But this would require speculation...

When the Future of Computing Academy recently suggested that authors of computer science papers should write about the possible negative implications of the technology they build or the research they conduct, one complaint I saw from scientists was that this would require speculation: How are we supposed to know what bad things might happen?