Writing use cases for AI systems question.

Hi,

I'm creating my first AI system for a project and am beginning by drawing some use cases.

I was wondering, is it acceptable to have the AI system as the actor in the use case?

The AI will be pro-active sometimes, it has it's own use cases to perform. when it detects something in the real world via a hardware device, it performs it's use case which assists the human also via hardware.

Do you think that is OK, or should an AI never be an actor?


AI and UML are new areas to me, and so am learning about how they are normally used together.

Many thanks for your thoughts.

:)

Quote:

---

Originally Posted by Orange Lozenge

I was wondering, is it acceptable to have the AI system as the actor in the use case?

The AI will be pro-active sometimes, it has it's own use cases to perform. when it detects something in the real world via a hardware device, it performs it's use case which assists the human also via hardware.

---

Hmm.. nope. You can't.

An actor is an external agent. More precisely the role that agent performs over your system. And it is not a subsystem (your AI) of the system you wish to describe through Use Cases. So, for the example above -- not knowing exactly what conditions are detected by the hardware device -- the actor will be whatever aggregate term you can use to describe the external input. If this was a device that measured weather conditions and you wished to describe a Use Case for your whole system, you would start by choosing what level of detail you wish to describe. The actor could be "Outside Temperature" if you wished to describe only the behavior of your system to that input, "Weather" if you wished to describe your system behavior in more general terms perhaps, or a series of actors like "Outside Temperature", "Humidity" and "Atmospheric Pressure" if you which to describe your system behavior in more detail and in the presence of multiple inputs.

Your AI system can only be an Actor in a Use Case, when you want to describe another system in which your AI is an external agent/role that drives it.

EDIT: Sorry for not having seen this question earlier. Somehow missed it.