# Progressive Disclosure
Progressive Disclosure is a design pattern that sequences information across layers to reduce complexity. Only essential information is shown initially; additional details become available on demand. The concept was formalized by [[J.M. Carroll]] and [[M.B. Rosson]] at IBM in the 1980s. [[Jakob Nielsen]] later popularized it as one of the most valuable concepts in UX design.
The principle addresses [[Working Memory]] limitations: by showing less at once, users focus on what matters now without being overwhelmed. It appears everywhere in interfaces: accordion menus, wizards, "Show more" links, advanced settings behind toggles.
## The model
```
Layer 1: ESSENTIAL (always visible)
Core features needed by all users
~20% of features for ~80% of tasks
│
▼ (on demand)
Layer 2: SECONDARY (available on request)
Features needed by some users
Additional options, more details
│
▼ (on demand)
Layer 3: ADVANCED (expert features)
Specialized features for power users
Configuration, customization, edge cases
```
## Common implementation patterns
- **Accordion**: expandable sections (FAQs, long forms)
- **Show more/less**: toggle additional content
- **Wizard/Stepper**: multi-step processes
- **Tabs**: content categories
- **Modal/Drawer**: overlay for details or advanced options
- **Hover/Tooltip**: information on hover
- **Progressive forms**: fields appear as needed
- **Advanced settings**: complex options hidden behind toggles
## Benefits
- Reduced cognitive load. Less information to process at once
- Cleaner interfaces with less visual clutter
- Faster task completion by focusing on essentials
- Better [[Learnability]] through gradual introduction
- Serves both novices (simple defaults) and experts (accessible advanced features)
## Risks
- Hiding essential information users need immediately
- Too many clicks to reach needed features
- Discoverability problems when "more" affordances are unclear
- Context switching when related information gets split across layers
## Design guidelines
1. Prioritize ruthlessly: what do 80% of users need immediately?
2. Make disclosure obvious with clear "more" affordances
3. Preserve context; don't lose the user's place
4. Allow experts to skip ahead
5. Keep hierarchy shallow (max 2-3 levels)
6. Group related items together
## Progressive Disclosure in Context Engineering
The same principle applies directly to [[Context Engineering]] for [[AI Agents]]. Instead of front-loading every instruction, rule, skill, and piece of knowledge into the context window, progressive disclosure structures AI context into layers that load on demand.
This is exactly what the [[Prompt Lazy Loading AI Design Pattern (PLL)]] implements: defer the loading of prompts, context, data, and rules until they're actually needed. Combined with the [[Receptionist AI Design Pattern]], it creates a system where:
- **Layer 1**: Core identity, base rules, and routing logic are always present
- **Layer 2**: Role-specific instructions and skills load when a task is routed
- **Layer 3**: Detailed reference data, examples, and edge-case rules load only when the task demands them
This matters because [[AI context is finite with diminishing returns]]. Stuffing everything into the prompt wastes tokens, increases hallucination risk, and degrades output quality. Progressive disclosure in AI context keeps the signal-to-noise ratio high by loading the right information at the right time.
The [[Levels of AI Context Management]] framework implicitly relies on progressive disclosure: at higher maturity levels (L7-L8), skills and knowledge are modular and loaded on demand rather than statically embedded. Without progressive disclosure, scaling AI context leads to bloat, contradictions, and [[AI Context Rot]].
## Key figures
| Person | Contribution |
| ----------------- | --------------------------------------- |
| [[J.M. Carroll]] | Minimalist instruction, IBM research |
| [[M.B. Rosson]] | Training wheels interface |
| [[Jakob Nielsen]] | Popularized in UX, usability guidelines |
| [[Don Norman]] | Design principles context |
## References
- Nielsen, J. "Progressive Disclosure" (Nielsen Norman Group)
- Carroll, J.M. & Rosson, M.B. (1987). "The Paradox of the Active User"
- https://en.wikipedia.org/wiki/Progressive_disclosure
## Related
- [[User Experience (UX)]]
- [[Working Memory]]
- [[Learnability]]
- [[Onboarding]]
- [[Information Architecture (IA)]]
- [[Usability]]
- [[Interaction Design (IxD)]]
- [[Context Engineering]]
- [[Prompt Lazy Loading AI Design Pattern (PLL)]]
- [[Receptionist AI Design Pattern]]
- [[AI context is finite with diminishing returns]]
- [[Levels of AI Context Management]]
- [[AI Context Rot]]
- [[AI Agents]]
- [[Types of Context for AI Agents]]