The ACS Canberra branch meeting last night was "Taming Complex Legislation and Policy: A Business-Friendly IT Approach" about the product "Oracle Policy Manager". This is software originally developed in Canberra by Softlaw, later Ruleburst and acquired by Oracle in 2008. It is a very Canberra product, being originally designed to automate bureaucratic decision making in the Australian Government. It is now used internationally for government and commercial applications.
The core idea of policy maker is to help people make consistent decisions based on complex legal and policy rules. Laws and policy documents are converted into executable business rules. The business rules are still in a human readable format, essentially a form of structured natural language. These rules can then be run using a software interpreter to make a decision. The rules can be used by an employee who enterws the data in response to questions from a client in per4son, via a telephone or from a form. Alternqatively the data for decision making can be from a databse. The most interesting implementation is where the client themselves interact with the system via a web interfaqce.
As an example the Swedish government use the system to decide on payment of dental treatment refunds. The UK government system use the system to decide who is an employee and who is a contractor for tax purposes. A key aspect of this is that the tax payer can use the system themselves and see the result. An example which goes beyond simply paying money, is the NSW system for professionals, such as doctors, reporting children at risk of abuse.
The Australian Government Visa Wizard is implemented using the system. In a more complex case, it is used by the US IRS.
A key part of the system is that along with an answer, the system can explain how it came to that conclusion.
What is interesting is that the system is essentially designed to take legal and bureaucratic documents and semi-automatically convert them into something in a very well defined format which a computer can understand. Obviously it would be better if those writing laws did not
use natural language and instead authored directly in a rigorously structured language and then generated the natural language version. But this would require retraining those who write the laws. The system can be used to identify ambiguities in existing laws and policies.
One advantage not mentioned in the presentation of using an automated system to give citizens advice is confidentiality. It is well established that people will tell a computer program things which they are not comfortable to tell a person.
All the examples given were of public sector uses, but the product is also used by insurance companies, retail and universities.