Monday, July 02, 2018

Artificial Intelligence

In March 2018, the Justice and Security Commission of the Netherlands Parliament Second Chamber held a round table on artificial intelligence in the justice domain. Here is the summary of my speaking notes, on AI and courts. The original was, of course, in Dutch. This English version I wrote for the session on Artificial Intelligence by the Commission for the Efficiency of Justice (CEPEJ) session on June 27 2018 in Strasbourg.

AI for courts, in brief (this is the summary)
What use can artificial intelligence (AI) have for courts, and what does that take? In court cases, judges reduce complexity, but all court work is by no means complex, bespoke work. Courts do not process all cases in the same way, and consequently, they need information technology suited to the different ways. Therefore, AI can be useful for different types of courts cases in different ways. Some forms of AI have already proven themselves in practice. But will robots replace judges, as some people have been claiming for more than twenty years? There is still no evidence to support it. Article 6 of the European Convention on Human Rights prescribes fair procedure. It will take a lot of work to make AI conform to that standard. Legal information needs to be structured and given meaning, in order to make the information not only readable, but also actionable, including decision making. Providing an explanation is, in the foreseeable future, not yet feasible for AI. AI can help people looking for information, parties in a case and judges with structuring information, and if legal information is enriched, also with advice and suggestions.  read the full note here

Monday, March 19, 2018

Court IT: we must, we can, but it’s not easy

A while ago, a judicial colleague published a column in one of the Dutch online papers. My colleague pointed to rising cost and longer duration of the digitalization of the courts. The digitalization was done in the Quality and Innovation Program. In Dutch the program is called Kwaliteit en Innovatie (KEI). The courts are, apparently, held ransom by the vendors. I reacted to his column in my monthly blog for one of the Dutch legal magazines. My blog went viral. Reason enough to give it some wider circulation by including it in here, in my Technology for Justice blog. For those of you who read Dutch: here is the original.

My question: if digitalizing turns out to be more difficult than expected, should we give it up?

Dear colleague,
Should I react to your column entitled “IT-vendors hold judiciary in a tight grip”? I wondered if I should.  Then, I read an earlier interview in which you wanted to abolish the redrawing of the judicial map, as well as the Council for the Judiciary. Fortunately, we judges enjoy freedom of speech. I am not going to debate the merits of your viewpoints. I had already decided not to publish this blog. But then, my IT team asked to please publish after all. The negative publicity made them very uncomfortable. So, therefore, I use this opportunity to list a few facts.

The parliamentary hearings on government IT-projects have made clear that an IT project can be considered a failure when the software that was developed is not implemented. Delay of rising costs are, in themselves, not a reason to consider a project a failure.
The first KEI-systems went into operation in 2015. Since April 2015, the system handled more than 20.000 asylum cases. Users in the courts are very happy with the system. Hearing planning is now a matter of hours instead of weeks. Since November 2015, bankruptcy supervisors can communicate with the courts in all types of bankruptcies. Well over 80% of bankruptcies are now handled this way. Professional supervisors have started to communicate digitally with all the courts in November 2017. Since September 2016, more than 650 (over 900 by mid-March, DR) commercial claims were filed with the new system, in two pilot courts. A few during a voluntary test phase, but most in the current compulsory filing pilot phase. Some cases have already been completed. One case, with a full hearing, was concluded within 7 weeks from filing. Users are not entirely happy with the system yet, and therefore the system has not been rolled out to all courts. That will happen later in 2018. So, the KEI software has been in use for a couple of years. 

The Council of State, in its 2014 advisory on the digitalization legislation, pointed out the risk of innovating procedure and digitalizing in a single operation. The Council also pointed out that earlier suggestions to drop the distinction between claims and requests and handle all cases on the basis of a request, were not followed, and that this complicates the digitalization.  De legislator did not follow this advisory. This has made designing the civil digital procedure more complex, hence more expensive than expected. There were a few other things that could not be foreseen in 2014. For example, article 113 of the Code of Civil Procedure that was added in a later phase of the legislative process. This article upended the design that we had already made and built, and sent us back to the drawing board. As a consequence,  we could only give lawyers access to the new systems, and no one else.
Digitalization needs to be kept simple. That is not a new insight. The Netherlands Accounting Chamber investigated government IT projects at the request of Parliament. The Chamber reported extensively in 2007. The dynamic of politics makes government IT-projects complex. Look at the Tax Office. That took a long time, but by now we all file our taxes on line.
Suppose: we know we have to keep things simple, but from the Accounting Chamber’s reports it is clear that complications will arise anyway. What do we do? Forget about digitalization? I cannot speak for the legislator or the Council for the Judiciary, but I can speak for myself, as the product owner of the civil procedure. I felt we should go ahead, and seize this opportunity that might not arise again for the next ten years. With every decision, we tried to apply the simplicity rule: can we make it simpler, is it necessary at all? It turned out we could not always do that. Legislation, the environment of bailiffs and lawyers, requirements and wishes from the users in the courts, technology itself, security needs, and a host of other factors kept us from applying the simplicity rule consistently. But some procedures have been digitalized, and they work. Civil procedural legislation changed so much that the courts need to put a lot of effort into implementation, and that is another reason why the new system has not been rolled out to all courts yet.

Dear colleague, we judges are the guardians of the existing legal order. Our work is looking back and deciding who need to get the blame for what went wrong. There is nothing wrong with that. But looking ahead and envisioning how to innovate does not come naturally to us. So, court innovation is difficult. If we are afraid to be blamed if something goes wrong, we will avoid innovation. Blaming in retrospect will not help innovation. But it must be done, because we are losing our market share to digital alternatives, and we do not do enough to give ordinary people access to justice.