Center for Analysis and Design of Intelligent Agents

2020 Lecture Notes

Using Models to Validate and Measure - a.k.a. Simulation

What simulation is	A simplified model of subject under study - that is, a simplification not of the key causal factors in the phenomenon, which must remain in our model for it to be useful, but rather a reduction (sometimes a radical one) of the “extra stuff that really doesn't matter”.
What it does	Simplifies! Makes it easier to (A) set up testing conditions, (B) control independent variables, (C) make changes to the independent variables,(D) measure the results.
When to use	When the complexity of that which is to be modeled/understood becomes so great that mathematical models are intractable and hypothesis falsification would take decades, centuries or millennia, or is simply out of the question (as in e.g. astrophysics).
Kinds of simulation methodologies	Continuous time and state: E.g. differential equations. Discrete time/state: E.g. automata.
Relation between scientific theories and simulations	To build a simulation we need a theory that tells us how things relate to each other.
Procedure	Pick methodology. Decide which kinds of questions to answer. Model major states/transitions or input/output/functional properties of system. Run simulations with variations in independent variables. Note outcome. Fix model. Repeat.

Why We Need Simulation and When to Use it

Simulation	Simulations that “tell a story of a system” by integrating several observable and non-observable variables into a coherent whole, they can act as proper scientific theories. Simulations can also act as hypotheses when they concretize theories. Simulations are the newest methodology that science offers in our study of the (natural) world.
Model	A model is a representation of a phenomenon. A model of the earth-sun system can be created by a ball and a flashlight. Simulations are an executable version of a model.
Software Simulations	Implement a model of a phenomenon as runnable code. The model, and its implementation as a simulation, are informed by a higher-level theoretical or philosophical stance (example: A materialistic view of the world).
Verification / Grounding	To ensure that the results produced by a simulation match to reality, a process of grounding must be performed whereby the truthfulness of the model implemented as a simulation is ensured to represent the phenomenon correctly. This process is difficult; typically only a small part of the full model is something that is well understood in the physical world. It is expensive because for any reasonably complex system this process takes time.
When to use simulation	When the complexity of that which is to be modeled/understood becomes so great that mathematical models are intractable, or when they are not possible for other reasons, and experiments with standard hypothesis falsification would take decades, centuries or millenia.

EOF