Dear Michel, I spent my career doing much the same thing with mutual information, which in this case quantifies the degree of constraint among the species. <https://people.clas.ufl.edu/ulan/publications/ecosystems/gand/> Encouraged by the suggestions of E.P. Odum, I hypothesized that ecosystems would naturally increase in the product of their gross activities times the mutual information among the network of interactions -- a product (fashioned after the Gibbs/Helmholz free energies) that I called system "ascendency".
After about two decades of measuring the ascendencies of diverse ecosystems, the data were telling me that my hypothesis was wrong. Ecosystems do not continually progress in increasing ascendency, but rather achieve a balance between ascendency (a surrogate for efficiency) and its complement, the system overhead (which mirrors reliability). Furthermore, the quantitative nature of the balance is notably insensitive to the type of ecosystem under study, averaging about 40% efficiency and 60% redundancy. (See Figure 7 on p1890 of <https://people.clas.ufl.edu/ulan/files/Dual.pdf>.) Now, you might argue that constraint is not information and so these results are not germane to our discussion, but I (and I think Stan) would propose that constraint is actually the most generalized form of information, and the Bayesian forms of the Shannon measure beautifully parse the division between efficiency and reliability. While I didn't set out to falsify my initial hypothesis, that is indeed what eventually happened. Notice that it was accomplished in quantitative fashion and without any recourse whatsoever to system dynamics. The decades-long exercise demonstrates, I think, a phenomenological approach to the science of life pursued in abstraction of (but not contradiction to) the underlying physics and chemistry. Je lis un peu francais et voudrais bien lire de votre travail sur l'information mutuelle. Cordialement, Bob > Dear colleagues > > Loet thinks that "Nobody of us provide an operative framework and a > single (just one!) empirical testable prevision able to assess > "information" > > In my ecological work, I try to know the relations between living > organisms and their environment, and I use Brillouin's formula (and > non-inferential statistics) to compute the "mutual information" between > each species of plant or animal and and each constraint of the > environment. The testable prevision is, for example, the potential area > of a species. > > The book where that method is explained is written in french, but I > could translate this example in english if you think that it could be > published. > > Cordialement. M. Godron > > > _______________________________________________ > Fis mailing list > Fis@listas.unizar.es > http://listas.unizar.es/cgi-bin/mailman/listinfo/fis > _______________________________________________ Fis mailing list Fis@listas.unizar.es http://listas.unizar.es/cgi-bin/mailman/listinfo/fis