Labeling for Automated and AI-Generated Data

There are many things that we can do to create better experiences for our users through the smart application of automatic and AI-assisted technologies. Some of these can happen under the hood of our product, while others need to be explicitly labeled for compliance reasons. In this document we will review the details and provide the best means of representing them on a screen.  

 

AI-generated content and automated data will use the same visual treatment in Forge, but labels will diverge based on features and content. Below, we provide recommendations for visual treatment and label styling, but Forge visual guidelines are NOT a replacement for legal and compliance content guidance. Before designing with AI or automated data, please read the following legal bulletins in full:

• Artificial Intelligence (AI) Development and Deployment Policy
• Artificial Intelligence (AI) Data Handling Standard

Product teams implementing content covered by these two legal bulletins are responsible for compliance.

 

Where can I go for questions surrounding the legality and compliance of AI features or automated data?

• askHub-Legal New Request
• LegalEase

• Artificial Intelligence (“AI”):
  • Any engineered or machine-based system that is capable of performing tasks typically requiring human intelligence, such as visual perception, speech recognition, decision-making, and natural language understanding (collectively, “AI”). AI includes, but is not limited to, algorithmic, interpretive, machine learning, and generative AI. AI does not include automation.
  • Example: a feature that provides generated responses for patient messaging.
• Automated Data (or automation):
  • A system that follows pre-programmed rules, often for routine or repetitive tasks.
  • Example: suggested zip code prefill based on deterministic population logic.

• Be Transparent & Avoid Deception: Be upfront and transparent about the use of AI or automated data. Don't use misleading labels that could make users believe the content is human generated.  
• Clarity: Make sure the label is clear, easy to understand, and noticeable.  
• Accuracy: Ensure the label accurately reflects the level of AI involvement or data automation.  

Automated data or automation refers to data or information that is delivered to the user from a system that follows pre-programmed rules. In its simplest form, this would include things like pre-populating information in a field from something that was previously entered. Another kind of automated data is the population of information based on deterministic logic, from either an internal or trusted external source.

 

Various types of automated data have different labeling requirements. These guidelines have been developed with Legal and Compliance to ensure safety and accuracy in our work, but please reach out to Legal using askHub-Legal New Request for any questions. The table below outlines more specifics around each type of automated data and its requirements:

AI-Generated features are product features made using any engineered or machine-based system that is capable of performing tasks typically requiring human intelligence. AI includes, but is not limited to, algorithmic, interpretive, machine learning, and generative AI. This includes things like AI written text and summaries.

 

AI-Generated Data must always be accompanied by a notification of its use. Please reach out to Legal using askHub-Legal New Request for any questions.

Both automated data (when required) and AI-generated features will use a variant of the bolt icon and appropriate labeling based on visual hierarchy needs.

The Bolt icon now has an info-purple variant specifically for use with Automated Data and AI-Generated use cases. A new alert icon variant called BoltAlert has also been added for these purposes. 

• Icon with Tooltip: Use for simple automated data labeling, such as simple prefill. Tooltip should contain source of prefill or information.
• Icon with Popover: Use Popovers that appear on click to reveal the AI-generated nature of the content or show source logic for complex automated data. This can be helpful for saving space and maintaining a clean UI.  
• Text with Icon: Use in scenarios where disclosure is needed but no additional source logic is needed. Such as in a form where information is prefilled from a previous field. 
• Visual Tag: Apply tags or overlays directly on the content, like on images or videos.  
  Include the label within the content itself, such as in a corner of an image.  
• Inline Alert: Use in  higher visibility scenarios where disclosure is needed but no additional source logic is needed. Or, combine Inline Alert with icon with popover for additional source logic.
• Content Label Box: Use in scenarios where strong visual hierarchy is needed, this labeling combines eye catching color with visual tags and the option for additional data with popovers or the removal of logic with buttons.

Refer to the ‘Alert and Status’ page in Figma libraries for these components. 

• Automated data
  • Refer to the chart discussing automated data labeling requirements.
• AI-generated features
  • Legal and compliance prefer that AI-generated content labels use the text "has been generated by AI." For example, "This content has been generated by AI" or "These label suggestions have been generated by AI".

 

For label, disclosure, or logic questions please reach out to Legal using askHub-Legal New Request.

When there isn’t enough room for a descriptive label, or there is more supplemental information than is practical to show on the screen, use Popovers and Tooltips. These assets can help interested users understand the use of automated data and AI-generated features without distracting power users who have visited the page frequently.

Use a simple text label to highlight a section of automated data that needs low-level visual hierarchy.

Note: In this instance, not all fields are related to the automation, so this designer opted to use the Bolt icon to clarify which fields were automated. This meets compliance needs as it is explicit in communicating which fields have been automated.

Use an InlineAlert with a Popover for a feature that needs labeling with medium-level visual hierarchy and hidden supplemental information.

Use high visibility content label boxes to draw attention to complex and/or lengthy uses of automated data or AI-generated features.