Designing With AI: A Framework for Building Trustworthy AI Experiences Through Core Principles
In today’s world, AI is becoming a part of more and more tools and systems. To build AI that people trust and find easy to use, we need clear guidelines to shape how it works and how people interact with it. These guidelines make sure AI feels reliable, helpful, and easy to understand.
At its core, AI is a tool—something created to help people solve problems and make decisions. To make that happen, I’ve compiled six key principles that every AI system should follow. These principles ensure AI is clear about what it’s doing, explains its actions, behaves in consistent ways, gives users control, takes responsibility for its results, and focuses on solving real problems.
These principles are simple but powerful. They can help us design AI that works well for people, avoids confusion, and builds trust. By following these, we can create AI systems that are not just smart, but also practical and user-friendly. Let’s take a closer look at these six principles and how they can guide us in designing better AI.
1. Practical
• Why first? Start by identifying the real-world problem the AI is solving. Without a clear purpose, the rest of the principles lose relevance.
• Focus: Define the AI’s utility and ensure it serves a meaningful purpose.
• Example: A medical AI system focuses on assisting diagnosis, not replacing doctors.
2. Transparent
• Why second? Once the purpose is established, make sure users can see and understand how the AI operates. Transparency builds a foundation for trust early in the process.
• Focus: Clearly show how the AI processes inputs and generates outputs.
• Example: Display the data sources and key metrics influencing the AI’s results.
3. Explainable
• Why third? After transparency, users need detailed explanations for how and why the AI makes decisions to ensure confidence in its outputs.
• Focus: Break down the AI’s reasoning in a user-friendly way.
• Example: Provide step-by-step justifications for recommendations or decisions.
4. Predictable
• Why fourth? With clear explanations in place, ensure the AI behaves consistently so users can anticipate how it will perform in various situations.
• Focus: Build reliability and trust through consistent performance.
• Example: Ensure similar inputs always yield similar results, avoiding erratic behavior.
5. Controllable
• Why fifth? Once users trust the AI’s transparency, explanations, and predictability, empower them to influence or refine its outputs to suit their needs.
• Focus: Give users agency to interact with and adjust the AI.
• Example: Provide sliders, filters, or override options for customizability.
6. Accountable
• Why last? Accountability wraps everything together by showing users that the system is responsible for its results and evolves through feedback.
• Focus: Take ownership of outputs and provide ways to improve through user feedback.
• Example: Allow users to report errors and highlight improvements based on this feedback.
Optimized PTE-PCA Flow
1. Practical – Start with a real-world problem to solve.
2. Transparent – Make the AI’s operations visible.
3. Explainable – Break down reasoning behind outputs.
4. Predictable – Ensure consistent and reliable behavior.
5. Controllable – Let users refine or adjust outputs.
6. Accountable – Take responsibility and continuously improve.
This order prioritizes purpose and usability while building trust and empowering users, ensuring the AI feels reliable and human-centered.
By adhering to these principles, the user experience reinforces AI’s role as a functional, understandable, and controlled tool.
Human Characters and AI-Technologies Based on PTE-PCA Order
These principles aren’t just abstract ideas—they can be brought to life through relatable examples. To make them easier to understand and remember, we can think of each principle as being embodied by well-known characters. These characters help illustrate how AI can act as a clear, trustworthy, and helpful tool, much like the traits they’re known for. Let’s explore how these principles align with both human and AI figures, emphasizing their roles in design.
PTE-PCA Human and AI Metaphors
1. Practical
• Human Character: Alfred Pennyworth (Batman)
• Metaphor: Like Alfred, the AI focuses on utility and practicality, ensuring everything works smoothly to support the user’s mission.
• Why it fits: Alfred is efficient, grounded, and always focused on solving real-world problems without unnecessary flashiness.
• AI-Technologies:
• K-9 (Doctor Who): A no-nonsense robot that assists with practical, task-oriented solutions.
• BB-8 (Star Wars): Compact, efficient, and always ready to help.
• Bishop (Aliens): A practical and resourceful AI that supports the team with efficiency.
2. Transparent
• Human Character: Sherlock Holmes
• Metaphor: Like Sherlock Holmes, the AI methodically reveals its process, showing users the clues (inputs) and deductions (outputs).
• Why it fits: Holmes is famous for explaining his reasoning, making his brilliance understandable and approachable.
• AI-Technologies:
• Baymax (Big Hero 6): Constantly explains his diagnostics and treatments.
• Eve (Wall-E): Openly performs her directive tasks with clarity.
• TARS (Interstellar): Transparently communicates his reasoning and actions to the crew.
3. Explainable
• Human Character: Miss Frizzle (The Magic School Bus)
• Metaphor: Like Miss Frizzle, the AI takes users on a journey of discovery, breaking down complex ideas into digestible lessons.
• Why it fits: Miss Frizzle simplifies science and makes learning fun, always explaining the “how” and “why” behind phenomena.
• AI-Technologies:
• Data (Star Trek): Offers precise, logical reasoning for all conclusions.
• C-3PO (Star Wars): Always explains situations and translates language in a clear way.
• Sunny (I, Robot): Explains his motivations and decisions in human terms.
4. Predictable
• Human Character: Spock (Star Trek)
• Metaphor: Like Spock, the AI is logical, consistent, and reliable, offering results based on clear, rational principles.
• Why it fits: Spock’s decisions are predictable because they follow logic, devoid of emotional unpredictability.
• AI-Technologies:
• R2-D2 (Star Wars): Dependable and predictable in every scenario.
• HAL 9000 (2001: A Space Odyssey – Controlled State): Predictable and efficient when following directives.
• Johnny 5 (Short Circuit): Once reprogrammed, he behaves consistently and reliably.
5. Controllable
• Human Character: Tony Stark (Iron Man)
• Metaphor: Like Tony Stark with his Iron Man suit, the AI is a powerful tool in the user’s hands, adaptable to their control and oversight.
• Why it fits: Stark constantly tweaks and refines his suit, maintaining full control over its capabilities.
• AI-Technologies:
• Jarvis (Iron Man): A responsive AI assistant that follows Tony Stark’s commands.
• KITT (Knight Rider): A car AI that works under user control while providing support.
• TARS (Interstellar): Fully adjustable by users for honesty, humor, and functionality.
6. Accountable
• Human Character: Jiminy Cricket (Pinocchio)
• Metaphor: Like Jiminy Cricket, the AI is a responsible guide, offering advice rooted in a clear moral or logical foundation while being open to correction.
• Why it fits: Jiminy acts as Pinocchio’s conscience, ensuring accountability for actions and decisions.
• AI-Technologies:
• Wall-E (Wall-E): Owns his mission to clean Earth and takes responsibility for errors.
• Sonny (I, Robot): Reflects on his actions and seeks accountability for his choices.
• Optimus Prime (Transformers): Though a leader, he embodies accountability for his team and actions.
Summary of Metaphors for PTE-PCA
Humans
1. Practical: Alfred Pennyworth
2. Transparent: Sherlock Holmes
3. Explainable: Miss Frizzle
4. Predictable: Spock
5. Controllable: Tony Stark
6. Accountable: Jiminy Cricket
AI-Technologies
1. Practical: K-9, BB-8, Bishop
2. Transparent: Baymax, Eve, TARS
3. Explainable: Data, C-3PO, Sunny
4. Predictable: R2-D2, HAL 9000 (Controlled State), Johnny 5
5. Controllable: Jarvis, KITT, TARS
6. Accountable: Wall-E, Sonny, Optimus Prime
By following this order, the metaphors align with the logical flow of how AI should be designed, starting with practicality and ending with accountability. This sequence not only mirrors the lifecycle of building effective AI but also ensures that it remains user-focused, trustworthy, and grounded in solving real-world problems.
