The year 2023 has been marked by technological advances and paradigm shifts that have redefined conventional approaches to testing. The adoption of artificial intelligence by some companies has helped optimize their test campaigns, increasing efficiency and test coverage. At the same time, automation has seen strong adoption, significantly accelerating production cycles. Accessibility issues have taken center stage, prompting test professionals to integrate practices that ensure applications are usable by all.
As we look back on these challenges, what are the prospects for the current year? Our colleague Marc has the answers. Through his perspectives, he presents the highlights of the past year, while anticipating the emerging trends that will shape the software testing landscape in the coming months.
The year 2023 was a rich one. However, if I had to pick out 2 highlights from 2023, they would be
Artificial intelligence (AI) represents a major complexity in the field of testing. The test perimeter is not clearly defined, as AI is not constrained to link a specific output to an input, making testing difficult. Working with data, selecting the right algorithms and verifying results are crucial. Moreover, AI is prone to bugs, often unanticipated by the team, as illustrated by the following example shared by a friend
—–
The army wanted to use artificial intelligence to determine on a battlefield whether the tanks present were friendly or enemy, in order to decide whether they should be targeted by drones. They set out to do this by processing a vast amount of data, including photos. However, when implemented, the system proved ineffective. They then identified the problem: all images of friendly tanks were taken in good weather, with the sun shining brightly, while those of enemy tanks were taken in bad weather with clouds. As a result, the artificial intelligence was actually just reporting weather conditions rather than distinguishing allies from enemies.
—–
These challenges, often obvious to humans, are not necessarily so to machines, underlining the importance of considering them during test campaigns. It’s also vital not to depend entirely on AI, to see it as a decision-making tool rather than delegating the final decision to it, and to carry out regular reviews with necessary adjustments in case of error, recognizing that AI, just like a human being, can make mistakes.
Practical experience and training are essential to familiarize yourself with AI, understand how it works and enable informed comparison with other tools.
In the context of digital services, the possibilities seem endless, but exhaustive testing is impossible. The key lies in focusing on the notion of value, and in particular on sustainable quality, which implies maintaining quality over time. However, digital technology presents three major issues: firstly, its growing environmental impact requires an eco-design approach to guarantee its long-term viability. Secondly, the issue of designers’ well-being is emerging, with sustained work rhythms sometimes leading to burn-out and disengagement. Finally, it’s crucial to remember that digital services are first and foremost services, designed for users. Deviations such as an orientation towards profit rather than the real needs of users are to be avoided.
Accessibility, although a long-standing issue, has never been a priority. Its current importance stems from its forthcoming mandatory nature, similar to the way the RGPD has increased the importance of security. Penalties for non-compliance are not very severe, but companies must anticipate and assume the costs involved in bringing their sites up to standard.
For me, the priority lies in sustainable quality, but everyone should determine their own priority topics according to their preferences and skills. I liken it to deciding what to test on a piece of software. The people involved in testing should think of themselves as the software itself, assessing their knowledge and preferences to decide which topics to invest their time in to progress and improve. This may involve one, two or three topics, depending on individual preferences, whether technical or otherwise. Guidance should also take account of market skills and demands. This is my recommendation.