I resonate with what you wrote. The inadequacy of traditional user stories for probabilistic AI systems truly highlights the critcal need for more robust specification.
Herman, this is very insightful, thank you! I used to extend the User Story format with BDD - basically stating acceptance criteria in Given, When, Then format. You make me think I need to have a look and see what needs to change/is it still useful?!!!
I think BDD is still very applicable. We must also use it to define the expectations around the uncertain behaviour and fallbacks for the AI parts of a product or feature.
I resonate with what you wrote. The inadequacy of traditional user stories for probabilistic AI systems truly highlights the critcal need for more robust specification.
Herman, this is very insightful, thank you! I used to extend the User Story format with BDD - basically stating acceptance criteria in Given, When, Then format. You make me think I need to have a look and see what needs to change/is it still useful?!!!
I think BDD is still very applicable. We must also use it to define the expectations around the uncertain behaviour and fallbacks for the AI parts of a product or feature.