4 ms·

It depends on what you consider anti-conservative. I've had a few friends mention that any links to something regarding "Benfords Law" (some forensic tool used

It depends on what you consider anti-conservative. I've had a few friends mention that any links to something regarding "Benfords Law" (some forensic tool used for flagging potential fraud) gets an immediate "disputed" label. My guess is that algorithmic checks are flagging posts and conservatives are interpreting that as censorship.

rodw 6y ago
Asserting that Benford's Law can be used to identify election fraud is a disputed, or just downright misleading, claim. Academic research has shown Benford's Law is not applicable to election tallies. (Also so does common sense when you think about it.)
- katzgrau 6y ago
  No one said identify fraud, I said flag potential fraud. There are some people who think it's a useful heuristic for uncovering potential oddities. Academics may not think so, but perhaps people who are professional analysts do. Ultimately it comes down to who's side you want to take, and whether that's going to result in something akin to censorship. Also, part of the reason I don't like commenting on any remotely political HN posts is that you get downvotes for saying perfectly reasonable things.
BearsAreCool 6y ago
Sad to see downvotes here but the Benford's law thing is absolutely one of the weakest claims of fraud finding. It sounds good initially (tool for finding some kinds of fraud used to find other kinds of fraud) but it is almost entirely flawed with how it is used in this context. It truly is the best current example of statistics lying. If you've seen the handy little graph of the first digit statistics that each candidate got in big cities, I dug into this a bit but the basic summary is precinct sizes aren't random at all and that messes up the entire premise of that analysis. I made a quick repo that altered the initial repo to show things like how just getting 70% of the vote would screw up benfords law in most counties [0]. [0] https://github.com/BeesAreCool/2020_benfords https://github.com/BeesAreCool/2020_benfords

rodw 6y ago

Asserting that Benford's Law can be used to identify election fraud is a disputed, or just downright misleading, claim. Academic research has shown Benford's Law is not applicable to election tallies. (Also so does common sense when you think about it.)

katzgrau 6y ago
No one said identify fraud, I said flag potential fraud. There are some people who think it's a useful heuristic for uncovering potential oddities. Academics may not think so, but perhaps people who are professional analysts do. Ultimately it comes down to who's side you want to take, and whether that's going to result in something akin to censorship. Also, part of the reason I don't like commenting on any remotely political HN posts is that you get downvotes for saying perfectly reasonable things.

BearsAreCool 6y ago

Sad to see downvotes here but the Benford's law thing is absolutely one of the weakest claims of fraud finding. It sounds good initially (tool for finding some kinds of fraud used to find other kinds of fraud) but it is almost entirely flawed with how it is used in this context. It truly is the best current example of statistics lying. If you've seen the handy little graph of the first digit statistics that each candidate got in big cities, I dug into this a bit but the basic summary is precinct sizes aren't random at all and that messes up the entire premise of that analysis. I made a quick repo that altered the initial repo to show things like how just getting 70% of the vote would screw up benfords law in most counties [0]. [0] https://github.com/BeesAreCool/2020_benfords https://github.com/BeesAreCool/2020_benfords