The Group graphics are a stylized folder of a randomized for color, with an outline of a person that suggests male.
My request is to generate some additional images so there's an even split between images that seem to be male, seem to be female, and those that are gender-ambiguous.
I'm not sure how this would work on the back-end, but one way might be this:
To ensure that the de-biased images populate to existing groups, you could replace existing images, taking the colors that exist now, dividing by three, and updating 2/3 of the total to gender diverse outlines. You would need to randomize the assignments in order to break cultural gender+color associations.
In any case, it would be desirable for this to be retroactive, or the bias will be retained.
