Hacker News new | past | comments | ask | show | jobs | submit login

ShotSpotter has (used to?) manually changed the location of detected gunfire to placate police who needed the evidence to point to their primary suspect.

Then they could go to court and say “this system showed the gunshots came from the suspect’s location”. But “the system” didn't show that, until ShotSpotter employees manually changed the data.

They spent a lot of effort hiding that as well.




They don’t change the data. They change their interpretation of it. The police say the classifier got something wrong, so the classification gets updated, both for the record and to improve the classifier. The police say that the location is wrong. The company looks at the data, maybe finds that the system automatically located on an echo at one or more sensor and then recomputes the location using the primary pulse.


The “data” which gets presented in court is the classification. So they changed the “data”.

The underlying algorithm wasnt subject to examination by defense teams until a rules change like, last week.


What gets presented in court is a "Detailed Forensic Report" that includes the timestamped audio waveforms, sensor locations, and screenshots of the multilateration computations. These reports, when used as evidence, are required to be available to the defense for rebuttal and crossexamination. One could, if one was so inclined, calculate shot locations by hand from these reports. The time stamped audio files are discoverable.

The underlying algorithms have been made available to defense teams in countless Daubert and Frye hearings. I have no idea what you're referring to by "a rules change like, last week."


How are you privy to the underlying algorithms/source-code being made available to defense teams? I thought the source code (algorithm) doesn't have to be made available to defense teams.


Daubert requires that expert testimony be based on "reliable principles and methods" and Frye requires "the thing from which the deduction is made must be sufficiently established to have gained general acceptance," both essentially requiring the disclosure of underlying algorithms if not the actual source code.



A proposed law, not a rule change. One introduced in response to a novel DNA analysis technique being rejected in a Frye challenge under existing law.


Which is why we need public counter validation. (https://github.com/hcfman/sbts-aru) :-)




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: