Monday, 4 August 2014

Decision making in biosystems

What is decision making and why do they say sometimes that biosystems make decisions? 

Decision making is, by definition, a choice from alternatives as part of planning. A choice is made by an agent, expert or decision maker. Choices are made for a purpose given some conditions. Imagine yourself in a shop with a list of goods to buy. The list is long but your finances are limited. You start making decisions: this one I will buy, that one I won't. In the shopping list example, the purpose could be to keep your better half happy by saving money the most while at the same time maximizing the nutrition value of your cart. So a purpose, or a goal, may be formulated as optimization with respect to one or more criteria. In the latter case, the final choice of an acceptable solution is made out of the so-called Pareto set of solutions where no solution is better than the other in all respects. The final choice of a single optimal solution in this case is down to the agent.

Fig.1. Optimization criteria (The poster featuring the oligarch Abramovich reads: "I am not rich enough to buy cheap stuff"). Here the criterion is reliability and, in the long run implicitly, financial savings.

As we can see, decision making is the prerogative of experts. However, it can be delegated to the 'automatics' or a program. Examples are many: programmed tools on the factory floor, Windows XP and other operating systems, protein synthesis, nanotechnologies (molecular logic gates, processors on crystals). But even if the system under investigation makes decisions automatically, informally speaking, the agent must have pre-programmed these decisions using forethought. More formally speaking, the expert determines the heuristic process of finding solutions (e.g. by writing code). By doing so, he effectively directs the future search towards those areas of solution space where, as he thinks from previous experience, the density of solutions is higher than average. The expert (explicitly or implicitly) transmits to the controlled system the so called active information [1] about the desired characteristics of the search and/or of the solutions.

Experts are necessary for code to appear. There are no other known ways of generating code as meaningful computationally halting sets of instructions. Why is it so? Based on the given pragmatic criteria (such as saving money or raising nutrition value) a program can choose between states of the system that are physically (or chemically) indeterminate equilibrium states. For this reason a choice between such states cannot be made by the environment without recourse to an agent. Therefore a program (and a programmer) is indispensable.

It is easy to see that decision making is not necessarily conscious as far as the system under investigation is concerned. This is true wherever the action of decision making has been pre-programmed in some way (explicitly or implicitly). Under specific conditions, we can legitimately say that it is highly probable that such systems must have been programmed to make decisions by an agent with the capability of forethought. Moreover, sometimes this probability is so close to 1 that in practical terms we can be sure that such a system as a whole is a result of intelligent actions on the part of an agent.

From the point of view of the theory of computation complexity, for a program to appear there must exist a non-algorithmic oracle, i.e. an agent. This, I think, is a consequence of a fundamental theoretic result, i.e. the intractability of the halting problem for a Turing machine. The Turing machine is a mathematical model of an algorithm. Its halting models the successful completion of the respective algorithm. It is debatable whether a program can randomly mutate and still remain functional, when and within what limits it is possible and what is necessary to ensure in practice that the program remains sound and halts. However, it is beyond any reasonable doubt that the first program that was coding up the behaviour of a living organism must have arisen exclusively intelligently.

Using objective metrics for what is called functional information, ID works out statistical lower bounds for functional information at or beyond which it is possible to assert that the dominant factor in the generation of these quantities of functional information is intelligence of a decision maker, whereas stochastic factors or factors of law-like necessity (i.e. physical/chemical regularities), though present, are much less significant. Note the statistical and abductive nature of this statement. As the amount of functional information in a system increases, the probability of this system being a result of purposive decision making increases. Likewise, the probability of the dominant role of stochastic perturbations and law-like necessity in generating this system decreases. Abduction here is understood in terms of inferring to the best plausible explanation based on massive observations of functional information in various systems. While high statistically significant levels of functional information are systematically observed in man-made or animal-made artefacts (such as human or animal vocal or sign languages, computer languages and other complex information processing systems), only trace levels of functional information can be observed in non-living nature. From a practical point of view, in cases where the amount of observed functional information is beyond a certain level (for different systems these thresholds are different) the influence of stochastic factors or factors of law-like regularity can be safely disregarded as noise.

Now concerning living organisms. Perhaps the most important observation of biology, biochemistry, bioinformatics, mathematics and cybernetics, i.e. the study of control and controlled systems, over the recent decades has been that all terrestrial life is program-based, or cybernetic

Living systems control their own state automatically. Biosystems are implemented so that decision making in them is pre-programmed, e.g. in the form of instincts. Instincts are programs of action implemented at various levels: cellular, tissular, at the level of separate organs or systems, and at the top level of the entire organism. One of many examples of automated decision making in organisms is bacterial chemotaxis. In addition to instinctual reactions, the human higher nervous activity enables conscious decision making.

  1. Winston Ewert, George Montanez, William A. Dembski, Robert J. Marks II, "Efficient Per Query Information Extraction from a Hamming Oracle," Proceedings of the the 42nd Meeting of the Southeastern Symposium on System Theory, IEEE, University of Texas at Tyler, March 7-9, 2010, pp.290-297.

No comments:

Post a Comment