The notion of mechanistic explanation is given a very specific meaning by Maturana. Does it influence research in human sciences? In this article an attempt is made to answer this question.

Maturana’s theory of autopoiesis, and particularly his ideas on language and the ensuing construction of reality, are used as the major line of thought in this contribution. Some basic notions from his theory are used and extended as a framework to provide some explicit thoughts about what is considered as a particular omission in his theory: an affirmation of the impossibility to empirically observe the internal states of an autopoietic system. It is maintained that the definition of autopoiesis entails a concept of ‘state’ that coincides with ‘state transition’. A researcher who is interested in observing the internal states of an autopoietic system will become enmeshed in the impossibility to perceive his own perceptual acts. This is related to Merleau-Ponty’s notion of a ‘negative philosophy’, i.e. a philosophy in which the central theme is the impossibility for consciousness to grasp itself as a consciousness of the world. Finally, the concept of a ‘one-sided boundary’ is presented as a metaphor for the inaccessibility of closure. It is maintained that entering a closed system is not so much a matter of observation, but rather of its cessation.

The fact that the programme of second order cybernetics (which was called by Von Foerster the theory of the observing systems), failed to come up to expectations on the one hand, and the simultaneous success of Artificial Intelligence (AI) on the other hand, are our starting points to inquire into the problem of the modelling of cognition. Why didn’t cybernetics consider AI as a prolongation of itself? Why did Von Foerster see first cybernetics as inadequate to model cognition? Why have Von Foerster’s tessellations of ‘cognitive tiles’ passed into oblivion, whereas connectionist models are considered to be serious propositions to get at a model of cognitive abilities? On the basis of Von Foerster’s criticism of first cybernetics and his interpretation of memory, we inquire into the particular view on cognition in cybernetics of second order. The concept of ‘inverted reductionism’ is meant to give expression to the specificity and the inadequacies of this view.

After introducing the problem and a short impression of the work of Humberto Maturana, I elaborate on his ideas concerning science and the scientific explanation (1.) and objectivity (2.). In an evaluation of these views I claim that we cannot accept some consequences of Maturana’s theory, because they leave no space for discussion (3.). I then turn to Bruno Latour, and after a short introduction (4.) proceed to claim that there are some interesting similarities between Latour and Maturana (5.), especially in their respective ideas on the praxis of explaining. I further claim that Latour’s idea of ‘infrareflexivity’ constitutes a way out of the problems with Maturana’s theory (6.). I conclude with a ‘plea for taking dialogue seriously’ (7.). Along the way, it turns out, I could show Latour a counterexample of his claim that systems theorists are not aware of their own role in creating ‘systems’.

It is argued that a confusion among correspondence, consensus and coherence truth values is counterproductive, notably in the social field. Refinements of this rough categorization are discussed, as well as novel interpretations like dynamic coherence, akin to ‘hermeneutic’ truth regarded as ‘the iterative refinement of meaning’. It is maintained that dynamic coherence is fundamental to the other forms of truth prior to them, especially in specific domains of interpretation. Societal implications are considered insofar as a democratic organization is intended to exist, notably the difference between ‘unity’ and ‘uniformity’.

This paper spells out differences and similarities between the older or classical cybernetics and the recent and very different cybernetics which has been produced as a product of philosophical and scientific evolution. It is argued that considerations such as modal logics (encompassing self and other reference for imperatives, deontic forms and so on) and truth valuations of coherence and distinction are already giving rise to further fundamental developments, some of which are indicated.

When cybernetics entered the scene during the forties, high ambitions immediately arose and quite unusual claims were made as compared to those of traditional epistemology: cybernetics would do away with the distinction between ’mind’ and ’body’ (Papert, 1965); it would bring a new interpretation, through the artefact, of the Kantian synthetic a priori; cybernetics would give, following Turing’s approach of computability, a mechanistic sense to Kantian schematism (Dupuy, 1985, pp. 105–106). In this text we shall analyse the statute of that sort of claims. In other words, we shall show what is the relation between epistemological questions which we call classical or theoretical and those occurring within cybernetics, and more in particular in the work of McCulloch. McCulloch was clearly an advocate of experimental epistemology. He emphasized, in contrast with Kant for instance, that the experimental inquiry into the functioning, the emergence and the consequences of knowledge can be important for epistemology in general. Furthermore, we shall inquire how McCulloch’s experimental epistemology may be relevant to classical epistemology. Firstly, we shall discuss the position of McCulloch within cybernetics. Secondly, we shall deal with the meaning of his net of formal neurons and with the meaning of the heterarchic nets. This will allow finally us to illustrate in what sense McCulloch is to be situated outside first cybernetics and how he anticipates second order cybernetics. It will also allow us to show that the experimental epistemology which he defends can get an interpretation beyond the frame of a traditional reductionism. What we are proposing here is a minimalistic interpreta­tion: the epistemology, insofar as it tends to be experimental, doesn’t have to serve as a support for reductionism, but rather aims at pointing out concrete limits within which a theoretical epistemology can be developed.