Modelling type and modelling theory - Representing influences between states

[chcorbato]

This excerpt from D2.2 understanding candidate raises a profound question.

States can easily be represented as Modelets, but representing influences requires modelling a procedural mechanism, \xie\ the exertion of Models by Engines ---the agent's abilities are a special type of influence, one under control by the agent. We discuss both in
section~\ref{s:us.engines}.

I am not sure that representing influences (I would use another term, relations?) between states requires a procedural mechanism. Equations represent such relationships, and they are not procedural. The equation is different from the procedure to use it to obtain information (e.g. future state values)

This raises the question of which different kinds of models do we consider?
E.g., models that are structural (e.g. the phenomena consists of this set of state variables, or a concrete set of {variable,value} pairs), and models that represent behaviour (e.g. the equations that represent the evolution of the values of those states relate those variables with their future values).

[alex-quin-gabriel]

I think we have to clearly separate two abstraction layers here:

Metacognition / Understanding Logic

- here we have models of engines and models of modelets
- we reason about their properties
_____________________________________________________________________

Cognition / Agent Behaviour

- here we have modelets and engines
- engines exert modelets

With that being established, within the understanding logic, we don't have procedural mechanisms other than the reasoner.
that means, the engine models, are not procedural, but they represent procedural phenomena.

we need the capability to represent procedural phenomena in the agent (engines, idle battery drain, heat accumulation)
but we also need the capability to represent procedural phenomena outside the agent (e.g. an apple falling, or a person walking) if we want to project the present into possible futures. ....the mechanisms we have for that are engines.

[alex-quin-gabriel]

I think we are back at the question of where an engine ends and where a model begins.

if engines are just facilities and algorithms (e.g. how do I use math to project motion) are models, then we have to get a lot of ROS programmers to think differently.

if algorithms are inherent to engines, then they have (implicit) models of what they do, that we have to make explicit to some degree to use them.