Frame the design decision
Treat user trust as a design outcome. Decide what the user needs to see to understand the AI’s role, assess its confidence, and take a safe next action.
Pathway 2 · Lesson 0460 min
Designing for Trust & Explainability
Transparency, fairness, and responsible AI UX
Learning Objective
By the end of this lesson, you will be able to: Apply explainability design patterns to an AI feature, identify three sources of AI bias, and create a responsible AI design checklist.
75%
of employees are concerned AI will make their jobs obsolete
Engage
"Trust is the foundation of AI adoption. Without it, even the best AI product fails. How do you design AI experiences that earn trust rather than erode it?"
Explore
Google's principles for designing AI that users trust — practical UX patterns for transparency and explainability.
Explain
Why Users Don't Trust AI
Three root causes of AI distrust:
1. Opacity: 'I don't know why it said that.' Users can't see the AI's reasoning, so they can't evaluate its reliability.
2. Errors: 'It was wrong before.' Past errors create lasting distrust, especially if they weren't caught.
3. Loss of control: 'I can't change it.' Users who feel they can't override or correct AI decisions disengage or resist.
Designing for trust means addressing all three — making AI transparent, reliable, and controllable.
Explain
Explainability Design Patterns
Four patterns that make AI more explainable:
1. Show reasoning: 'I recommended this because...' Display the factors that drove the AI's decision.
2. Confidence indicators: Show when the AI is uncertain. A confidence score or 'I'm not sure about this' message prevents over-reliance.
3. Source attribution: 'Based on [document/data]...' Show where the AI's information came from.
4. Alternative options: Show 2-3 alternatives alongside the recommendation. This signals that the AI's output is a suggestion, not a decree.
⚠️
Automation Bias: The Hidden Risk
Automation bias is the tendency to over-rely on automated systems — accepting AI outputs without sufficient scrutiny because 'the computer said so.' Good AI UX actively fights automation bias by making AI uncertainty visible, requiring human confirmation for high-stakes decisions, and designing friction into the acceptance of AI recommendations.
Elaborate · Worked Example
Explainability in Practice: AI Loan Decisions
Context
A bank uses AI to make initial loan approval decisions.
The Situation
Without explainability: 'Your application was declined.' The customer has no recourse and no understanding. With explainability design: - Show the factors: 'Your application was assessed on: credit score (weight 40%), income stability (30%), debt ratio (30%)' - Show confidence: 'This was a borderline decision — a human reviewer can reconsider' - Show alternatives: 'You could qualify for a smaller loan of £X' - Show control: 'Request a human review of this decision'
The Outcome
Explainability transforms a black-box rejection into a transparent, actionable, trustworthy interaction — even when the answer is no.
Try It · Reflect & Apply
Audit an AI feature you use or are designing. Score it on the four explainability patterns (0-3 each): Does it show reasoning? Confidence? Sources? Alternatives? What would you change to improve trust?
Your notes are saved in your browser for this session.
Evaluate · Knowledge Check
1 / 3What is 'automation bias' and why is it a design problem?
Design Practice Studio
Trust Design Studio
Create the artefact
Create a Trust Canvas covering explanation, sources, confidence, alternatives, override control, error recovery, and a route to human support.
Test with evidence
Ask five users: ‘Would you act on this? Why or why not?’ Then test an intentionally wrong recommendation to see whether they notice, question, and recover.
Spaced revisit
Revisit after one week and prioritise the biggest trust gap: opacity, error handling, or loss of control. Make one visible design change.
External video references
Keep Learning on YouTube
Open in a new tab
Ready to move on?
Mark this lesson complete and continue to "AI Design Sprint Facilitation".