4 Examples of Smart Case Weighting

To illustrate the power of smart case weighting, take a look at two examples.

First, consider the simple example mentioned above (see table 2.1): A court system with just three courts, ten judges and two case types (A and B). We saw above how applying traditional case weighting to this system would lead us to recommend that - in order to balance workload - the courts should have 2, 3 and 5 judges (a transfer of one judge from Court 1 to Court 3).

Table 4.1 shows how the asssesment would be if we had instead applied smart case weighting, either with 20% weight uncertainty or with relative weights (A>B). It turns out the two smart case weighting approaches end up with exactly the same recommendation for the optimal allocation as did the traditional case weighting approach7. All the case weighting assessments are furthermore different from both the current allocation and the “naive” allocation we would end up with if we had wanted to balance unweighted cases.

Table 4.1: Optimal allocation of judges according to different asessment methods
Assessment methods
Court Current number Naive: No weights Traditional: Fixed weights Smart: 20% Weight uncertainty Smart: Relative weights
Court 1 3 1 2 2 2
Court 2 3 1 3 3 3
Court 3 4 8 5 5 5

This is a strong result. Imagine people had been complaing about the case weighting system, implying they were treated unfairly by the way weights were set. Now we can show that even allowing for a faily large uncertainty about the weight (plus/minus 20%) - or just relative assumptions about the weights - we will reach exactly the same result.

As a second example, we return to the data from the Danish courts. Imagine we had applied smart case weighting with either 20% uncertainty around the weights, or with just the knowledge about the relative order of the 17 case weights.

Table 4.2: Comparison of recommendations from alternative models
Assessment methods
Traditional: Fixed weights Smart: 20% Weight uncertainty Smart: Relative weights
Courts to reduce number of judges 16 14 14
Courts to increase number of judges 18 14 13
Number of judges to transfer 19 16 15
Number of reductions the model agrees with the fixed weights model on 16 15

Table 4.2) shows the results from these three models. The traditional case weighting model would - in order to establish balanced workload - recommend transfering 19 judges. 16 courts would have to decrease the number, while 14 courts would have an increase.

The traditional case weighting model require us to assume the weights are exatcly right. However, if we have less confidence in our ability to measure the weights exactly, we can opt for the model with 20% weight uncertainty. This model identifies 16 transfers. In other words, 3 out of the 19 transfers suggested by the traditional case weighting model are not justified if we allow for weight uncertainty.

Finally, the smart case weighting model with relative weigths identify 15 transfers. So, just by knowing the relative order of the weights we are able to identify 15 out of the 19 transfers recommended by the traditional case weighting model.

  1. To be precise: The smart case weighting approach with relative weights identifies a need to remove a judge from Court 1. Some possible weight combinations may, however, lead to the conclusion that the judge should be moved to Court 2 instaed of to Court 3. There are different approaches to determining which court is most likely to need the additional judge. The most common of these approaches will all identify Court 3 as the one most likely to deserve the additional judge.