The DWP has denied using algorithms in the course of making LEAP decisions that have prevented many thousands of PIP claimants from getting arrears, but appear to have admitted to using ‘business rules’ instead. However, there is a big question mark over how ‘business rules’ differ from algorithms and we are now looking for a specialist who can help us investigate this further.
As regular readers know, the LEAP review was set up by the DWP to identify claimants who had missed out on large sums of PIP, from £3,000 to £12,000 each, because the department got the law wrong.
Originally, the DWP estimated that 164,000 claimants were due arrears of PIP due to an error in relation to the mobility component. Yet, with well over half of 1.6 million claims allegedly reviewed, the DWP have only made awards in 3,700 cases.
Benefits and Work made a Freedom of Information request for the Data Protection Impact Assessment (DPIA) carried out in connection with the use of Automated Decision Making (algorithms) for the LEAP review.
If the DWP had used algorithms, there would be a legal duty to carry out a DPIA.
However, the DWP have replied that no such document exists and instead referred us to a parliamentary answer by Baroness Stedman-Scott in response to a question about which algorithms the DWP use to decide on eligibility for benefits:
“Decisions on applications for welfare payments and services are made by the Department colleagues. The Department does not use algorithms to make decisions in this way. The Department does use business rules, some of which are automated and focus on everyday repetitive processing tasks so that colleagues can spend more time supporting vulnerable claimants.”
We know that the DWP have drawn up a list of the conditions they consider are most likely to result in an award of arrears under LEAP. We also know they have a table of which age groups are most likely to get an award.
But that was how they arrived at a figure of 164,000 claimants being eligible.
Clearly other rules must be being used as well to exclude claimants from being considered.
So, what we don’t know is when a set of ‘automated business rules’ becomes an algorithm.
A business rule defines or constrains some aspect of business and always resolves to either true or false . . . For example, a business rule might state that no credit check is to be performed on return customers. Other examples of business rules include requiring a rental agent to disallow a rental tenant if their credit rating is too low, or requiring company agents to use a list of preferred suppliers and supply schedules.
The programs designed specifically to run business rules are called rule engines. More complete systems that support the writing, deployment and management of business rules are called business rules management systems (BRMSs).
Many commercial rule engines provide the Rete algorithm, a proprietary algorithm that embodies many of the principles of Rete.
Why automate decisions?
Automating decisions through a business rules management system (BRMS) enables businesses to create and manage business logic independently from applications and processes. Businesses can also leverage AI and machine learning to make decisions with precision, targeting each customer interaction intelligently.