Automated case assignment in courts: Is everything okay?
To prevent society from losing faith in judicial reform (trust in courts is the lowest after political parties), an automated distribution of cases among judges has been proposed. This is a serious innovation. But! In developed democracies, everything can be distributed manually because Europeans trust their courts. And even with such innovations, questions arise in our country.
Those eager to “solve” cases in courts haven’t disappeared. It’s worth recalling the “Schemes” investigation into the mysterious illness of 29 judges of the Kyiv District Administrative Court, which left only one candidate for selection by the computer, or analyzing the DEJURE schemes involving the submission of many lawsuits or manipulation with the specialization of judges.
And although such scandals have not been written about lately, it should be understood: the deeper and more effective the judicial reform, the more eager there will be to exploit the loopholes in legislation regarding the case distribution algorithm.
And they exist.
One of the most high-profile cases recently is the decision by the High Anti-Corruption Court to impose preventive measures on former MP Serhiy Pashinsky, who is suspected of illegally misappropriating nearly a billion hryvnias worth of oil products.
Sergiy Pashinsky (Photo: via Sternenko/Telegram)
Pashinsky was not the first to be tried in this case, so the decision on the preventive measure, by law, was handed over to the previously elected composition in the High Anti-Corruption Court and in the Court of Appeals.
Therefore, it is necessary to analyze how the initial automatic distribution occurred overall in Pashinsky’s case when the investigative judge Natalia Movchan was chosen in 2022, and then how in the Court of Appeals of the High Anti-Corruption Court, a panel with reporting judge Mykola Hlotov was selected for the consideration of the appeal of Pashinsky’s business partner Serhiy Tyshchenko.
Below in Fig. 1, graphs are presented, constructed based on the respective protocols, reflecting the probability of selecting judges during the automatic distribution depending on their workload coefficient.
Figure 1
In both cases, the probability of randomly selecting a judge with the lowest workload is determined by a principle distinct from other judges and amounts to 50%. In both cases, this particular judge was selected.
When the suspect is an opposition figure, the case will always have political undertones. There will always be those willing to believe in political persecution.
However, these graphs do not in any way prove the existence of abuses in Pashinsky’s case. Such an algorithm is built into the automated document management system used by this court for case allocation.
Since the establishment of the High Anti-Corruption Court and its Appellate Chamber, there has always been a 50% probability of selecting a judge with the lowest workload. Therefore, based solely on probability theory, in such a number of cases, the selection of such a judge occurred, and the coincidence in both instances, as happened in Pashinsky’s case, would occur in every fourth case.
The court’s apparatus already knows which judge will have a 50% probability of selection when allocating the next case. When the desired judge is needed, the responsible person within the apparatus can easily coordinate the submission of the case through the Electronic Court or enter documents previously submitted in paper form into the system at the right moment. Court leadership may also be tempted to influence this process.
Simpler options exist as well. For example, cases are not assigned to judges 14 days before their vacation if the duration of the vacation is at least two weeks. However, the High Anti-Corruption Court has established that these days are considered in the workload calculation (section 3 of section II of the Principles of using an automated document management system). As a result, during the first few days after returning from vacation, a judge will have a constant 50% probability of selection during case allocation, until they catch up with their colleagues in workload. Anyone can take advantage of this, and the court may not even be aware of attempts at manipulation by applicants.
However, the “possibility of abuse” does not necessarily mean the “existence of abuse.”
The specificity of the described scheme lies in the absence of traces because there is no interference with the operation of the automated system, and based on the results of allocation, it is theoretically impossible to determine whether there was a leak of insider information and artificial selection of a judge or not because all allocations look equally suspicious.
For example, if Pashinsky accuses the authorities of manipulation in selecting a judge to hear his case, he will not be able to prove anything, even if something like this did occur. Similarly, the authorities will never be able to prove the falsity of the accusations, thereby appearing guilty to many.
The situation becomes even more interesting when comparing different courts.
On figure 2, the graphs are constructed based on randomly selected distribution protocol data in the scandalous District Administrative Court of Kyiv for a day (December 12, 2022) before the law on its liquidation was passed and the protocol on the same day in the Kyiv District Administrative Court, which has been tasked with handling the work of the District Administrative Court of Kyiv until now.
Figure 2
Absurd. However, using the example of the most corrupt court, one can demonstrate what an honest probability determination algorithm should look like.
There were no manipulations at this stage in the scandalous District Administrative Court of Kyiv. There was no need for this – towards the end of its work, the choice of judge did not affect the possibility of obtaining any ordered decision.
In the Kyiv District Administrative Court, which is not hopeless in terms of justice, the choice of judge influences many things, so the same dubious algorithm is used there as in the Higher Anti-Corruption Court.
Although Ukraine is a unitary state according to the Constitution, there is no unity in approaches in the same courts in different regions. For example, out of six appellate commercial courts, the probability of choosing a judge with the lowest workload is set at 30% in the Northern (Kyiv) and Northwestern (Rivne) regions, while in the other four it is 50%.
Out of seven working appellate administrative courts, a normal distribution is observed in the Second (Kharkiv) and Sixth (Kyiv) regions, automatic 30% probability for the judge with the lowest workload in the First (Kramatorsk) region, and 50% in the other four.
At first glance, a 50% probability for an individual judge may not seem like much.
For example, it significantly increases the efficiency of schemes described by DEJURE. If identical claims could be filed against a judge 5-10 times, the same reliability is achieved with just one or two claims under the new scheme. This makes it less likely to attract attention.
In appellate and Supreme Courts, a panel is selected, not an individual judge, so the probability of selecting someone from the panel with this distribution is already 55-65%, depending on the number of permanent panels with the relevant specialization.
The impact of the Public Integrity Council’s work can be effectively halved, as influential individuals will have half the probability of selecting a judge they consider undesirable. The more successful judicial reform is, the more significant this scheme will become. In a court with several dozen honest judges, only one dishonest one may remain, but there will still be a 50% chance of selecting that one.
The most interesting part is this: it mysteriously originated somewhere within the judicial system, with blank spots in legislation. There are no legal requirements for the principle of determining the probability of selecting a particular judge.
The current Regulation on the automated court document management system simply states that “the selection of a judge by a random number is carried out according to the judge’s workload coefficient at the time of automated case distribution,” with no clarification on how this random number should be determined.
In the absence of any regulatory acts, centralized programs developed under the guidance of the State Judicial Administration took an algorithm and embedded it, allowing each court to set in its parameters the probability they desire for selecting a judge with the lowest workload.
Sometimes, the assemblies of individual judges endorse manipulative norms in their own Regulations, such as: “In the field ‘Probability of distribution… to a judge with the least workload,’ the percentage probability of distributing a case to a judge with the largest interval is set at 50%.” However, as in the case of VAKS, the assemblies of judges usually do not make any decisions on this matter, and courts simply use the program with parameters set according to unknown criteria.
However, this described mechanism has the potential to transform from illegal to legalized scheme at any moment.
By approving the draft Procedure for the Operation of the Subsystem of Automated Case Distribution as part of the Unified Judicial Information and Telecommunication System, the Council of Judges of Ukraine stated that “for a judge with the least workload (workload coefficient), the probability range is calculated to be no less than 50% of the sum of the probability ranges of all judges participating in the automated distribution.”
As a result, firstly, even if the majority of judges in a particular court are honest, they will no longer be able to establish fair rules for automated distribution at their meetings, leaving an effective loophole for manipulating judge selection. Secondly, the formulation “no less than 50%” opens up possibilities to securely embed an even higher probability of random selection of a specific judge into the distribution algorithm.
Sweetening this pill could only be achieved by considering only the days when cases were assigned to a specific judge when calculating workload, with the judges’ assemblies stripped of the right to change this. While this is not very fair in terms of workload distribution among judges, it slightly complicates the use of the scheme for manipulation because after a judge returns from vacation, there will be a few days where the probability of selection will not be set at 50%, as described above using the example of the current provision in VAKS.
Next, the approved project by the Council of Judges through the State Judicial Administration was sent to the High Council of Justice, which on January 25, 2022, approved it as changes to the Regulation on the functioning of individual subsystems (modules) of the Unified Judicial Information and Telecommunication System.
A month later, at the request of the Supreme Court, the High Council of Justice made additional, albeit fundamental changes, establishing that when calculating workload, the number of working days worked by the judge in the current year at the time of automated case distribution shall be taken into account, unless otherwise provided by the decision of the judges’ assembly. Thus, by default, the most convenient and dangerous mechanism described above is already provided for.
However, after a full-scale introduction of the new procedure for automated case distribution was delayed several times by the decision of the Chairman of the Supreme Court. Then, the newly elected High Council of Justice on February 28, 2023, deferred it indefinitely, “until changes are made to this Regulation and the software of the automated court document management system is revised.”
Does such artificial inflation of the selection of a particular judge make sense to ensure fair and equal workload distribution among judges?
Abuses of probabilities in automated case distribution may be justified by the need to ensure fairness even for the judges themselves – allegedly necessary for the equitable distribution among them.
However, although parliament did not pass laws of mathematical statistics, they still apply. Even workload distribution among judges can be ensured without abuse, especially if not fixating on the idea of strict equality throughout the current calendar year.
In the case of distributing a large number of cases to judges, the workload gradually evens out. Even in courts where the number of cases throughout the year is not very high, it is sufficient to take into account indicators from the past few years.
And everything else will be done by the existing algorithm responsible for linear dependence on the graphs mentioned above. For example, in the mentioned story with Pashinsky in the Appellate Chamber of VAKS, even on the linear section of the graph, a judge with 1.6 times less workload already has a 8.4 times higher probability of being randomly selected. This approach effectively evens out the workload of leaders and outsiders by the end of the year, without creating convenient schemes for abuse.
The best evidence of this can be obtained simply by comparing the workload distribution in courts using different approaches. For example, let’s take the appellate administrative courts, as there are many cases, and look at the data for December 1, 2023 (towards the end of the year, but when judges have not yet started their New Year vacations).
In the Sixth Appellate Administrative Court, where there are no anomalies in linear distribution, the difference between judges with the lowest workload coefficient (3.5204) and the highest (3.5736) is only 1.5%. In the Fifth Appellate Administrative Court, where a judge with the lowest workload automatically receives a 50% probability of selection, the difference between the lowest workload (43.8961) and the highest (44.2100) is 0.7%.
There is no need for complex statistical studies to see the obvious: the “50% for the judge with the lowest workload” approach, which creates significant risks of abuse in case distribution, practically does not differ from fair distribution in terms of its impact on the evenness of judges’ workloads.
Specially for “Mirror of the Week.”
