The purpose of a policy model is to make decisions on your behalf, often applying complex sets of rules. A lot of work goes into automating the decision making process so that every decision is fair, consistent and traceable. So why on earth would you want to allow a ‘human’ to over-ride this decision? There are times when some human involvement is desirable because to automate all aspects of the process may not be practical - a policy model can only make a decision if it has all the rules it needs. Examples where some discretion might be required include: medical diagnosis, staff promotion, bonus payment, application for a grant and travel approval.
Allowing discretion in rules must be carefully thought through. A policy model should always be allowed to arrive at a decision itself where it is possible for it to do so. If permitted, the rules must be modelled to clearly state discretion has been applied, and the decision report as a consequence must reflect this.
The following fictitious example provides a simple illustration of how two areas of the rules have been constructed to allow the user (a Deciding Officer) to over-ride the automated decision.
A policy model has been authored to determine whether an applicant has a Centre of Interest in Atlantis based on a number of criteria. A decision can also be based on answers provided by a Deciding Officer. A positive decision will impact an applicant’s ability to receive certain government related payment credits.
Rules to provide discretion have been modelled in Sections 1 and 3 as highlighted by the red boxes. If the applicant’s family members are not located in Atlantis, the Deciding Officer is provided with the option to over-ride the decision.
The Deciding Officer is also able to over-ride the overall decision at the end of the assessment using a similar procedure. If the system determines that the applicant’s Centre of Interest is not in Atlantis, then it is possible to over-ride this and add other factors which might be important to the determination.
It is possible to limit the ability to over-ride and use discretion on the rule decision to a limited number of end-users (i.e. Senior Deciding Officers). In this example however, all users
are provided with this opportunity. Any reasons for granting discretion must be entered by the Deciding Officer during the assessment, and these are saved as part of the interview. The journey for the user commences with the Summary Screen, which provides an introduction and a link to start the assessment:
Personal details are entered on the next screen and then the user is presented with a series of questions based on the rules to determine the applicant’s Centre of Interest:
If the user answers ‘no’ to the question relating to the close family members, the policy model contains rules to allow for discretion at this point. The Deciding Officer is provided with an option to state whether other factors should be taken into account:
The Deciding Officer is required to enter other factors which they believe should be taken into account for this assessment in a free-text format:
The policy model checks the reasons have been entered, but does not investigate whether specific factors are relevant or valid. Rules can be created to allow the user to select from pre-defined lists of options prior to this to categorise the discretion more tightly. In this example the discretion is more open.The Summary Screen is displayed and the determination is presented:
Whilst the decision is positive for the applicant, the Deciding Officer is reminded that discretion has been applied. Selecting the ‘Why’ link displays the decision report, which shows the decision was made with discretion applied by the Deciding Officer:
A record of the reason for discretion is then stored within the session case file, that is then saved and stored in the host application.If the policy model determines that an applicant does not have a Centre of Interest in Atlantis, the rules will provide an option for the Deciding Officer to use discretion and over-ride this decision as well:
In summary, the ability to over-ride a policy model decision should be treated with care. In some circumstances, factors may impact a decision where it is not feasible to model them in rules.In all cases, allowing the policy model to automate the decision making process should be considered by default.
This Blog is brought to you by Oracle Consulting Services. Further details about OPA enablement, coaching and mentoring services can be obtained by contacting the author.