Abstract
Important links appear between bounded rationality (Herbert Simon's view) and axiological cognitivist rationality (Raymond Boudon's approach). Understanding science as a kind of human action requires an explicit discussion of models of human being involved in its analysis.
I propose that a cross-fertilisation of the rational-actor model and the sociological role-theory, as proposed by Boudon or Hollis, could shift our present conception of rationality towards a better understanding of human action (including scientific research).
I maintain that our conceptual nets actually are informational filters, that our concepts are information filters, that, we are, as it were, some kind of semipermeable membrane. Our descriptions and explanations can only be constructed according to those filters. We select information not only by our internal capabilities, but it also depends on the context (just as happens with semipermeable membrane inside solutions of different concentrations).
The target of our actions is not, or not only, to optimize some singular variables (such as truth, rhetoric force or consistence); instead, human beings try to satisfy a set of values that we regard as relevant, for example, own authorship or agency. Contextual features can generate rules; we are prone to ascribe rules to cognitive capabilities of participants when those rules might be the output of the relationships themselves. We do not need to presuppose olympic participants with absolute and common knowledge (each one knows what the other one knows, and so on) neither in our human interactions nor in human-computer relationships; all we need to suppose is some flesh and blood human beings in contextual interactions.
Many philosophical approaches to language are built on a standard notion of rationality that shares some idea of optimization and some kind of generic principle that speakers try to adopt. The idea is very similar to utility and its optimization in neoclassic economic theory. If we try to understand the process of dialogue as a means to obtain optimal communication we would not develop a pragmatic approach. A pragmatic view requires only to satisfy some values rather than to optimize them. Real agents must not be blurred, they must always remain as, at least, a parameter of the interaction. In the standard view of rationality, the olympic agents could be eliminated because every agent is identical to the other; if every one was an ''epistemic god'', no one would be necessary. On the contrary, I argue that we will always need a concrete agent: the objectivity is not the view from nowhere or from everywhere, it is the view from somewhere (as Amartya Sen said). The foraging information theory (Pirolli and others), a great turn in human-computer interfaces, could be a first step towards an agent-based information theory that would change our preconceptions of information, optimization and rationality.