US 12,474,938 B2
Feature exposure for model recommendations and feedback
Ratna S. Desai, Oakland, CA (US); and Alex M. Cook, III, San Francisco, CA (US)
Assigned to GOOGLE LLC, Mountain View, CA (US)
Filed by GOOGLE LLC, Mountain View, CA (US)
Filed on Jun. 30, 2023, as Appl. No. 18/217,371.
Application 18/217,371 is a continuation of application No. 17/671,284, filed on Feb. 14, 2022, granted, now 11,734,034.
Application 17/671,284 is a continuation of application No. 16/425,461, filed on May 29, 2019, granted, now 11,249,772, issued on Feb. 15, 2022.
Claims priority of provisional application 62/677,615, filed on May 29, 2018.
Prior Publication US 2023/0350705 A1, Nov. 2, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 9/451 (2018.01); G06F 3/0482 (2013.01); G06N 5/04 (2023.01); G06F 3/023 (2006.01)
CPC G06F 9/453 (2018.02) [G06F 3/0482 (2013.01); G06N 5/04 (2013.01); G06F 3/0237 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method implemented by one or more processors, the method comprising:
selecting, using an action selection model that processes decision data, a recommended action,
wherein selecting the recommended action is from a plurality of actions that a user device may take and that are each different from each other action, and
wherein the decision data comprises user data describing selections responsive to recommended actions by one or both of a particular user or interests of the particular user;
providing, by the user device, an interaction element that describes the recommended action, wherein the user device is configured to perform the recommended action in response to a particular selection of the interaction element;
receiving user input to select the interaction element;
in response to receiving the user input to select the interaction element that describes the recommended action, providing, by the user device, (i) at least a portion of the decision data processed by the action selection model used to select the recommended action described by the interaction element and (ii) a selectable feedback element that, when selected, generates feedback data describing user satisfaction with the recommended action; and
in response to receiving an indication that the feedback element has been selected by a user, causing the feedback data to be included in the user data, wherein subsequent selections using the action selection model are adjusted based on the feedback data included in the user data.