Generating and visualizing bias scores representing bias in digital segments within segment-generation-user interfaces

Patent metadata

Field	Value
Publication number	US-11556567-B2
Application number	US-201916411515-A
Country	US
Kind code	B2
Filing date	May 14, 2019
Priority date	May 14, 2019
Publication date	Jan 17, 2023
Grant date	Jan 17, 2023

Abstract

This disclosure relates to methods, non-transitory computer readable media, and systems that generate and visualize bias scores within segment-generation-user interfaces prior to executing proposed actions with regard to target segments. For example, the disclosed systems can generate a bias score indicating a measure of bias for a characteristic within a segment of users selected for a proposed action and visualize the bias score and corresponding characteristic in a segment-generation-user interface. In some implementations, the disclosed systems can further integrate detecting and visualizing bias as a bias score with selectable options for a segmentation-bias system to generate and modify segments of users to reduce detected bias.

First claim

We claim:

1. In a digital medium environment for dynamically identifying and selecting a digital population segment prior to distributing digital content across computer networks, a computer-implemented method for efficiently visualizing biases for digital population segments within graphical user interfaces comprising:
providing, for display together within a segment-generation-user interface, a segment visualization depicting a segment of users selected for a proposed action, a bias-detection option selectable for generating bias scores for the segment of users, a bias indicator element for depicting the bias scores, and a reduce-bias option selectable for modifying a distribution of the users within the segment of users to reduce the bias scores;
based on receiving an indication of a selection of the bias-detection option:
generating a bias score specific to a user-selected characteristic and representing a measure of group bias and a measure of individual bias corresponding to the user-selected characteristic;
updating the bias indicator element within the segment-generation-user interface to reflect the bias score specific to the user-selected characteristic; and
generating a modified segment visualization for the segment of users depicting the distribution of the users within the segment according to the user-selected characteristic; and

based on receiving an indication of a selection of the reduce-bias option, modifying the distribution of the users within the segment to reduce the bias score in relation to the user-selected characteristic.

2. The computer-implemented method of claim 1 , further comprising providing, for display within the segment-generation-user interface, the modified segment visualization for the segment of users depicting the distribution of the users within the segment according to the user-selected characteristic.

3. The computer-implemented method of claim 1 , further comprising:
providing, for display within the segment-generation-user interface, user-characteristic options corresponding to characteristics of users within a user dataset and a machine-learning-segment-generation option for generating segments of users according to selected user-characteristic options; and
based on receiving indications of selections of one or more of the user-characteristic options and the machine-learning-segment-generation option, generating the segment of users selected for the proposed action using a segment-generation-machine-learning model.

4. The computer-implemented method of claim 1 , wherein modifying the distribution of the users within the segment to reduce the bias score in relation to the user-selected characteristic comprises adding users to the segment of users or excluding users from the segment of users.

5. A non-transitory computer readable medium storing instructions thereon that, when executed by at least one processor, cause a computer system to:
provide, for display together within a segment-generation-user interface, a segment visualization depicting a segment of users selected for a proposed action, a bias-detection option selectable for generating bias scores for the segment of users, a bias indicator element for depicting the bias scores, and a reduce-bias option selectable for modifying a distribution of the users within the segment of users to reduce the bias scores;
based on receiving an indication of a selection of the bias-detection option:
generate a bias score specific to a user-selected characteristic and representing a measure of group bias and a measure of individual bias corresponding to the user-selected characteristic;
update the bias indicator element within the segment-generation-user interface to reflect the bias score specific to the user-selected characteristic; and
generate a modified segment visualization for the segment of users depicting the distribution of the users within the segment according to the user-selected characteristic; and

based on receiving an indication of a selection of the reduce-bias option, modify the distribution of the users within the segment to reduce the bias score in relation to the user-selected characteristic.

6. The non-transitory computer readable medium of claim 5 , further comprising instructions that, when executed by the at least one processor, cause the computer system to determine the measure of group bias by:
identifying an excluded segment of users not selected for the proposed action from a user dataset; and
comparing a first probability that users within the segment of users correspond to the user-selected characteristic with a second probability that users within the excluded segment of users correspond to the user-selected characteristic.

7. The non-transitory computer readable medium of claim 5 , further comprising instructions that, when executed by the at least one processor, cause the computer system to determine the measure of individual bias by:
identifying a user dataset comprising the segment of users selected for the proposed action and an excluded segment of users not selected for the proposed action;
applying a classification algorithm to the user dataset to generate clusters of users from the user dataset based on characteristics corresponding to users within the user dataset, wherein each cluster comprises a set of users sharing at least one common characteristic; and
determining the measure of individual bias based on distributions of the segment of users and the excluded segment of users within the generated clusters of users.

8. The non-transitory computer readable medium of claim 5 , further comprising instructions that, when executed by the at least one processor, cause the computer system to generate the bias score representing the measure of group bias and the measure of individual bias by combining a weighted measure of group bias and a weighted measure of individual bias corresponding to the user-selected characteristic of users within the segment of users selected for the proposed action.

9. The non-transitory computer readable medium of claim 5 , further comprising instructions that, when executed by the at least one processor, cause the computer system to determine that a bias exists by comparing the bias score to a bias-score threshold.

10. The non-transitory computer readable medium of claim 5 , further comprising instructions that, when executed by the at least one processor, cause the computer system to modify the distribution of the users within the segment to reduce the bias score in relation to the user-selected characteristic by at least one of adding users that reduce bias in relation to the user-selected characteristic to the segment or removing users that increase bias in relation to the user-selected characteristic from the segment.

11. The non-transitory computer readable medium of claim 5 , further comprising instructions that, when executed by the at least one processor, cause the computer system to:
identify an excluded segment of users not selected for the proposed action from a user dataset;
provide, for display within the segment-generation-user interface together with the segment visualization, the bias-detection option, the bias indicator element, and the reduce-bias option, an excluded segment visualization for the excluded segment of users; and
based on receiving the indication of the selection of the bias-detection option, modify the excluded segment visualization to depict an additional distribution of users corresponding to the user-selected characteristic within the excluded segment of users.

12. The non-transitory computer readable medium of claim 5 , further comprising instructions that, when executed b

Assignees

• Adobe Inc

Inventors

• Sinha Moumita
• Srinivasan Shankar
• Sawant Pari

Classifications

• G06F3/0482

  Interaction with lists of selectable items, e.g. menus · CPC title

• G06F16/355

  Creation or modification of classes or clusters · CPC title

• G06F16/24578

  using ranking · CPC title

• G06F16/287 Primary

  Visualization; Browsing · CPC title