Feedback

A powerful way to improve your AI models is to collect feedback from users about the responses they receieve and using that tdata to improve the models. Datawizz provides a simple way to collect feedback from users and use it to improve your AI models.

You can simply add feedback for any inference log - marking it as good or bad - and the feedback will be used in future model training to refine responses. In many cases, this simple feedback loop can significantly improve the quality of responses.

​

Adding feedback through the dashboard

You can use the Datawizz dashboard to manually add feedback to your logs. This can be useful when you want to review a batch of existing logs and provide feedback on them.

To do so, head over to the logs section and open any log. In the log details page, you will see a feedback section where you can mark the log as good or bad - simply click the thumbs up or thumbs down icon to provide feedback.

​

Sending feedback programmatically (via API)

If in your use case users interact with AI generated responses, it can be useful to collect feedback from them directly. You can do this by sending feedback programmatically via the Datawizz API.

​

When to collect feedback

​

Explicit feedback

Explicit feedback is when users provide feedback explicitly by clicking a thumbs up or thumbs down button or by providing a rating. In this flow, you’ll ask the users for feedback after they receive a response from the AI.

​

Implicit feedback

Implicit feedback is when you collect feedback based on user behavior. In implicit flows, you infer the feedback based on what the user does after receiving a response. For example, if a user saves the response generated by the AI, you can infer that the response was good. If the user skips the response or asks for another response, you can infer that the response was not good.

Implicit flows might be more natural for users as they don’t have to explicitly provide feedback, but they can be harder to implement and might not always be accurate. Explicit flows are more straightforward but might require more effort from the user and have lower participation rates.

​

Sending feedback via API

To send feedback via the API, you can use the PATCH /log/:log_id endpoint with the quality_rating field set to 👍 or 👎.

When you generate a response via the OpenAI / Anthropoic compatible API endpoints, Datawizz will send back the log id in the response. Simplt use this log id to send feedback via the API endpoint.