Principles of AI Design

The Principles of AI Design.

The following principles of designing AI products are inspired by Dieter Rams now legendary Principles of Design. These are intended to explicitly call out what makes an AI application useful, good and capable of evoking delight.

1. Good AI design is innovative

The possibilities for innovation are not exhausted. In design and technology, we are at the beginning of the beginning. Technological development continuously offers new opportunities for innovative design. As the context moves, so too does the necessity and opportunity for design innovation. 

2. Good AI design makes a product useful

Products are made to be used. Good AI design emphasises the usefulness of a product whilst removing superfluous functionality and unnecessary interactions.

3. Good AI design is aesthetic

What we use affects who we are and how we feel. Imbuing an AI product with aesthetic quality is integral to its usefulness and predictive of its used-ness. This is not surface detail. Only well-executed objects can be beautiful. If adding AI to the product doesn’t make it more pleasing to use, it has no place in the product. 

4. Good AI design makes a product understandable

AI abstracts away decision-making and interaction. Good AI clarifies the product’s structure. It explains itself simply and clearly. You know what it does, and what to expect from it. 

5. Good AI design is unobtrusive

AI products fill a purpose. They don’t exist to show off their algorithmic sophistication. They have no business wasting the user’s time so they can greedily shake them down for data. AI design should be restrained and ask as little as possible from the user.

6. Good AI design is honest

Passing a dumb product off as intelligent is dumb. It disrespects the user and sets them up for disappointment.  Good AI design signals the boundary of its intelligence and makes no pretences or apologies for what it cannot do. 

7. Good AI design learns

Good AI design is not a one-hit wonder; a static application of a single model. Good AI design is built to evolve with changing user needs, user habits and improving technical art. It forms the basis of future AI, not throwaway models for a throwaway crowd. 

8. Good AI design is detailed

Nothing must be arbitrary or left to chance. Care in the AI design process shows respect towards the consumer. A brittle model in a brittle technical context disrespects the user. Good AI design considers how the user will interact with it within the constraints of data collection and labelling. 

9. Good AI design preserves attention

Attention is the non-renewable resource of digital design. Good AI design values the user’s attention. Good AI design doesn’t make the user think. It values and defends the attention of the user.

10. Good AI design is as little design as possible

Less, but better. AI abstracts away unnecessary, decision-making and removes friction-filled interactions because it concentrates on the essential aspects of the product. Back to simplicity. Back to the job to be done. 

11. Good AI design builds a relationship

Good AI design learns from the users’ interactions. It builds a relationship of trust over time. The more the user interacts with the product, the more the user invests, the more they get in return. Like friendships and communities, Good AI design is designed for the long term. 

1. Good AI design is innovative

2. Good AI design makes a product useful

3. Good AI design is aesthetic

4. Good AI design makes a product understandable

5. Good AI design is unobtrusive

6. Good AI design is honest

7. Good AI design learns

8. Good AI design is detailed

9. Good AI design preserves attention

10. Good AI design is as little design as possible

11. Good AI design builds a relationship