The New Age of Crime-Solving: AI Steps into the Spotlight
The New Age of Crime-Solving: AI Steps into the Spotlight
It's totally understandable if you have no idea what to make of this.
The Mystery Machine enters the picture
In a groundbreaking—and for some, perhaps troubling—move, NJ.com reports that the Middle Township Police Department in Cape May County, New Jersey, has turned to artificial intelligence (AI) to help solve the mysterious disappearance of Mark Himebaugh. This unsolved case is three decades cold. The police department has partnered with Tabtu Corp and its service provider Terawe, utilizing their Vollee Artificial Intelligence to analyze the extensive data accumulated during the investigation. The Vollee AI is reportedly built on Microsoft's Azure cloud service. The company behind it promises advanced computational and processing capabilities. The AI can supposedly spot patterns and connections within the massive amount of data collected over the years.
Mark Himebaugh, an 11-year-old with red hair and blue eyes, vanished from his house in Del Haven, Middle Township, on November 25, 1991. He was last seen watching firefighters battle a marsh blaze in a nearby park. Over the years, the case has seen a few leads, but none have come to fruition. The only physical trace of Himebaugh ever discovered was his left sneaker, found not far from where he was last seen. The case, described by investigators as a "non-family abduction," has confounded local authorities and the FBI. At least one viable suspect showed a witness what appeared to be footage of him assaulting a boy who looked like Himebaugh, but investigators could never nail him for the crime.
This partnership with Tabtu Corp and Terawe reportedly marks a significant turning point in the investigation. AI tools like facial recognition technology could uncover new leads and hopefully solve this long-standing, terrible mystery. Currently, the police department is in the process of uploading all of its data into the Vollee AI system. The FBI, State Police, and the National Center for Missing & Exploited Children are expected to do the same. It's a hope-filled moment for the Himebaugh family and the local community, who have spent decades waiting for answers.
ViCAP
Far be it for me to throw cold water on the endeavor. To be clear, I support anything that might conclusively solve a cold case, including artificial intelligence. And there’s no doubt that AI technology could be uniquely geared toward finding meaningful patterns that even seasoned human investigators might miss.
Heck, the dawn of behavioral analysis in criminology incorporated computer pattern-matching abilities when the FBI began building ViCAP in 1985—a nationwide data information center designed to track and correlate information on violent crimes, especially murder. From the beginning, the system collected and analyzed data on homicides, sexual assaults, missing persons, and unidentified human remains. ViCAP’s goal is to find patterns and link crimes that could be related, assisting local, state, and federal investigators in efficiently solving cases.
A simple analogy explains ViCAP: Imagine you're working a jigsaw puzzle, but you only have a few pieces. The pieces are crimes. The completed puzzle is the full pattern of criminal activity.
One of AI’s fundamental strengths is pattern-matching. It’s at the core of how it works. AI systems are designed to learn from data, identify patterns, and make decisions or predictions based on those patterns. This capability is particularly evident in machine learning, a subset of AI, where algorithms are used to model and understand complex structures and phenomena. A match made in crime-solving heaven, right? (I hope there isn’t a crime-solving heaven; that’s just weird.) Artificial Intelligence could make for a super-charged ViCAP.
Obviously, though, I wouldn’t have written all this if I didn’t have some doubts. Or, to be precise, reservations.
Reservations
I was struck by the extremely low Google profile of the company the Middle Township Police Department in Cape May is working with on the Himebaugh case. Vollee.ai has a great-looking website. Very professional and slickly designed. Sure, it’s fleshed out with stock images, but so are half the sites you look at daily.
Terawe.com is much more informative and clearly well-established. It says the company focuses on working in various industries, including education, finance, healthcare, and—perhaps the most relevant industry regarding crime-solving—government. My guess is that the Himebaugh case may be a big experiment, a kind of trial balloon for an AI tool that, if successful, will be made available to law enforcement worldwide.
It’s fair to note here that Parabon NanoLabs, the company behind groundbreaking (but in my opinion, potentially flawed) DNA phenotyping technology, was very secretive when it began offering its services to law enforcement, rarely responding to press inquiries in any detail. This is probably just smart from a corporate perspective; people will steal such ideas.
About Parabon
Speaking of Parabon—as someone who has covered crime for years and cares about those left behind by these unsolved crimes, I remember being excited about its potential and immediately wary. That’s how I feel about AI’s entry into crime-solving.
My reservations about Parabon were simple. The company touts its profiles that appeared to nail an unknown suspect appearance and ancestry-wise. At this point, it has been around long enough to notch some real successes. But I couldn’t shake what I know personally about the unreliability of DNA’s ability to predict your appearance. 23andMe attempts to predict traits, too, including what you look like. According to that service, I should have light brown hair and no dimpled chin. In fact, 23andMe’s analysis said I was least likely to be a redhead.
I’m a redhead with a dimpled chin. That reads like nitpicking, but if you consider that Parabon made a big splash by presenting what could be portraits of unknown suspects and John and Jane Does to find new leads, the difference between a brown-haired guy with a regular square chin and my appearance is substantial. Assuming 23andMe and Parabon were working with the same data and getting similar results, a mockup using my DNA would be missing two major identifying traits.
Dept. of Precrime?
The problems with AI are many—too many to lay out here. I’m not afraid of it and experiment with AI tools regularly, but I’m too cynical to view it as some magic bullet for all humanity’s problems. Discussion of what should and should not be done with it won’t end any time soon. One thing far too many people seem to ignore in pivoting to using it for content or creation is how wrong it can be. In writing this, I went to Google’s Bard and asked it, “In what ways could AI help solve crime, especially cold cases?”
Bard provided a typically well-formatted and detailed answer. Through analyzing mountains of data, AI can identify patterns and trends. It could potentially link cases previously considered unrelated (the reason ViCAP was created). Bard also stated that AI could “be used to predict where and when crimes are likely to occur” and that this “information can then be used to deploy law enforcement resources more effectively and prevent crimes from happening in the first place.” Which sounds uncomfortably like precrime forecasting, as conceived by Philip K. Dick in the story that inspired Minority Report. Bard also said, “AI can be used to re-examine cold cases that have gone unsolved for years. By applying new technologies and techniques, AI can help law enforcement identify new leads and finally bring justice to the victims.”
Its answer concluded, in part:
…AI has been used to help solve a number of high-profile cold cases, including the Golden State Killer and the Zodiac Killer. In both cases, AI was used to analyze DNA evidence that had previously been inconclusive. By using AI to re-examine the evidence, law enforcement was able to identify new suspects and finally bring the cases to a close.
If you know your recent true crime history, you immediately see the problem here. Artificial intelligence didn’t solve the Golden State Killer case. Joseph James DeAngelo was the first major cold case solved by forensic genealogy.
And despite troubled people regularly coming out of the woodwork to claim their dad/stepdad/fairy godmother was the Zodiac Killer, that case remains unsolved unless Bard’s AI is truly some monolithic intelligence and knows something I don’t. One of the killer’s famously difficult ciphers was solved, and AI could have played a role there, but that’s all I know about.
I corrected Bard regarding the Zodiac and received this response: “You are correct. The Zodiac Killer was not solved by AI. The ciphers sent to the press were solved by a team of codebreakers, including Kevin Knight, who used a supercomputer called CARMEL. However, the Zodiac Killer's identity remains a mystery.”
Okay, so…
I believe AI will play some role in crime-solving. Perhaps a huge role. It may even solve cases. I do, however, fear the idea that any law enforcement entity might rely on it too soon. Just as digital publications pivoting to AI is the height of “business over people” idiocy, there could be a huge legal and criminal justice minefield here that no one involved wants to acknowledge or knows enough to see—one that could take courts decades to figure out.
Mark Himebaugh was likely the victim of a homicide—the kind of stranger-snatching parents used to terrify kids into coming home before dark when I was young. His killer should be brought to justice as soon as possible. Let’s hope that if that person is alive and can be brought into court that he doesn’t have some immediate grounds for appeal due to any role artificial intelligence could play in finding him.
However the case is solved the Himebaugh family deserves to see it done the right way.