Consensus on the Path to the Answer

I have concluded that Behavior Driven Development (BDD) and backdoor verification are a bad mix. To force BDD to introspect a system violates the nature of the BDD vision of treating a system as a black box. When BDD is used to measure the value of a System Under Test, then the system should emanate  an ‘externalized’ signal (as output) which can validate the value of the system as providing something. To assume that a backdoor process can verify that a system ‘arrived’ to the correct conclusion means the ‘arrival process’’, as context, needs to be verified; Otherwise the conclusions can be either chalked up as ‘lucky guess’, or a contrived example using carefully selected inputs. The end result is a test process which could produce a trivial or simplistic example of success. As systems become increasingly more complex in their problem solving ability, there will be an increasing need to understand how these systems arrived at their answers. This will involve some level of transparency in the process which still avoids opening the system up for inspection. In other words, this tool should allow 3rd party observers to judge if the arrived conclusion was valid or not. The challenge of keeping proprietary algorithms as intellectual property and at the same time allowing consensus on safety as an observable characteristics seems to be a market necessity and a public right.

As a parent I often wonder if rote learning is too often the mode of education and not meaningful cognitive problem solving skills. As we approach a new age of machines quickly out pacing our capacity to unravel problems, I hope we as engineers can learn to engage these black boxes and engineer the necessary tooling to have these systems clearly explain how they arrived at their answers in a way that we as slowly evolving carbon life forms can all understand.