> Precrime only works if they can separate signal from noise.

IMO a lot of these debates depend on implicit assumptions about the threat and how it operates. For example:

1. Lawful Evil: They care about good data and going after the "worst" offenders, even if I might disagree about what is "bad".

2. Lazy risk-averse evil: The data needs to give them something to justify the existence of the program, they'll go after whomever is convenient.

3. Cover-your-ass evil: The data archive exists to let them make a plausible case for someone they've already decided to persecute for other reasons.

4. Fraudulent evil: The data archive is just to make it easier to fabricate a fake reason to go after someone.

5. Blatant evil: The data doesn't matter because they can just do stuff to you by fiat.

Some of those groups would be hampered by noise, some would benefit from noise, and the last just won't care.