Response regeneration
Enables users to generate an alternative response in generative AI chat.
Key UX concepts
Regeneration
A user can request a new version of a model's response, which generates an alternate response to replace the original one. This approach maintains the conversation's prior context while allowing users to explore alternative responses to their questions. The same process applies whether the original response was completed or interrupted by the user through a stop action.
Model reprocessing
During regeneration, the model performs a new sampling process to generate an alternative response. While using the same prompt and conversation context, the model can produce varied outputs due to the probabilistic nature of language model generation. This is similar to getting different arrangements each time you shuffle a deck of cards.
Distinguishing regeneration from re-prompting
Regeneration happens when a user selects an explicit regenerate icon button. The model then replays the exact same input to generate a new response. Alternatively, typing a phrase like “try again” or “regenerate that” is a new, natural language prompt that the system interprets contextually as a new request.
How regeneration affects context
Context builds a shared understanding between user and system throughout their conversation. When a user regenerates a response, it changes the conversation's path from that point, like traveling back in time to take a different direction forward. To maintain consistent context, the system removes all messages that came after the regenerated response. This ensures the conversation can continue naturally along its new path without conflicting with messages from the previous direction.
Building blocks
A. User prompt
The user's original question using the chat bubble outgoing message with an avatar to indicate it's a sent message.
B. Generated response
The system’s response is displayed using the chat bubble incoming message and generative AI avatar to indicate it's a received message.
C. Regenerate button
An icon button within the button group with a refresh icon that allows users to request an alternative response to the same prompt. Appears consistently below each eligible system response.
D. Stop action - optional
After a user has sent a prompt, the send button changes to a stop button which enables users to interrupt the generation or processing of a response.
Common use cases
Regenerate a completed response
Users can regenerate a completed response to see alternative versions of the same response. The action reuses the original user prompt and context so the conversation stays coherent while exploring new outputs.
Regenerate a stopped response
Users can stop generation as soon as they notice an unwanted response direction. This prevents the model from consuming additional tokens generating an unwanted response. The regenerate option becomes immediately available, letting users quickly pivot to better alternatives. For example, if a user sees the response taking an overly technical approach, they can stop the generation and use regenerate to get a different response to their original prompt without changing anything.
General guidelines
Do
- Place regeneration action in button group to create a more accessible tab order, since regenerating is typically selected after a user sent a response in the prompt input.
- Add a regeneration button only to user-initiated responses that can be regenerated without breaking context.
- Allow users to regenerate after stopping a prompt and also after system has completed a response.
- Follow error states for Generative AI chat to handle failed regeneration attempts.
Don't
- Don't add additional friction to stopping generating responses. Users should be able to stop generation immediately.
Writing guidelines
General writing guidelines
Use sentence case, but continue to capitalize proper nouns and brand names correctly in context.
Use end punctuation, except in headers and buttons. Don’t use exclamation points.
Use present-tense verbs and active voice.
Don't use please, thank you, ellipsis (...), ampersand (&), e.g., i.e., or etc. in writing.
Avoid directional language.
For example: use previous not above, use following not below.
Use device-independent language.
For example: use choose or select not click.
Accessibility guidelines
General accessibility guidelines
Follow the guidelines on alternative text and Accessible Rich Internet Applications (ARIA) regions for each component.
Make sure to define ARIA labels aligned with the language context of your application.
Don't add unnecessary markup for roles and landmarks. Follow the guidelines for each component.
Provide keyboard functionality to all available content in a logical and predictable order. The flow of information should make sense.