One private firm’s artificial-intelligence system is deemed insufficient evidence
ShotSpotter, a gunshot detection firm contracted by police departments nationwide, has recently received criticism for its audio forensics system that, it claims, incorporates “sensors, algorithms, and AI” to identify gunshots and locate their source. While several precincts have praised the company for increasing police response to incidents of gun violence, its accuracy as evidence in court remains questionable.
There are two primary reasons for skepticism: 1) studies have indicated that its algorithm has a propensity for generating false positives, and 2) employees are able to modify the database after alerts come in. Since its system is protected as a trade secret, it has been generally inscrutable from oversight.
As seen in this Associated Press investigation, a State’s Attorney’s Office used ShotSpotter’s data for evidence in a case against a Chicago man. This left him in prison for 11 months before the judge dismissed the case. The report eventually released by ShotSpotter showed that the alert in question was identified differently at first. It alerted to a “firecracker” several blocks away from the alleged scene of the crime — but an employee later revised the identification and location. As a result, prosecutors decided that the “evidence was insufficient to meet [their] burden of proof.”
How could it be improved?
This case emphasizes the importance of accountability in regards to digital evidence on either side of a case. The Health Information Portability and Accountability Act (HIPAA), for example, requires retention of Electronic Medical Records (EMR) stored in Health Information Systems (HIS). Healthcare firms must record a permanent record of all additions, changes and deletions of EMR, including the time and person making those changes.
While ShotSpotter obviously isn’t in healthcare, its system would still benefit from similar transparency. It would help improve the reliability of such information. In this case, such logs would have revealed human intervention earlier on. This would have saved the defendant from the 11 he spent months in prison. In other cases, transparency could support prosecution. Regardless, it would bolster ShotSpotter’s credibility when used as evidence.
It’s possible that we could examine information recorded — when the stored data was originally entered and changes to that stored data — without violating trade secret status to a software provider’s algorithms. HIS software providers have trade secret protection to their software. Still, they are required to disclose all record EMR, as well as the revision history to those records.
Where we can help.
Asking the right questions and gathering all available digital evidence is important to achieving an equitable outcome. Enigma Forensics has experience auditing and authenticating digitally stored electronic evidence. We can assist with validating such claims as genuine.