What does it take for the word ‘system’ to be a tool rather than an ornament? Mazur's answer is that the concept of a system becomes useful only when we use it according to strict rules, and he gives five of them: precision in defining the system, its invariance in the course of the reasoning, completeness and disjointness of the division into subsystems, and functionality, that is, singling out systems according to what they do, not according to what forms a separate solid body. He illustrates the last rigour with the example of three workers lifting a beam at the command of one of them: in terms of functions, these are not three subsystems but two, the directing one and the executing one.
Applied to decision problems, the rigours lead to a division into three kinds. Since the new system, the old system and the transformation between them exhaust everything that can be specified, there are only postulation (what to achieve), optimisation (how) and realisation (from what). A diagram in which the optimiser and the realiser are coupled with the postulator and with the environment, and with each other only through the postulator, shows that goals, ways and means adjust to one another mutually, not in a simple sequence; the caption of the figure sets this diagram beside an autonomous system.
The author takes the starting point of the whole argument from the place of cybernetics in science. In his view, it fills the gap between mathematics, which is general and detached from the concrete, and the sciences that deal with segments of reality, so its results can be transferred between fields. The concept of a system is meant to serve this end: to cover as much as possible without blurring the differences between objects. Hence a system is defined by elements and relations, and a structure by relations alone, so that two different systems can have the same structure. Mazur does not enter here into disputes about definitions; he simply asks what the concept is meant to serve, and he regards systems theory as part of cybernetics, not the other way round.
The bibliography records the first printing in the proceedings of a systems engineering school in 1976; the text we use comes from the reprint in Postępy Cybernetyki of 1987, published after the author's death. The same division of the decision process, shortened and with emphasis on the qualifications of decision-makers, returns in the 1980 article; on autonomous systems, see Cybernetyczna teoria układów samodzielnych (‘A Cybernetic Theory of Autonomous Systems’, 1966).
Our own account, working version; definitions of concepts are in the dictionary, with their sources.