Today I had the great privilege to eat lunch with professor Juan Ilorens, also the CEO of a company on the technical forefront, The Reuse Company. They work with natural language processing, requirements quality and information reuse.
I tried to provoke him by asking: “Do you think the written requirement as we know them are doomed to disappear by the increased use of different modelling techniques? Everybody today seems to talk about Model Based Systems Engineering, MBSE, as the future of Systems Engineering.”
He stopped eating and looked at me with his typical smile and said: “The challenge for a modern Systems Engineering organization today is to know how to manage their knowledge. When we have tested the level of understanding we see that people tend to understand text better than a plain model without any text in it.”
We must remember that requirements as well as models are to be used to capture knowledge and design decisions to pass information downstream to the persons in need of that information. The goal for many organizations today of using exclusively models to capture this kind of knowledge will simply not work. The different stakeholders don´t have the same levels of modelling skills and will not truly understand what they see”.
But isn’t models any good, I asked?
“Yes, models have obviously benefits if you use them correctly and in the right way. The tendency today is basically “all or nothing” and that’s wrong. Models are good for automatic processing, simulation, analysis, code development, test case development, etc. But today the MBSE community says that models are great for stakeholder communication and I believe this is just wrong or deeply exaggerated”.
Juan continued: “We assume that an organization understand the modelling notation just as good as natural language like English, but we have learned to interpret models later in life at university or later and theoretically it works to communicate through models, but in real life it doesn’t.
Engineering today seems to be a fight against complexity and in order to defeat it we introduce modelling as the universal medicine solving every issue for everyone. It’s just about adding views and diagrams and we can manage every need. But we are creating monster models that way. Today’s technology and systems are so complex that we more than ever need communication between different groups of people and that communication is reduced by models not the other way around.
Engineering is to communicate and agree on different things and we work in teams, often distributed. We need a way of working that maintains the information and knowledge over time. I think this needs requirements written in natural language and we need traceability between different knowledge assets. Our notion in the Reuse Company that engineering teams needs help in analyzing correctness, completeness and consistency of different knowledge assets like requirements and specifications will soon be expanded to include even models. Imagine comparing a System Specification with a System Model to find errors like inconsistency or incompleteness. That will truly mean a revolution for engineering! That’s why we today have started to talk about Knowledge Engineering at Reuse.”
So, Juan. You mean that Requirements written in plan and natural text will co-exist with models in the future and not become replaced by them?
“Yes absolutely. When we help people to write high quality requirements using structured techniques like patterns we can start to build powerful reasoning around the requirements themselves. If we also add the possibility to compare them against other assets like models we will become super effective and much smarter than today. We will soon release methods and tools that handles this. We are talking about an engineering knowledge revolution my friend”.
Now it was time to say goodbye after an extremely nice and interesting lunch. I can’t wait to see what Juan has up his sleeve. Next post will be on MBSE. I need to look more into what is said out there about this subject. But I am now more than ever convinced that requirements are here to stay. If we speak to one another using natural language, we will need to capture the different stakeholder needs with natural language requirements. But we shall use modelling as a complimentary technique to manage complexity, make simulations and other cool stuff. Requirements are here to stay, mark my words for it.