1) The document discusses different approaches to modeling complex systems, noting that current approaches are often limited and idealized. It proposes that a new formalism is needed that avoids ideals and encapsulates all potential modeling perspectives.
2) As an example, it compares rational modeling, based on ideals, to relational modeling using large datasets. Relational modeling inverts the concept of empirical research by using "data deluge".
3) Self-organizing maps are presented as a method that creates abstractions from data while preserving topological relationships, providing both a transformation and primal-dual representation. This allows SOM to be used for various tasks from simulation to computing with contextual numbers.