Assumptions on Assumption
I assume that after reading this post, you will use this blog post as a reference when talking about the "assumptions" in Software Testing. We engineers, believing or hoping that the software we build will work itself is an assumption. An assumption that is carefully thought over and evaluated. Do you agree with that?
When was the last time you assumed the battery charge leftover in your cell phone is enough to make a quick call or to make a banking transaction? How did you know that the charge remaining is enough? Anytime that assumption was broken to you?
In another way to put it out, the software and hardware we are using is an assumption that is functioning as we expected to an extent. That also means we are testing the sets of programmed assumptions with conditions, data, states, and events when we test the software.
Did we assume when solving a problem in Math? Did assumption help to solve? We use theorem, hypothesis, corollary, and set of defined assumptions for the data we take in problem solving. But some might not agree and say we do not use assumptions.
Assumptions, Mathematics, Engineering, and Testing
I'm not sure who all will agree and disagree on saying assumptions are part of the evolution. If there are no assumptions, probably the evolution will cease. I buy groceries for a month assuming, I will survive this month. Should I call this -- assumption, hope, confidence, determination, evaluation, accuracy, etc Actually there is a very thin line of difference between the meaning of these words. In a way, these all look alike at the certain phases in the context.
When I started testing the Machine Learning systems, this learning started to become much cognizance to me. In fact, the AI/ML model is an assumption! But this assumption is evaluated on the set of data that we aware of it and know what it is to an extent.
I have to come to this understanding for now on series of evaluations in my Software Testing practice:
Testing is a science of problem learning, problem solving, and deductions. We assume certain things, and we infer conclusions from them.
If I replace the word "Testing" with Learning, Engineering, and Mathematics, I see it suits very well to my learning that I have been making as a Software Testing Practitioner. Should I keep these assumptions I made over certain data to improvise the solution and use it as "the solution"? That's the decision one has to take from learning out of assumptions valuing against what is expected.
Every problem solving will start, progress, and stops (not end) on the set of known assumptions made. If one is not aware of assumptions made and being done, then is that a problem? Or is this just as any other assumption?
Model, Assumptions, Architecture, and Testing
Do we build any engineering product without a model? The architecture, design representation, requirement, and strategy documentation are few models to mention here. Then what actually is the model?
A model is a simplified version of the observations. The simplification is for helping to focus on what has to be focused on while knowing what isn't being focused on and why. Isn't this an assumption? Yes, it is an assumption that is evaluated to an extent based on some (well thought?) assumption.
If we do not make absolutely no assumption about the data, then there is no reason to prefer one model over any other. Was there any day or instance, you came to a conclusion this one test data is enough to sample the system you are testing? Then why do people generate different payloads for XSS attacks? Isn't that payload test data? That payload is a model; it works or might not work. If worked till what extent and what did it uncover?
That payload when built by a Test Engineer (an Engineer) wasn't it an assumption? An assumption that will help her or him to discover information and help to learn about the system in a given context?
Every test data we identify, build, use or ignore is a model -- a modeled data on thought over assumptions.
Testing, Models, and Assumption
We cannot test without an assumption. Then we cannot build an engineering system without assumptions. At a layer, a working system is a model and in turn which is an assumption.
Rather than saying, do not assume, saying list out what you have assumed and why so. There is no model that is prior guaranteed it works. So the name model. The only to know for sure which model is best in a given context is to evaluate them all. Is it possible to evaluate all models? In other words, is it possible to test all the models?
Since this is not possible, in practice we make assumptions that look reasonable about the system, data, state, event, user, technology, and ourselves i.e. engineers. On making these assumptions we evaluate a few reasonable models.
Having models in the testing and automation will help in understanding and approaching the testing in the layer appropriate to the context.
Assumptions
It is not bad. It is needed. We use them every day in learning, problem identifying, problem solving, and decision making. Watch out for the coverage of the assumptions made and being made. Testing nor coding nor debugging cannot proceed further if we do use sampled and evaluated assumptions. Question the assumption. Questions the assumptions you are making.
