Here is the first blog post where I discuss the talks I attended at the UKSTAR 2019 conference. This covers the first keynote by Angie Jones, and the Deep Dive session run by Anne-Marie Charrett.
Keynote 1 – The Reality of Testing in an Artificial World
This first keynote of the conference was given by the amazing Angie Jones. I confess, I’d already watched this talk once before at the STAREAST techwell conference last year. They make a selection of their talks available online to watch on demand for a few months after the event and this was one of the talks I chose to watch. Angie is such an engaging speaker that, even though I knew the story she was going to tell, I was still on the edge of my seat wondering what was going to happen next.
Angie challenges the misconception that an application doesn’t need testing because the “AI is doing it!”. She produces several examples where machine learning has gone wrong, which may indicate that the application was not tested adequately enough. This is especially worrying as there may come a point where AI is incorporated into applications where reliability is essential. For example, an application that predicts if a patient is likely to get cancer. Some applications are too important not to test so we cannot be relying on them just working.
So how do we test it? Angie walks us through the process she followed when testing an AI application for the first time.
- First she learnt how the application works. This is really important as AI will have no pre-determinable results. Therefore, we need to know how it got to this unknown result and test that the AI is calculating the result correctly.
- Second, we train the system to see if the outcome is correct. Using test automation, we generate large amounts of data. We then test the outcome and see if that matches what we expect to appear based on the data we fed into the system. We repeat this multiple times with different sets of training data to see if the outcome matches what we expect.
AI is all about calculating results that cannot be pre-determined. We should be tested the method for calculating that result, not the outcome itself. If we are putting all our faith in an AI application and relying on them to getting the correct result, how can we NOT test this? People often ask if AI is something to be feared. Without testing, the answer is yes AI is something that should be feared.
Deep Dive – API Exploratory Testing
When attending this conference, my main focus was on test automation. However, I also wanted to learn something new. I was attracted to this deep dive session because API testing is something I’ve not done much before and really think I should start doing. Also, with so much focus on test automation, it was nice to learn something new about exploratory testing for a change. After all, both are equally important.
This talk started by examining how the use of mind maps can encourage testers to take a more systematic approach when doing exploratory testing. She started with the GUI and then went on demonstrate how this same idea can be used to explore the API. She walked us through examples of tests that might be carried out.
This talk was especially useful for API testing novices as she taught us how basic API commands worked, and how they can be used to test the API layer within an application. It was very basic, but useful for getting us started. Exploratory testing is all about learning and gaining experience. It is an excellent testing method to use for learning more about the application being tested and API testing.
Having never attended a deep dive session, I didn’t know what to expect. I really enjoyed the interactivity of the session. Anne-Marie is very good at encouraging audience participation. We were encouraged to suggest tests to run. During each ‘test’, before actually revealing the outcome, she would ask the audience “Whats the hypothesis?” encouraging us to think and talk about what we’d expect the outcome to be.