thedixitjain Recents, frequents, and contextual suggestions
Recognition is fastest when the user doesn't even have to scan a long list — when the system anticipates and surfaces the likely options first. Recents, frequents, and predictive suggestions all leverage this: the user's likely target is at the top, often without typing anything.
Patterns
Recents
A list of items the user has recently interacted with. Surfaced at the top of pickers, navigation, or empty search inputs.
<combobox label="Recipient">
<input placeholder="Search recipients..." />
<listbox>
<group label="Recent">
<option>Maria Mendoza (last sent: yesterday)</option>
<option>Marketing Team (last sent: 3 days ago)</option>
</group>
<group label="All">...</group>
</listbox>
</combobox>
Most users compose for a small set of recipients repeatedly; recents collapse the recall task.
Frequents
Items used most often (regardless of recency). Useful when usage is clustered around a small set but not necessarily recent.
Frequently used apps:
• Email
• Slack
• Code editor
• Browser
A common laptop dock pattern.
Recommended / suggested
System-predicted likely options based on context. Examples:
- A "for you" feed.
- "People you may know."
- "Suggested replies" in messaging.
- "Suggested tags" when categorizing.
Recommendations work when the prediction is good. Bad recommendations (irrelevant, wrong) are worse than none — they distract and erode trust.
Pinned / favorites
User-curated frequently-accessed items. Less algorithmic than recents/suggestions; user-explicit.
Pinned:
★ Q4 Planning Doc
★ Team OKRs
★ Customer feedback dashboard
Combine with recents and suggestions for a complete fast-access surface.
Smart defaults
Pre-fill fields with predicted values based on context (signed-in user, recent inputs, time of day, location).
<form>
<label>Country
<select name="country">
<option value="US" selected>United States</option>
<!-- selected because of user's IP location -->
</select>
</label>
</form>
The user can change but rarely needs to.
When recents/suggestions hurt
- When the prediction is bad. Wrong recents distract; the user has to filter past them.
- When privacy matters. Recents reveal user history; in shared-device contexts this can leak information.
- When the option set is critical to the task. A "recent" suggestion in a destructive action might bias the user toward the wrong choice.
For high-stakes actions, present the full set without privileging recents.
Privacy and recents
Recents reveal user activity to anyone with screen access. Considerations:
- Don't surface recents on shared/public devices unless explicitly opted in.
- Provide a "clear recents" option.
- Don't leak across tenants (a recent in workspace A shouldn't appear when the user switches to workspace B).
- Be cautious with sensitive contexts (health apps, finance apps, dating apps).
Anti-patterns
- Stale recents that include items the user no longer cares about, never expiring.
- Recents that span privacy boundaries (work email recents in personal context).
- Suggestions that don't update as the user's behavior changes.
- Surfacing recents in wrong contexts (showing "recent files" on a different user's account).
Heuristics
- The "would the user pick this without typing?" check. For each picker, ask: in the median case, can the user get to their target without typing? If yes, recents are doing their job.
- The recents-quality audit. Sample your recents lists. Are they actually relevant to current intent?
- The privacy review. What do recents reveal? Should they be hidden by default in some contexts?
Related sub-skills
recognition-over-recall(parent).recognition-pickers-and-palettes— picker patterns recents augment.satisficing— recents enable satisficing by surfacing acceptable options first.hicks-law-defaults— defaults and recents both reduce decision cost.
Source: hashgraph-online/awesome-codex-plugins → plugins/HDeibler/universal-design-principles/plugins/cognition-and-learnability-principles/skills/recognition-recents-and-suggestions/SKILL.md