Automated judging of criminals: Future of law P3
Automated judging of criminals: Future of law P3
There are thousands of cases around the world, annually, of judges handing out court verdicts that are questionable, to say the least. Even the best human judges can suffer from various forms of prejudice and bias, of oversights and errors from struggling to stay current with the rapidly evolving legal system, whereas the worst may be corrupted by bribes and other elaborate profit-seeking schemes.
Is there a way to sidestep these failings? To engineer a bias and corruption-free court system? In theory, at least, some feel that robot judges can make bias-free courts a reality. In fact, the idea of an automated judging system is beginning to be discussed seriously by innovators throughout the legal and tech worlds.
Robot judges are a part of the automation trend slowly seeping into almost every stage of our legal system. For example, let's take a quick look at policing.
Automated law enforcement
We cover automated policing more thoroughly in our Future of Policing series, but for this chapter, we thought it would be helpful to sample a few of the emerging technologies set to make automated law enforcement possible over the next two decades:
Citywide video surveillance. This technology is already used widely in cities around the world, especially in the UK. Moreover, the falling costs of high definition video cameras that are durable, discrete, weather resistant and web-enabled, mean that the prevalence of surveillance cameras on our streets and in public and private buildings is only going to grow over time. New tech standards and bylaws will also emerge that will allow police agencies to more easily access camera footage taken on private property.
Advanced facial recognition. A complementary technology to citywide CCTV cameras is the advanced facial recognition software currently being developed around the world, especially in the US, Russia, and China. This tech will soon allow real-time identification of individuals captured on cameras—a feature that will simplify the resolution of missing persons, fugitive, and suspect tracking initiatives.
Artificial intelligence (AI) and big data. Tying these two technologies together is AI powered by big data. In this case, big data will be the growing quantity of live CCTV footage, coupled with the facial recognition software that's constantly paring faces to those found on said CCTV footage.
Here the AI will add value by analyzing the footage, spotting suspicious behavior or identifying known troublemakers, and then automatically assigning police officers to the area to investigate further. Eventually, this tech will autonomously track a suspect from one side of town to another, collecting video evidence of their behavior without said suspect having any clue that they were being watched or followed.
Police drones. Augmenting all of these innovations will be the drone. Consider this: The police AI mentioned above can employ a swarm of drones to take aerial footage of suspected criminal activity hot spots. The police AI can then use these drones to track suspects across town and, in emergency situations when a human police officer is too far off, these drones can then be used to chase and subdue suspects before they cause any property damage or serious physical injury. In this latter case, the drones would be armed with tasers and other non-lethal weapons—a feature already being experimented with. And if you include self-driving police cars into the mix to pick up the perp, then these drones can potentially complete an entire arrest without a single human police officer involved.
The individual elements of the automated policing system described above already exist; all that remains is the application of advanced AI systems to bring it all together into a crime-stopping juggernaut. But if this level of automation is possible with on-the-streets law enforcement, can it also be applied to the courts? To our sentencing system?
Algorithms replace judges to convict criminals
As mentioned earlier, human judges are susceptible to a variety of very human failings that can taint the quality of verdicts they pass down on any given day. And it's this susceptibility that's slowing making the idea of a robot judging legal cases less farfetched than it used to be. Moreover, the technology that could make an automated judge possible isn't that far away either. An early prototype would require the following:
Voice recognition and translation: If you own a smartphone, then by now you probably have already tried using a personal assistant service like Google Now and Siri. When using these services, you should have also noticed that with each passing year these services are getting much better at understanding your commands, even with a thick accent or amid a loud background. Meanwhile, services like the Skype Translator are offering a real-time translation that's also getting better year to year.
By 2020, most experts predict these technologies will be near perfect, and in a court setting, an automated judge will use this tech to collect the verbal courtroom proceedings needed to try the case.
Artificial intelligence. Similar to the point above, if you've used a personal assistant service like Google Now and Siri, then you should have noticed that with each passing year these services are getting much better at offering correct or useful answers to the queries you ask of them. This is because the artificial intelligence systems powering these services are advancing at a lightning pace.
As mentioned in chapter one of this series, we profiled Microsoft’s Ross AI system that was designed to become a digital legal expert. As Microsoft explains it, lawyers can now ask Ross questions in plain English and then Ross will proceed to comb through "the entire body of law and return a cited answer and topical readings from legislation, case law, and secondary sources."
An AI system of this caliber is no more than a decade away from developing above a mere legal assistant into a reliable arbiter of law, into a judge. (Going forward, we'll use the term ‘AI judge' in place of ‘automated judge.')
Digitally codified legal system. The existing base of law, currently written for human eyes and minds, needs to be reformatted into a structured, machine-readable (queryable) format. This will allow AI lawyers and judges to effectively access relevant case files and court testimony, then process it all through a kind of checklist or scoring system (gross oversimplification) that will allow it to decide on a fair judgment/sentence.
While this reformatting project is currently underway, this is a process that currently can only be done by hand and could, therefore, take years to complete for each legal jurisdiction. On a positive note, as these AI systems become more widely adopted across the legal profession, it will spur the creation of a standardized method of documenting law that's both human and machine readable, similar to how companies today write their web data to be readable by Google search engines.
Given the reality that these three technologies and digital libraries will fully mature for legal use within the next five to 10 years, the question now becomes how will AI judges truly be used by the courts, if at all?
Real world applications of AI judges
Even when Silicon Valley perfects the technology behind AI judges, it will be decades before we see one independently try and sentence someone in a court of law for a variety of reasons:
- First, there will be obvious pushback from established judges with well-connected political affiliations.
- There will be pushback from the wider legal community who will campaign that AI tech isn't advanced enough to try real cases. (Even if this weren't the case, most lawyers would prefer courtrooms managed by a human judge, as they have a better chance at persuading the innate prejudices and biases of said human judge as opposed to an unfeeling algorithm.)
- Religious leaders, and a few human rights groups, will argue that it's not moral for a machine to decide the fate of a human.
- Future sci-fi television shows and movies will begin featuring AI judges in a negative light, continuing the killer robot vs. man cultural trope that has frightened fiction consumers for decades.
Given all these roadblocks, the most likely near-term scenario for AI judges will be to use them as an aid to human judges. In a future court case (mid-2020s), a human judge will manage the courtroom proceedings and listen to both sides to determine innocence or guilt. Meanwhile, the AI judge will monitor the same case, review all the case files and listen to all the testimony, and then digitally present the human judge with:
- A list of key follow-up questions to ask during the trial;
- An analysis of the evidence provided in advance of and during the court proceedings;
- An analysis of the holes in both the defense and prosecution’s presentation;
- Key discrepancies in the witness and defendant testimonies; and
- A list of biases the judge is predisposed to when trying a specific type of case.
These are the kinds of real-time, analytical, supportive insights that most judges would welcome during their management of a case. And over time, as more and more judges use and become dependent on the insights of these AI judges, the idea of AI judges independently trying cases will become more accepted.
By the late-2040s to mid-2050s, we could see AI judges trying simple court cases such as traffic infractions (the few that will still exist by then thanks to self-driving cars), public intoxication, theft, and violent crime—cases with a very clear-cut, black and white evidence and sentencing. And around that time, should scientists perfect the mind reading tech described in the previous chapter, then these AI judges may also be applied to far more complex cases involving business disputes and family law.
Overall, our court system will see more change over the next few decades than it's seen in the last few centuries. But this train doesn't end at the courts. How we jail and rehabilitate criminals will experience similar levels of change and that's exactly what we'll explore further in the next chapter of this Future of Law series.
Future of law series
Next scheduled update for this forecast
The following popular and institutional links were referenced for this forecast:
The following Quantumrun links were referenced for this forecast: