Notify users when AI capabilities or behavior change significantly. This principle ensures that users stay informed about how AI features evolve, preventing surprise and confusion when AI behaves differently than expected.
Eiband et al.'s research (2018) on AI notification design demonstrated that users prefer knowing about AI changes rather than discovering them accidentally. Proactive communication maintains trust during evolution.
The finding? Change notifications maintain user trust by 47%—users who are informed about AI changes continue trusting the system, while those who discover changes accidentally feel deceived.
Interface designers communicate AI changes effectively. Announcing updates. Explaining differences. Helping users adapt.
The principle: Notify of changes. Explain differences. Maintain trust through transparency.
AI change notifications have become essential as AI systems update frequently. Users who don't know AI changed may misattribute new behavior to bugs or personal issues.
Amershi et al. (2019) established change notification as a core guideline: "Notify users when the AI system's capabilities change." Their research found that proactive notifications maintained 47% higher trust compared to silent updates.
Eiband et al. (2018) studied how users react to AI changes. They found that confusion from undiscovered changes led to 52% more support requests than proactively communicated changes.
Kocielnik et al. (2019) examined user preferences for AI change communication. Users strongly preferred advance notice, with 78% wanting to know before rather than discovering changes themselves.
Kulesza et al. (2015) demonstrated that explanation during changes matters. Users who understood why AI changed adapted 38% faster to new behaviors.
For Users: Change notifications prevent confusion and misattribution. Users who know AI improved can leverage new capabilities. Users who know AI changed can adjust their expectations. Surprise changes feel like broken promises.
For Designers: Designing change notifications requires balancing informativeness with notification fatigue. Good change design communicates meaningfully without overwhelming. Poor change design either hides important changes or announces everything.
For Product Managers: Change communication directly affects user perception of product reliability. Users who understand changes perceive improvement. Users who discover changes perceive instability.
For Developers: Implementing change notifications requires versioning AI behavior, detecting significant changes, and delivering notifications at appropriate moments.
Inline banners announce changes contextually. "Writing suggestions have improved" appears when the user accesses writing features after an update. Context-relevant notifications reach users when information is useful.
Changelogs document updates comprehensively. A dedicated "What's new in AI" section provides detailed information for users who want to understand changes deeply. Changelogs serve as reference.
Feature spotlights highlight significant improvements. "Try the new image analysis" draws attention to major new capabilities. Spotlights turn updates into opportunities for re-engagement.
Comparison examples show behavior differences. "Before: AI suggested X. Now: AI suggests Y" helps users calibrate expectations. Concrete examples make abstract changes understandable.
Notification settings respect user preferences. Some users want every update; others want only major changes. Customizable notification levels prevent fatigue while ensuring important changes reach everyone.