In the last article, we looked at test cases related to Apple Intelligence predictive test scenarios. In this article, we will look at Apple Intelligence predictive edge cases.
Generally, there are three types of edge cases that one could perceive on predictive app suggestions. Note that this list is not exhaustive, and subjective to the imagination of possible combinations of the tester. The three types of tests are:
- Rare or unpredictable user behaviors (e.g., sudden change in routines)
- Users with minimal historical data
- Conflict scenarios where multiple suggestions could apply
Let’s look at these scenarios in detail.
Rare or unpredictable user behaviors
A predictive suggestion for the apps to launch can be made only if the patterns of usage of the user is predictable! This looks like stating the obvious but honestly, if the user’s usage is erratic or not falling into a habit pattern, then it would become impossible for a prediction engine to suggest recommendations. There could be several reasons why an user’s behavior cannot be predicted
- Lack of a regular routine
- Sudden change of routine
- Loss of historical data
- Change of user using the smartphone
and so on. Again, this list is not exhaustive, but gives a very good idea about the kind of scenarios that can occur commonplace.
Users with minimal historical data
While loss of historical data is covered under the previous topic, the users might have minimal historical data which would make it incapable for the prediction algorithm to suggest apps. In this case, the users have to build history through constant usage of apps. From a testing perspective, this is a valid edge scenario and should be tested that unrelated apps are not shown as suggestions just because user does not have enough historical data.
Conflict scenarios where multiple suggestions could apply
There might be scenarios where multiple apps could be competing for being suggested. For example, if someone uses three different apps frequently and they generally use nine different apps sparsely, which ones of those apps are recommended should be thoroughly tested. The sparsely used apps should not be suggested. If there’s space for recommending only two of the apps and there are three apps that are constantly being used, it should be analyzed which two of those three are suggested. These test cases could get pretty interesting when all the three apps are all constantly used and there is conflict or competition between them to be listed.
Conclusion
Apple Intelligence Predictive edge cases could get really interesting depending on how inventive the tester gets on testing the predictions. There are several edge cases to be considered, but a few have been highlighted in this article. If you need a thorough review of your organisation’s test strategy on behavior of predictive algorithms, please feel free to get in touch with me. Happy to help!