Challenges of AI implementation in Testing

Challenges of AI implementation in Testing

Length: 3 Minutes
Almost everyday and everywhere in newspapers and on social media we are listening about Artificial Intelligence(AI). Almost all industries are transforming the way they work till yesterday. We are moving towards automation. Now we have started adding Artificial Intelligence to the automation. Software industry is also not behind among those.

Definitely using AI will not be as simple as we think, there are multiple challenges we need to overcome while we implement or use AI in this transformation. All those challenges are applicable for Software testing as well.
What would be those challenges?

Deterministic testing Vs Non deterministic (Limitless) testing:

Human always do finite things, witch he can determine and analyse about particular thing.

For example: If individual has asked to test particular website. He will follow standard testing phases and will come up with few scenarios to test, can be called as deterministic or finite tests.

When we add AI to the automation, automation should consider how human is thinking. Otherwise we might land up in doing limitless testing and need to face below challenges.

  • Exhaustive testing.
  • Testing the areas of the application for which business has very less importance.
  • Less priority defects.
  • Duplicate defects etc.

To overcome such challenges, we should design learning model, used by AI applications in such way, that it will depict human behavior i.e. how human think.

How AI applications will think like human?

AI uses Machine learning and Deep learning which is actually a part of AI.

Brief about Machine learning, It makes ability to “learn” in machines. Which is achieved by using algorithms that discover patterns and generate insights from the data they are exposed to, for application to future decision-making and predictions, a process that sidesteps the need to be programmed specifically for every single possible action.”

How AI learning process would happen in Testing?

Let’s us assume we need to test Shopping cart website ( Example for the reference purpose only, in real world AI is used and will be use in the testing of more complex applications).

AI application might need to be exposed to the real world human shopping experience or might be exposed to dummy model which replicates human behavior.

After learning phase gets completed, we might faced above discussed challenges if not implemented properly.

Confirmation about AI application learned right or what we are expecting:

To overcome this challenge we need to define and implement a model which evaluates AI results based on some parameters.

Data privacy:

If AI applications are exposed to observe or learn or read live shopping cart user behavior, then we might land up in data privacy issues. As individual shopping cart experience could contain payment details, individual choices etc.

We might need to put some restrictions on AI applications or need to create separate test environment for this.

Return on investment:

This is also a very important factor to consider, however this will depend on industry type and size of the industry.

Readiness for acceptance:

Last but not the least is acceptance of this change and faith on results generated by AI applications.

These are few challenges I could think, if you think any more, you can add it in comment section.

Thanks for reading! You can share your thoughts in comment section.You can click here to Subscribe.

Leave a Reply