A New Perspective for Risk Management: A Study of the Design of Generic Technology with a Matroid Model in C-K Theory - Principia Designae: Pre-Design, Design, and Post-Design - page 202

We say that the process will at least go through the three steps: actually
additional steps are possible. For instance, the designer might himself add a new
“function” to be addressed—this is far from unusual: it would just means that the
designers add a constraint. Of course this new function should be then be handled to
keep independence.
13.3.4 Main Properties for the Design of Generic
Technologies
The model with matroid helps to understand critical properties on the design of
generic technology (GT): GT results from two “genericity building” operations: G1
(in step 2) connects by preserving (necessary) past connections; and G2 (in steps
1 and 3) completes the graph, ie creates new dependent structures in G based on
the first new expansive edge. These two processes explain apparent paradoxes in the
design of GT:
1. Because of G1: A generic technology includes both the longest cycle, hence the
“most constrained” one, and a local property (just one missing edge). This
corresponds to the logic of genericity in steam engine, where genericity was
made by working on reciprocating movement to link the technologies of steam
engine with the technologies of machine tools. More generally, the design of
generic technology consists in designing to creates new compatibilities between
“islands” of disconnected technologies.
2. Because of G2: Every new edge is a consequence of the first expansive edge
(hence a function of it) but, in the end, every new edge is also a possible
substitute for the initial expansive edge. Hence a generic technology also
encloses variances and alternatives. It is not one single solution but rather a set
of interdependent solutions.
3. Discussing the evolutionary model of GT design. We can easily represent in the
matroid model the evolutionary process mentioned above for the design of GT: it
consists in building new edges for each single new market. It is interesting to
note that this kind of random, evolutionary process can certainly help to com-
plete a connected graph but will hardly manage to create the edge that connect
two separate connected components: of course one market might require to link
to vertices from disconnected components, but this market would hence allow
for a technology that is not necessarily compatible with all already existing
technologies in the connected component. This only possible for such a connec-
tion is a market that needs all the function of the two disconnected components.
Hence a random, evolutionary process is quite unlikely to lead to a generic
technology; it should at least be guided by the requirement to connect discon-
nected components.