Gossip modeling
In the previous period of project development, a series of models often need to be proposed. What is the role of these models? How should I evaluate these models?
Briefly, the purpose of the model is to strive to describe the reality and explain certain relationships in reality. We have too much insignificant details for realistic cognition, and they often mislead our work, so we need model to simplify reality.
The essence of modeling is cautiously makes appropriate assumptions, omitting realistic details to achieve a degree of simplification, so that reality can be effectively described. In general, we consider more details when we need to omit, rather than which details should be included in the model. This requires determining a proper simplified level, so we need a trade-off between versatility and robustness. On the one hand, we will not need a model containing all the phenomena, and on the other hand, it can only explain the model of a single problem, and there is no great value.
Imagine a map, we can treat it as a model of a geographic area. There will be many different maps in the same city. Some maps have marked tourist attractions in the city. Some maps have marked the names and location of the major buildings in the city, and some marked the bus and subway lines. and many more. You will find that there is no map, the reason is clear, you won't use such a map that cannot be read - it lost the value of modeling, such as simple.
These maps provide different models of the current status of the same geographic area, each of the maps are simplified by many details on the geographic area. On the one hand, the specific purpose using the map has limited its versatility, and each map is just a number of target services, without a map that can be fully replaced with other maps. On the other hand, it is necessary to maintain a certain versatility, we can't draw a map in order to mark the No. 1 bus route, and draw another different map in order to mark the 2 bus route.
How should we evaluate these different models? A basic fact is: No scientific method can absolutely judge that the map containing the bus route must be better than the map containing the tourist attraction. Obviously, depending on your purpose, only the modeling target of the modeling can decide which model is better. Therefore, the evaluation of the model must first explicitly model the goal, which is that these goals tell us how to simplify reality and should omit what details.
