I'm not sure precisely how the scenario you're describing differs from current police investigative techniques, where you can become a suspect for, for instance, walking past the wrong streetcorner when a crime was committed and being picked up on camera.
... which is just a digital version of being misidentified by the local corner store owner when the cops go around and ask if they saw anyone suspicious in the neighborhood.
Its simple. Imagine a bomb is set off in an indoor mall. Lets say for the sake of argument that no legitimate evidence exists.
The bomb maker wore gloves, walked up, wore a mask, didn't bring his phone. Even so the more tech we deploy the greater the number of hits we are going to get.
GPS on everyone else's phones will tell you the name of hundreds to thousands of people that passed by within a plausible scenario, license plat readers can get you even more, facial recognition more yet! We can pull in even more if we look for people that looked up anything vaguely related to bombs and the mall within the last month.
Spurious info from eyewitnesses and traditional police work certainly happen but technology can trivially produce 100x the volume regardless of whether that data includes the guilty party or not.
Furthermore once you have your mark guilty or not our present system is designed to trivially secure the convictions of the innocent by punishing them 10 times over if they insist on the actual process they are owed. Here have 10 years 5 years with good behavior or how about a life sentence and in order to have competent representation I hope you have 20-50k to pay or we will shovel you someone who will themselves pressure you to take the deal because they don't have the time to deal with the caseload they have now.
Wrongful conviction is a fact of life not a hypothetical. Some estimates are as high as 10%
In that context normalizing tech that shovels loads of low quality evidence that governments can use to increase their clearance rates by intimidating anyone who might be good for the crime into copping would seem to have predictable results.
If the goal of this hypothetical is to show how technological methods are a societal risk, I would suggest couching the hypothetical in something other than "a bomber that is otherwise un-findable." Technology making it possible to find them through data correlation---even if the technology also has some false-positive risk---seems like an argument in favor of these methods.
It's probably also worth noting that these technologies can be used on the opposite end of the story too. "but officer, I can't possibly have been at the site of the bombing, here's my alibi data trail" is also enabled by these mechanisms.
The false positive rate is the whole point of hypothesizing a mad bomber with stringent data practices.
If you’re relying on circumstantial web searches for “bomb” or “fertilizer” or “SubUrban Mall map” that correlate with cell phone records of who was there on that day, you’re pulling in farmers, and people who wanted bug bombs to kill spiders, and people who wanted to know how to make a Jaeger Bomb, and people who wanted to know the fastest way to get to Clothing, Ltd. in that mall, etc. And you’re totally missing the nutjob who printed out “bomb-making-101.pdf” a year ago and left his phone at home (or in a friend’s car) when he struck.
In an area with 100k nearby people, which is a reasonable search metric since we have cars and malls live in populous areas, if the false positive rate is 1% you now have 1000 people as suspects.
What’s your success rate on guessing 1/1000? What tools do you think law enforcement has to further narrow this down?
> What tools do you think law enforcement has to further narrow this down?
Further cross-correlation of more data. Have the farmers ever even been to the same town as where the bomb was detonated? Were they there that day? Do they have any apparent motive? What's their alibi? You still have to prove a criminal case beyond a reasonable doubt, so all the points you're raising about alternative explanations for such searches are exactly the things the defense would raise, and why police don't use this data collection as evidence, but to focus the search from 100,000 people to 1,000 people.
It's probably worth thinking about it that way: absent this data, police have to assume everyone at the mall might be a suspect, so the pool is 100,000 potential bombers, not 1,000. What's the success rate on guessing 1/100000?
In a world of ever-increasing "data noise," that's true regardless of whether the cops are using the data.
A person who has a credit card for five years, even with a few missed payments, has better credit than a person who's never had a credit card. The latter is an unknown, and unknowns are risky.
A person who was let go from their last job last month has a stronger résumé than someone who's last job was five years ago.
A person who reads Hacker News for five years but never comments on anything lacks the downvote privileges that a person who comments heavily for a few months has.
... which is just a digital version of being misidentified by the local corner store owner when the cops go around and ask if they saw anyone suspicious in the neighborhood.