Levels of Usage

The creators of Decision Model and Notation intended it to be used in a variety of ways by different organizations and also within the same organization. The division of the models into Requirement,  Logic and Implementation levels pre-empts usage scenarios for:

• Modeling Human Decision-Making
• Modeling Requirements for automated decision-making
• Modeling automated decision making

Some text books and white papers talk about these as levels of maturity in the field of decision making, but fail to explain that there are many situations - even in highly automated systems - where it is important or even imperative for human beings to be making the final decisions, particularly in areas of human safety; these will not be automated, but are required to be described for governance and regulatory reasons.

Modeling Human Decision-Making

DMN can be used to model enterprise, organization or initiative level decisions that are performed by staff rather than computer systems.  Typically the decisions are described at quite a high level and the rules are commonly written in a natural language but can be described using more formal mechanisms such as Decision Tables. At this level it is useful to model how knowledge relates to decisions using Business Knowledge Models that capture specific areas of business knowledge and its applicability to one or more decisions for example  a set of Standard Operating Procedures. Knowledge Sources can also be modeled that describe the sources of Business Knowledge for example a Standard Operating Procedures Manual. This type of modeling can be both descriptive and prescriptive.

Modeling Requirements for Automated Decision Making

When DMN is used for modeling the requirements for an automated decision making system it is important that the inputs and outputs are defined and the models are developed to the level of fidelity needed for an automated system.

Modeling and Implementing Automated Decision Making

Enterprise Architect is a full-life-cycle tool that can be used to model decisions for Automated Decision Model systems and can generate programming code in a number of languages, which allows the decision requirements and their accompanying definitions to be enshrined in implementation code. This is a useful mechanism that leads to great productivity and ensures that there is an unbroken chain between the highest level thinkers in an organization and the final automated systems that process the decisions and determine their output, all within the one tool.