scenarios must be explained or expected.
- Stable behavioral patterns must be exposed over internal complexity.
- System variability must be constrained and clearly communicated.
- Users must be told when identical inputs can produce different outputs.
- Output variability must stay within communicated bounds.
- Meaningful changes to system logic must not occur silently.
- Behavioral transparency must increase as automation increases.
- Scaling and updates must preserve established behavioral expectations.
Executive brief
AI systems must behave predictably to earn user trust.
Core questions Predictability
“Is the system’s behavior predictable enough for users to rely on it?”
“Does it behave the way users expect?”
Focus areas
Expectation-first
“Can users form reliable expectations about behavior?”
“Do similar inputs produce similar outcomes?”
Boundary-aware
“Where does predictability end? Is that clear?”
“Are edge cases predictable or at least signaled?”
Consistency-aware
“Are patterns uniform across screens, states, and contexts?”
“Does the system reinforce learned expectations over time?”
AI-aware
“Is variability explainable and bounded?”
“Does AI learning improve reliability without surprising users?”
UX directives Predictability
Explain variation across comparable scenarios.
When similar inputs yield different outputs, show why or show that variation is expected.
Expose stable behavioral patterns, not internal complexity.
Prioritize outcome clarity over algorithmic explanation.
Constrain and signal system variability.
Communicate when behavior may differ from prior patterns.
Make output variability explicit.
Tell users when the same input can yield different results.
Keep variability within stated bounds.
Do not let outputs vary beyond the range users were led to expect.
Prevent unannounced behavioral drift.
Communicate meaningful changes to system logic.
Strengthen predictability in automated workflows.
Increase behavioral transparency as autonomy rises.
Preserve expectation stability during scaling and updates.
Ensure expansion does not invalidate prior learning.
Executive summary
- Predictability is reliable expectation alignment, not rigid uniformity.
- It ensures users can anticipate outcomes based on prior interaction.
- The system must produce consistent results in comparable contexts and signal meaningful variability.
- Probabilistic variability must be explicit and kept within stated bounds.
- Behavioral drift and logic changes must be communicated before they invalidate prior learning.
- Predictability succeeds when users can act confidently because system behavior remains stable, transparent, and foreseeable.
Success indicators
- System behavior matches user expectations.
- Similar actions produce consistent results in similar situations.
- Users can anticipate the outcome of their actions.
- Output variability is clearly communicated.
- Behavioral changes are announced and explained.
One-line summary
Before users can rely on intelligent behavior, they must be able to anticipate it.