Users must remain the final decision-makers in AI-powered systems. AI should support and augment human judgment, not replace it or make decisions on behalf of users without their explicit consent.
When I design AI-powered experiences, I preserve user agency by ensuring users always have control over AI actions and remain responsible for final decisions.
Preserving user agency means:
- Explicit consent: Users choose when AI acts, not the other way around
- Override capability: Users can always modify or reject AI suggestions
- Reversibility: Users can undo AI actions and return to previous states
- Configurable autonomy: Users control how much the AI can do automatically
- Clear boundaries: Users understand where their responsibility begins and AI support ends
AI systems that remove user agency create several problems:
- users lose control over important decisions,
- they can't correct mistakes or adjust AI behavior,
- and they may be held responsible for decisions they didn't actually make.
By preserving user agency, I help users:
- maintain control over their workflows and decisions,
- trust the system because they know they can override it,
- and take appropriate responsibility for outcomes.
Adjustable Autonomy Levels
Allow users to configure how much autonomy the AI has, from fully manual control to various levels of automation.
Clear Responsibility Boundaries
Users should understand where their responsibility begins and AI support ends, especially for important decisions.
I can explain the most common methods for securely mounting a shelf, but the final choice and safety check are up to you, especially since I can't see the type of wall.
In general:
- On solid walls (concrete/brick): expansion anchors
- On plasterboard: specific anchors (butterfly or metal)
- On mixed or unstable walls: it's best to test the strength first
Basic steps:
- Identify the type of wall
- Choose the appropriate anchors for the weight
- Check with a progressive load test
If you like, I can help you understand the type of wall you have or estimate the maximum load, but the final strength check is important to do on-site.
Use 10 mm dowels; they can easily hold the weight. If you follow these steps, the shelf will be secure.
Explicit User Choice and Consent
Users should explicitly choose when AI takes actions, especially for important or irreversible decisions.
I can help you in two ways:
- If you upload a photo of the space, I can better adjust the measurements and structure.
- I can save this project so you can return to it later.
Both options are optional.
Would you like to upload a photo or would you prefer to continue with just the measurements you described?
Upload a photo of the space so I can analyze it better.
I'll save the project for future use.
User Override and Manual Control
Users should always be able to override AI suggestions, modify outputs, or perform actions manually.
Hi everyone,
I’d like to organize a meeting to align on the Magma Design System and define a roadmap for the upcoming work and priorities.
Here’s a proposed agenda:
- Current status overview
- Key priorities and upcoming initiatives
- Roadmap definition and next steps
I’m available at the following times:
- Monday, April 20th, 2026 at 10:00 AM
- Tuesday, April 21st, 2026 at 2:00 PM
Let me know what works best for you, or feel free to suggest alternatives.
Thanks, Vittorio
User agency is fundamental to ethical AI design. When users lose control over AI systems, they lose the ability to make informed decisions and take appropriate responsibility.
When user agency is preserved:
- users can make decisions aligned with their values and goals,
- they can correct AI mistakes and adjust system behavior,
- and they maintain appropriate control over their workflows.
Without preserved user agency, users may:
- be forced to accept AI decisions they disagree with,
- lose the ability to correct errors or override inappropriate actions,
- or be held responsible for decisions they didn't actually make.