Whenever we read about a fictional detective like Sherlock Homles and Hercule Poirot, we are often amazed by their ability to remember and put to use the sheer amount of data and observations they’ve made over the years. Abilities like that might not be fictional anymore, although it won’t really be a human brain doing the hard work. UK Police is trying out a system where police data is being used by a computer to analyse and propose how a particular crime was conducted.
The system, called VALCRI, is intended to replace a crime analyst so that the tedious part of their job is done easily and in a short time. This will allow the analysts to focus on the more important part of their job, which is the case itself. Additionally, this system would ensure that a possibility or a theory hadn’t been missed by human analysts by exploring all possible outcomes.
How does it work?
So the main objective with VALCRI is to generate plausible theories and possibilities based on the data available about the crime, as well as the millions of police records available – which include previous investigations, interviews, evidence files etc. This entire analysis is then presented to the analyst in the form of two screens, all thanks to artificial intelligence.
At a basic level, this involves a lot of pattern recognition and then learning. So, for instance, if the AI notices a particular footprint of a specific size and shoe pattern on multiple scenes, it can bring it to the attention of the analyst, who can then inform the system about its significance. This way, the system keeps learning about the importance of several criteria when it comes to evidence and observations, along with contextual awareness. What would take multiple manual search queries, VALCRI would be able to do with a single action.
The police department in West Midlands is trialling the system with 3 years worth of anonymised data. Other sections of the police in Europe are also undertaking similar actions. The next step would be to feed it data in real time and observe the results.
A crime is never a simple thing to deal with. A case itself might be simple and offer a straightforward explanation, but for an ongoing case, evidence and testimony are often challenged and proved invalid in court. So feeding data to VALCRI too soon might eventually reduce the accuracy of the system.
Also, we, as a society, aren’t exactly comfortable with the idea of an AI making the final call on something like a death sentence or a juvenile offence. Also, data in a case is not always conclusive and human instinct is required more often than not. On the other hand, even though VALCRI is being developed to cover all avenues, it might miss something that a human analyst would have noticed otherwise – and the amount of trust placed on VALCRI might end up being a bias on the analyst’s part.
VALCRI deals with these problems currently by making the entire process available for scrutiny and backtracking. The end result is bound to be a much more detailed case being presented at a court hearing.
AI is inevitably going to take over many jobs and roles whether we like it or not. A way to look at it is that AI is enhancing the performance of certain activities, leaving humans free to pursue more detailed, time-consuming activities. On the other hand, something like VALCRI will leave fewer excuses for human detectives and analysts to not fulfil their responsibilities. Overall, Sherlock might not approve, but we know a certain Watson that would.
Source: New Scientist