Carbon for AI

AI introduces a range of new responsibilities when designing experiences across our software, particularly regarding trust and transparency. Carbon for AI is the framework for identifying AI-generated content and delivering explainability in our products.

Transparency of AI presence is key to user trust. Having a consistent identity helps build awareness and anticipation of AI presence across our experiences.

Whether AI exists within a single word or across the entire page, users should clearly understand how pervasive an AI is implemented.

This integration requires you to minimize distractions and show explanations only when needed or requested by the user.

While these new styling elements are enticing, they are strictly intended to identify any instances of AI being used in an experience.

Carbon for AI styling builds on the core Carbon library principles, using light as a metaphor to “illuminate” AI-generated content. This makes AI-generated or AI-recommended content distinctive and stands out for users. For IBMers only: To learn more about the Carbon for AI styling, see Design for AI: Style elements.

Carbon for AI uses light-inspired effects like brightness, glow, and gradients to emphasize AI instances across the experience.

Carbon for AI is designed to coexist alongside the Carbon themes but has its own unique visual characteristics that distinguishes it as AI while following the IBM Design Language DNA and principles.

To accommodate for these AI styles, we are introducing a new suite of color and style tokens that can be found within the main Carbon themes. Using the AI tokens does not require a different theme or elements package. When building custom AI component variants or instances in your UI use these new color tokens to stay consistent with the AI styling. For a full list of the new tokens, go to color tokens.

Explainability is a set of processes and methods that make AI decisions more comprehensible to humans. For IBMers only: To learn more about explainability, see Design for AI: Explainability.

The AI label is the primary indicator to communicate that AI is present, and it’s how users can access explainability content.

As a first layer of explainability, use the explainability pop-over attached to the AI label to provide a quick, in-context option to explain AI results for a particular instance.

The AI label is a component used as an indicator of AI instances in the UI. It is intended for any scenario where AI is generating content to reinforce AI transparency, accountability, and explainability at any interface level. It enables more effective recognition and recall of AI instances in a way that is identifiable across any IBM product.

The AI label is also the trigger for the explainability popover which serves as the first layer of explainability. It provides a consistent, up-front way to access explainability with the option to dig into more details if needed.

AI transparency is key, and the AI label is an accessible, interactive element that can mark these instances as presented in the interface. The label has been designed to work in many instances where AI-generated content can be surfaced as in components, see the component section for more details.

For recognition and recall, users need a consistent, identifiable visual reference to look to when they need to understand more details about how the AI was built.

Users must clearly understand how to access more details about how an AI was built. By providing a bridge to explainability from the AI label, users are given a recognizable location to learn more about the AI model.

For more information about using the AI label, see the AI label component page.

AI chat is a conversational framework between a user and an AI that can aid in creating tasks, finding insights, tracking documents, and more. For IBMers only: To learn more about chat, see Design for AI: Chat pattern.

For IBMers only: The chat component Figma library includes design assets, components, and generals chat usage guidelines.

Carbon components with an AI presence are styled using Carbon for AI. Each AI component is required to have an embedded AI label and explainability popover that alerts users to AI-generated content. When the AI label is turned on, the component transforms to the AI visual style but still functions like the standard one.

The AI presence styling for core Carbon components is available in the “(v11) All themes” Figma library. To use it, simply grab the normal component asset, and then in the layers panel, go to “Change variable modes” and choose “AI presence,” followed by “All AI elements.”

Components in Carbon come with the AI style properties already available for use through a mix-in. For more details on how to use the AI label in code, see the React and Web-components storybooks.

These components can toggle between the AI variant and the default variant depending on the user’s interaction. If the user manually overrides the AI-suggested content then the component will change from the AI variant to the default variant. Once edited, the user should still be able to switch back to the initially AI-generated content via a revert to AI button.

The following list of components are the core components that have an available AI variant. All of the AI components in Carbon have been designed to work harmoniously together, as parts of a greater whole.

When designing an AI experience, ask if the user knows AI is being used. The base principle of Carbon for AI is to mark where AI is present while providing explainability whenever available. This level of transparency helps build trust between users and the system. This transparency is crucial as the line between AI and human is becoming increasingly blurred.

The Carbon for AI style elements and components are tested for accessibility to ensure that colors, type, and contrast adhere to Carbon’s existing accessibility standards.

The AI label is the primary indicator to communicate that AI is present. To ensure that the AI label remains accessible across light and dark themes, use the provided component.

Carbon for AI utilizes light sources to distinguish it from base Carbon components. These light sources glow on container edges, but the spread is limited and subtle to ensure accessible contrast with UI content.

Help us improve this component by providing feedback, asking questions, and leaving any other comments on GitHub. For IBMers only: Questions and feedback can be directed to Slack in the channels #carbon-design-system or #carbon-for-ai.