AI Trust & Transparency
Design interfaces where users can see into the AI's reasoning, calibrate their trust appropriately, and verify claims independently. The GLASS framework makes AI decision-making visible without overwhelming users.
Core Principle
Trust is not a boolean. Users should not "trust AI" or "distrust AI" - they should develop calibrated trust: high confidence when the AI is reliable, healthy skepticism when it's uncertain. Your job is to give them the signals to calibrate correctly.
The GLASS Framework
| Letter |
Principle |
Design Question |
| G |
Ground in Sources |
Can the user trace every AI claim back to a verifiable source? |
| L |
Layer Explanations |
Can the user get a 5-second answer AND a 5-minute deep dive? |
| A |
Advertise Limitations |
Does the interface proactively tell users what the AI is NOT good at? |
| S |
Show Confidence |
Can the user see how certain the AI is about each output? |
| S |
Support Override |
Can the user correct, override, or reject AI outputs without friction? |
The Trust Calibration Spectrum
Design for the right trust level - not maximum trust.
| Trust Level |
User Behavior |
Design Goal |
When Appropriate |
| Over-trust (Automation Bias) |
Accepts all AI outputs without checking |
Introduce friction to encourage verification |
High-stakes decisions (medical, financial, legal) |
| Calibrated Trust |
Verifies selectively based on confidence signals |
Maintain - this is the target state |
Most AI interactions |
| Under-trust (AI Aversion) |
Rejects AI outputs even when correct |
Build trust incrementally through track record |
New users, after AI failures |
Trust Erosion Events (TEEs)
A single trust violation can undo weeks of reliable performance. Design for recovery:
| TEE Type |
Example |
Recovery Pattern |
| Confident hallucination |
AI states a false fact with no hedging |
Immediately acknowledge the error class; show what changed to prevent recurrence |
| Silent failure |
AI gives an answer but misses a critical constraint |
Add constraint-checking signals: "I accounted for X, Y, Z in this answer" |
| Inconsistency |
AI gives different answers to the same question |
Surface version/context differences: "This differs from my earlier answer because..." |
| Opacity |
User cannot understand why AI made a choice |
Retroactive explanation: "I recommended X because of [factors]. Here's what would change if..." |
Confidence Display Patterns
The Confidence Triad
Every AI output should communicate three dimensions of confidence:
| Dimension |
What It Tells the User |
Display Pattern |
| Certainty |
How sure is the AI about this specific output? |
Color-coded badge (green/amber/red) + percentage if available |
| Basis |
What evidence supports this output? |
Inline citations, source cards, "Based on..." prefix |
| Scope |
What does this answer cover, and what doesn't it cover? |
Explicit boundary statements: "This covers X but does not account for Y" |
Confidence Display Decision Matrix
| Context |
Show Numerical Confidence? |
Show Color Badge? |
Show Source Links? |
| Casual information lookup |
No - feels clinical |
Optional |
Yes, inline |
| Professional decision support |
Yes - precision matters |
Yes |
Yes, with expandable detail |
| Creative generation (writing, images) |
No - subjectivity makes numbers misleading |
No |
Show inspiration sources if applicable |
| Code generation |
Yes (test pass rate) |
Yes |
Link to documentation used |
| Medical/legal/financial |
Yes - accountability demands it |
Yes, conservative (amber default) |
Mandatory, with recency indicator |
Citation Architecture
Citations are the single highest-impact trust pattern for LLM-based products.
Citation Depth Levels
| Level |
What Users See |
When to Use |
| L0: No citation |
Raw AI output |
Only for creative/casual use cases with no factual claims |
| L1: Source attribution |
"Based on [source name]" |
Minimum for any factual claim |
| L2: Inline citation |
Numbered references linked to specific claims |
Professional, research, and decision-support contexts |
| L3: Quotable evidence |
Direct excerpts from sources with highlighting |
High-stakes contexts where users must verify independently |
| L4: Auditable trace |
Full reasoning chain + every source consulted + sources rejected |
Regulated industries, compliance, legal discovery |
Citation UI Patterns
| Pattern |
Implementation |
Best For |
| Superscript numbers |
Claim text[1] with footnotes |
Long-form responses, research |
| Inline source chips |
"According to WHO Guidelines 2025..." |
Conversational interfaces |
| Expandable evidence cards |
Collapsed by default, expand to show excerpt + link |
Decision-support dashboards |
| Side-panel source viewer |
Click citation, source appears in adjacent panel |
Document review, analysis tools |
| Confidence-colored highlights |
Text segments colored by source reliability |
Professional research tools |
Explanation Layering
Different users need different explanation depths at different moments. Design explanations that telescope from simple to deep.
The 5-Second / 5-Minute / 50-Minute Rule
| Layer |
Depth |
Content |
UI Pattern |
| 5-second |
Headline |
One sentence: what the AI did and its confidence |
Always visible - the response itself |
| 5-minute |
Summary |
Key factors that influenced the output, top 3 reasons |
Expandable section: "Why this answer?" |
| 50-minute |
Audit trail |
Full reasoning chain, all sources consulted, alternative answers considered |
Link to detailed view or export |
Anti-pattern: Dumping all three layers at once. The 50-minute layer should never appear unless explicitly requested.
The "Why?" Menu
Every non-trivial AI output should support a "Why?" interaction:
| "Why?" Question |
What to Show |
| "Why this answer?" |
Top 3 factors that influenced the output |
| "Why not [alternative]?" |
What would need to change for the alternative to be recommended |
| "What are you uncertain about?" |
Specific elements with lower confidence + what additional info would help |
| "What did you ignore?" |
Factors the AI considered but deprioritized, and why |
| "How would this change if...?" |
Sensitivity: what inputs would flip the recommendation |
Anti-Patterns
| Pattern |
Why It Fails |
| Showing confidence scores without context |
"87% confidence" means nothing without a baseline. Is 87% good or bad for this task? |
| Using green for everything |
If all outputs are green-badged, the badge system is meaningless. Users need contrast to calibrate |
| Burying explanations behind 3+ clicks |
If users can't reach the "why" in one interaction, they won't bother |
| Making citations look like legal disclaimers |
Dense, tiny-font footnotes signal "cover our liability" not "verify this yourself" |
| Explaining the model instead of the decision |
Users don't care about transformer architecture. They care about "why THIS recommendation for MY situation" |
| Only explaining when wrong |
If explanations only appear after errors, users associate explanation UI with unreliability |
Quick Reference
| Task |
Framework Element |
Key Deliverable |
| Add explainability to AI product |
Full GLASS framework |
Explanation layer architecture + citation depth map |
| Design confidence indicators |
Confidence Triad + Display Matrix |
Visual system with color, basis, and scope signals |
| Recover from trust violation |
Trust Erosion Events table |
Recovery flow with acknowledgment, explanation, and prevention |
| Audit an AI product for transparency |
Trust Calibration Spectrum |
Assessment of where users fall on the spectrum + design interventions |
| Add citations to LLM outputs |
Citation Architecture (L0-L4) |
Citation system matched to use case risk level |
Integration
Works with: ai-error-resilience (transparency about failures), ai-conversation-architect (confidence in dialogue), ai-safety-guardrails (transparency about content filtering), ai-feedback-loops (user corrections as trust signals).
1---2name: ai-trust-transparency3description: Design explainability interfaces that help users understand AI decisions, build calibrated trust, and verify AI outputs. Use when: AI explainability, XAI UX, confidence indicators, citation design, source attribution, trust signals, AI transparency, why did AI do this.4---56# AI Trust & Transparency78Design interfaces where users can see into the AI's reasoning, calibrate their trust appropriately, and verify claims independently. The GLASS framework makes AI decision-making visible without overwhelming users.910## Core Principle1112Trust is not a boolean. Users should not "trust AI" or "distrust AI" - they should develop **calibrated trust**: high confidence when the AI is reliable, healthy skepticism when it's uncertain. Your job is to give them the signals to calibrate correctly.1314---1516## The GLASS Framework1718| Letter | Principle | Design Question |19|---|---|---|20| **G** | Ground in Sources | Can the user trace every AI claim back to a verifiable source? |21| **L** | Layer Explanations | Can the user get a 5-second answer AND a 5-minute deep dive? |22| **A** | Advertise Limitations | Does the interface proactively tell users what the AI is NOT good at? |23| **S** | Show Confidence | Can the user see how certain the AI is about each output? |24| **S** | Support Override | Can the user correct, override, or reject AI outputs without friction? |2526---2728## The Trust Calibration Spectrum2930Design for the right trust level - not maximum trust.3132| Trust Level | User Behavior | Design Goal | When Appropriate |33|---|---|---|---|34| **Over-trust (Automation Bias)** | Accepts all AI outputs without checking | Introduce friction to encourage verification | High-stakes decisions (medical, financial, legal) |35| **Calibrated Trust** | Verifies selectively based on confidence signals | Maintain - this is the target state | Most AI interactions |36| **Under-trust (AI Aversion)** | Rejects AI outputs even when correct | Build trust incrementally through track record | New users, after AI failures |3738### Trust Erosion Events (TEEs)3940A single trust violation can undo weeks of reliable performance. Design for recovery:4142| TEE Type | Example | Recovery Pattern |43|---|---|---|44| **Confident hallucination** | AI states a false fact with no hedging | Immediately acknowledge the error class; show what changed to prevent recurrence |45| **Silent failure** | AI gives an answer but misses a critical constraint | Add constraint-checking signals: "I accounted for X, Y, Z in this answer" |46| **Inconsistency** | AI gives different answers to the same question | Surface version/context differences: "This differs from my earlier answer because..." |47| **Opacity** | User cannot understand why AI made a choice | Retroactive explanation: "I recommended X because of [factors]. Here's what would change if..." |4849---5051## Confidence Display Patterns5253### The Confidence Triad5455Every AI output should communicate three dimensions of confidence:5657| Dimension | What It Tells the User | Display Pattern |58|---|---|---|59| **Certainty** | How sure is the AI about this specific output? | Color-coded badge (green/amber/red) + percentage if available |60| **Basis** | What evidence supports this output? | Inline citations, source cards, "Based on..." prefix |61| **Scope** | What does this answer cover, and what doesn't it cover? | Explicit boundary statements: "This covers X but does not account for Y" |6263### Confidence Display Decision Matrix6465| Context | Show Numerical Confidence? | Show Color Badge? | Show Source Links? |66|---|---|---|---|67| Casual information lookup | No - feels clinical | Optional | Yes, inline |68| Professional decision support | Yes - precision matters | Yes | Yes, with expandable detail |69| Creative generation (writing, images) | No - subjectivity makes numbers misleading | No | Show inspiration sources if applicable |70| Code generation | Yes (test pass rate) | Yes | Link to documentation used |71| Medical/legal/financial | Yes - accountability demands it | Yes, conservative (amber default) | Mandatory, with recency indicator |7273---7475## Citation Architecture7677Citations are the single highest-impact trust pattern for LLM-based products.7879### Citation Depth Levels8081| Level | What Users See | When to Use |82|---|---|---|83| **L0: No citation** | Raw AI output | Only for creative/casual use cases with no factual claims |84| **L1: Source attribution** | "Based on [source name]" | Minimum for any factual claim |85| **L2: Inline citation** | Numbered references linked to specific claims | Professional, research, and decision-support contexts |86| **L3: Quotable evidence** | Direct excerpts from sources with highlighting | High-stakes contexts where users must verify independently |87| **L4: Auditable trace** | Full reasoning chain + every source consulted + sources rejected | Regulated industries, compliance, legal discovery |8889### Citation UI Patterns9091| Pattern | Implementation | Best For |92|---|---|---|93| **Superscript numbers** | Claim text[1] with footnotes | Long-form responses, research |94| **Inline source chips** | "According to `WHO Guidelines 2025`..." | Conversational interfaces |95| **Expandable evidence cards** | Collapsed by default, expand to show excerpt + link | Decision-support dashboards |96| **Side-panel source viewer** | Click citation, source appears in adjacent panel | Document review, analysis tools |97| **Confidence-colored highlights** | Text segments colored by source reliability | Professional research tools |9899---100101## Explanation Layering102103Different users need different explanation depths at different moments. Design explanations that telescope from simple to deep.104105### The 5-Second / 5-Minute / 50-Minute Rule106107| Layer | Depth | Content | UI Pattern |108|---|---|---|---|109| **5-second** | Headline | One sentence: what the AI did and its confidence | Always visible - the response itself |110| **5-minute** | Summary | Key factors that influenced the output, top 3 reasons | Expandable section: "Why this answer?" |111| **50-minute** | Audit trail | Full reasoning chain, all sources consulted, alternative answers considered | Link to detailed view or export |112113**Anti-pattern:** Dumping all three layers at once. The 50-minute layer should never appear unless explicitly requested.114115---116117## The "Why?" Menu118119Every non-trivial AI output should support a "Why?" interaction:120121| "Why?" Question | What to Show |122|---|---|123| "Why this answer?" | Top 3 factors that influenced the output |124| "Why not [alternative]?" | What would need to change for the alternative to be recommended |125| "What are you uncertain about?" | Specific elements with lower confidence + what additional info would help |126| "What did you ignore?" | Factors the AI considered but deprioritized, and why |127| "How would this change if...?" | Sensitivity: what inputs would flip the recommendation |128129---130131## Anti-Patterns132133| Pattern | Why It Fails |134|---|---|135| Showing confidence scores without context | "87% confidence" means nothing without a baseline. Is 87% good or bad for this task? |136| Using green for everything | If all outputs are green-badged, the badge system is meaningless. Users need contrast to calibrate |137| Burying explanations behind 3+ clicks | If users can't reach the "why" in one interaction, they won't bother |138| Making citations look like legal disclaimers | Dense, tiny-font footnotes signal "cover our liability" not "verify this yourself" |139| Explaining the model instead of the decision | Users don't care about transformer architecture. They care about "why THIS recommendation for MY situation" |140| Only explaining when wrong | If explanations only appear after errors, users associate explanation UI with unreliability |141142---143144## Quick Reference145146| Task | Framework Element | Key Deliverable |147|---|---|---|148| Add explainability to AI product | Full GLASS framework | Explanation layer architecture + citation depth map |149| Design confidence indicators | Confidence Triad + Display Matrix | Visual system with color, basis, and scope signals |150| Recover from trust violation | Trust Erosion Events table | Recovery flow with acknowledgment, explanation, and prevention |151| Audit an AI product for transparency | Trust Calibration Spectrum | Assessment of where users fall on the spectrum + design interventions |152| Add citations to LLM outputs | Citation Architecture (L0-L4) | Citation system matched to use case risk level |153154## Integration155156Works with: `ai-error-resilience` (transparency about failures), `ai-conversation-architect` (confidence in dialogue), `ai-safety-guardrails` (transparency about content filtering), `ai-feedback-loops` (user corrections as trust signals).
varunk130