From Agent-Based Models to Network Analysis ( and Return ): The Policy-Making Perspective
Finally , we restore the original positions of the agents / nodes , showing that it is , in any case , possible to calculate and to display the indicators based on the network algorithms . The original positions are important in spatially critical networks , e . g ., a network of hospitals ( Figure 5 ).
Figure 5
This approach innovates with respect to other interesting works on ABM and networks ( e . g ., De Caux et al ., 2014 ; Hamill & Gilbert , 2009 ) where the agents act to generate the network ; in our context , agents are activated , following their internal rules and capabilities , by the events , and the network emerges as a side effect , as in the real world . 6
Moreover , with respect to the other approaches ( Carayol , Roux , & YildizoGlu , 2008 ) that attempt to build networks through various characteristic recombination of links , our approach has no intrinsic limitation in the number of agents to be modeled , in that it does not work in combinatory terms . It follows that it can be used on a scale that can satisfactorily approximate real-world phenomena . In addition , the possibility of linking ABM and
NA in such a straightforward way is ripe with implications for policy analysis . Let us see them in more detail .
4 . A RecipeWorld for Economic Policy : Exploring Reverse Engineering
In Figure 6 , we represent the relationships between the real world , the simulated world , and the conclusions that can be drawn from the model . ( A ) Is the actual world populated by entities e1 , e2 , …, en and by their actual network ; ( B ) is an ABM where agents a1 , a2 , …, an are mimicking the actual behavior of the entities in ( A ) via the execution of orders and recipes ; ( C ) represent the agents of ( B ) generating a network , which — if the agents as construction are correctly managed — is similar , or very close , to that in ( A ).
84