The discussion highlights the importance of documenting biases and lineage in datasets and models, yet questions the feasibility of widespread adoption. There's concern that requiring datasheets may discourage data publication altogether. Emphasizing the need for transparency, it is suggested that conferences should mandate the sharing of code and data to improve the replication process in research.
