Repairing the harm: Toward an algorithmic reparations approach to hate speech content moderation
Previous
Previous
Elephant motorbikes and too many neckties: epistemic spatialization as a framework for investigating patterns of bias in convolutional neural networks
Next
Next