As stated in the Cover Note, we need to add examples of harms actually or hypothetically caused, particularly in cases of outliers, by AI tools. These examples are intended to make the discussion of algorithmic discrimination more concrete.
As discussed with @claytonhalllewis, we may need to proceed in two phases:
-
Include whatever high-quality examples are available, if any, prior to the review period. Revise the Cover Note as a request for comments from reviewers, that retains the suggestion to provide examples of harms caused by AI systems.
-
Write the missing section in response to comments received and independently conducted research.
If we are fortunate enough to receive a good set of examples prior to the review period, we can avoid the two-phase approach, although it may still be useful as a means of soliciting additional (or better) examples.
As stated in the Cover Note, we need to add examples of harms actually or hypothetically caused, particularly in cases of outliers, by AI tools. These examples are intended to make the discussion of algorithmic discrimination more concrete.
As discussed with @claytonhalllewis, we may need to proceed in two phases:
Include whatever high-quality examples are available, if any, prior to the review period. Revise the Cover Note as a request for comments from reviewers, that retains the suggestion to provide examples of harms caused by AI systems.
Write the missing section in response to comments received and independently conducted research.
If we are fortunate enough to receive a good set of examples prior to the review period, we can avoid the two-phase approach, although it may still be useful as a means of soliciting additional (or better) examples.