Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 2

Automated Testing in AI Applications

As AI and ML are becoming more dominant in our lives, it's crucial to ensure these systems are
thoroughly tested to work as planned. So first it is important to understand how AI works to test AI
based applications / systems. The QA has to use and understand the AI / ML testing to ensure that
the AI models produce accurate and reliable results. Successful testing begins with a deep
understanding of the AI model. This involves understanding the underlying algorithms, data inputs,
and expected outputs.

It is important to understand the anatomy of the Test case design to generate a test case. Test Case
refers to a structured and sequential list of action items that attempt to verify a specific feature.

A test case design is a sequence of steps…..

 describing actions to be performed on different elements / controls.


 the test data to be used
 and an expected response for each action performed.

The process for Testing AI Applications include:

1. Understanding the AI application


2. Identify the controls / elements on the layout / screen
3. Understand what is on the screen, as an human operator, essentially called as an Object
intelligence , Reading Intelligence, Test Case design Intelligence.
4. After understanding all the controls, the tool has to generate a Test Case along with the Test
Data including edge cases.
5. Execute the Test Case .

The Global AI market is expected to reach a revenue $126 billion by 2025. So, it is crucial to ensure
that these AI based systems are tested systematically. Most often or not AI and ML based systems
are non-deterministic, which means they tend to show different behaviour for the same input data.
So it is important to have sufficient test data that accurately represents different behaviours. Since
the AI/ML works on different input models, QA has to ensure that the inputs are upto mark
otherwise the defects become more complex and a lot of time and effort is consumed to resolve
them.

Domain specific information is very necessary to create an appropriate Test Case otherwise we
might end up in false positives in defect reports. QA needs to develop test scenarios that reflect real-
world situations. For example, if the AI application is a chatbot, create scenarios where users post
common queries, unusual questions, or complex inquiries. Evaluate how well the AI handles each
situation.

In any AI based application data is the new code, so the application should be tested quite
frequently for the change in input data. This is quite similar to the traditional way to testing where
the change in the code needs testing of the application.

In health care industry, test suite plays a major role which consists of Algorithms and Test Data sets.
For example: the test scenarios include the patient demographics, patient Therapy and patient
outcome.
Very importantly the AI applications require extensive performance and security testing as they
often need to handle a high volume of data and requests. QA need to do the performance testing for
assessing how the application scales under different loads. This ensures that response times and
resource consumption remain optimal. Similarly QA need to perform security testing for identifying
the security vulnerabilities so that businesses do not face application downtime or data theft. The
QA need to continuously update the Test Case scenarios based on new security vulnerabilities.

You might also like