ModelDriven-AI: the integrated Artificial Intelligence approach
Currently interest in AI is mainly focused on data-driven, machine learning based AI. Cheap data storage, fast processors and advancements in algorithms and neural networks have fueled this interest. There is certainly a touch of magic in predicting what will happen next.
Data-Driven-AI - looking into the past and predicting the future
The data-driven, machine learning based AI approach identifies what is the right answer based on having “seen” a large number of examples of question / answer pairs and “training” it to get to the right answer. For example fraud patterns and car insurance claims. The use of data-driven AI approach extends to unstructured data like pictures of for example what is a dog or identification of types of boats for fisheries. This approach with machine-learning is based on what we have seen so far.
The data-driven, machine learning based AI approach has value to many businesses, but is unexplainable by its nature and unsuitable for situations where explanations and audit-ability are needed. It is non-deterministic and cannot cope with the enforcement of changing regulations. And once it has identified a pattern or determined the likelihood of something, it does simply not know what to do with it on its own.
Expert-Driven-AI - the policies of now and acting compliant
This is where expert-driven and knowledge-based AI approaches are complementary to the data-driven approach in the AI-spectrum. And adding something very important: the ability to decide based on regulations and polices. To automate and to keep these decisions up-to-date when regulations change. For example with tax-calculations, where business-rules ensure the equal and transparent treatment of taxpayers. Doing this by capturing expert-knowledge in sets of explicit, unbiased and explainable rules. This approach is deterministic and uses proven technology.
In the context of process automation, we see that we have basically two ways to combine to build systems that automatically can determine what they are supposed to do next: data-driven and expert-driven. But how to combine these two complementary approaches to Artificial Intelligence for the maximum currently possible? This is where since 2015 the new industry standard Decision Modeling and Notation (DMN) kicks in.
ModelDriven-AI combines the two approaches
ModelDriven-AI uses the international modeling standard DMN to integrate the two different approaches: data-driven machine learning and expert-driven rule-based Artificial Intelligence. With DMN, Business Analysts now have the tools to clearly point out which part of the decision is to be answered by expert-knowledge and which part through data analysis and machine learning. And which various data sources can or should be used. Thus providing data scientists in their team the much needed guidance too.
As the technology matures we need to learn to capture and share knowledge and experience in the right, appropriate ways. Artificial Intelligence is here to stay and Explainable-AI currently is a hot topic. There is a lot of existing work within artificial intelligence research that can help with this. Being open to this results in better understanding intelligent systems, using combined data-driven and expert-driven AI approaches to solve many existing and new problems.
