Rationality and Intentional Systems

By: Aaron Schwartz
Submitted: 2007-01-17 16:25:24
Print this article | Tell a friend | For publisher |

Rating:

 

In what way can we predict the behavior and reaction of some system, its interaction with the environment and response to our actions? What are the mechanisms of such prediction and what are the properties common for such systems? These are some of the questions that Daniel Dennett is trying to answer in his work “Intentional systems”.

First of all let us define the terminology. “Intentional systems”, according to Dennett, are the systems the behavior of which “can be - at least sometimes - explained and predicted by relying on ascriptions to the system of beliefs and desires (and hopes, fears, intentions, hunches, . . .) “ The concept of rationality and intentional systems, by Dennett, is based on the suggestion that analyzing the behavior of a certain system we adopt three stances. The first of them is the design stance – in suggestion that we know how the system is designed, we can analyze this design and predict the behavior of the system in practically any situation. This stance is highly applicable to the mechanical systems and other systems, designed by people. There is a natural question: what exactly do we perceive as a “design”? In a most common understanding this term means functional parts of a system, we suggest that we know how the small components of a system work and therefore, using the design stance, we can predict the behavior of the whole system.

The second stance we may use is the physical stance. Our experience and experience previously gained by other people gives us some knowledge about natural laws; and therefore we can make conclusions about the behavior and condition of the system, basing on our knowledge about its physical condition and previous experience. When do we use the physical stance rather than design stance? First of all, when it’s difficult or impossible to know the design of a system, or when it is too difficult to analyze in terms of design stance. Secondly, physical stance may be used in the situation of a breakdown or a system malfunction. In this case the design stance is mostly inapplicable, for in terms of design stance we know how the system will behave itself in “normal” situation, but we seldom know how it will act in a situation when some part of her will not be functioning, and moreover, it’s 99% that we won’t be able to make a diagnosis of the breakdown in terms of design stance – but we will be able to do this if we adopt physical stance.

And thirdly, there is the intentional stance (which is naturally made for what we call “intentional system”). This stance appears when the system is too complex to be analyzed using both design and physical stances; in this case we can try to predict the behavior of the system if we ascribe some goals to it and make the suggestion that the system is acting optional (or designed optional) to reach these goals. This stance is rather relative because it implies a lot of relative suggestions. First of all, we ascribe the goals and strategy that is common only for human beings to the regarded system. Strictly speaking, we don’t have more information on other human beings’ motives than on any regarded system’s motives – so we can only ascribe our own properties of thinking and acting to the system. This ascription can be criticized and doubted, because of its high level of relativity, but there are some global laws that are likely to be common for such systems.

The ascription of intentions to a certain system implies the suggestion that this system is rational, i.e. that it has
1) certain input information about the situation
2) the system has certain goals
3) it is optimally designed to reach these goals, or, in other words:
“What is the most rational thing for the computer to do, given goals x,y,z,..., constraints a,b,c,... and information (including misinformation, if any) about the present state of affairs p,q,r, ...?” The above-mentioned understanding of rationality is good for mechanical systems, but it is not appropriate for people and natural systems. But if we call constraints and current information as “beliefs”, and the goals of the system as “desires”, we obtain the appropriate understanding of rationality for systems other than mechanical.

Why is the assumption that something is an intentional system the assumption that it is rational? Let us suggest that the system is not rational. Then its actions are defined neither by the input conditions, nor by the current situation, nor by the suggested goals. In this case there is a possibility that the behavior of the system is not influenced by the environment at all, or is very slightly influenced. Therefore the behavior of the system can not be predicted with good probability – which contradicts with the suggestion that the system is intentional (i.e. predictable to some extent). We obtain that in order to be able to ascribe intentions to tee system, we need it to be rational.

We do not have to regard the question whether the system really has beliefs and goals, we only use the suggestion that it might possess such qualities, and we can make predictions on the system’s behavior based on this suggestion. It doesn’t matter whether the system really has the above-mentioned qualities or does not.

Intuition tells us that there have to exist some laws or at least common rules for rational and intentional systems. Indeed, there are a few suggestions, that, though they are to some extent transferring human qualities on the system, prove to be true for all intentional systems. In suggestion that the system has certain beliefs and desires we only ascribe to it categories of rationality, perception (in other words, ability to get information) and ability to react and to act in general. In my opinion, this transfer is not narrowing the space of regarded systems; these properties are consequences of the definition of intentional system and behavior. D. Dennett says on this purpose: “ascriptions of beliefs and desires must be interdependent, and the only points of anchorage are the demonstrable needs for survival, the regularities of behavior, and the assumption, grounded in faith in natural selection, of optimal design”.

Let us analyze the terms “beliefs” and “goals” in understanding of Dennett’s theory. If a creature or agent believes into the logical rules, it is supposed to follow these rules. A perfectly rational system has all beliefs that are “true”. As in real situation we cannot interact with a perfectly rational system, some of the beliefs of the regarded agent are “not true” (the more such beliefs are, tee less rational system actually is). What is, in terms of this theory, a totally irrational agent? It is the system the beliefs of which are all “not true”. As was said earlier, the assumption of rationality and therefore intention can not be done for such a system.

Supposed that the system is rational, we can conduct that it possesses special functions and reactions, that are appropriate and useful (in relation to the system’s goals) in the environment, and answer the current circumstances.

Let us regard the connection between beliefs, desires and behavior. Sennett states on this: “For the concept of belief to find application, two conditions, we have seen, must be met: (1) In general, normally, more often than not, if x behoves p, p is true. (2) In general, normally, more often than not, if x avows that p, he believes p [and, by (1), p is true]. Were these conditions not met, we would not have rational, communicating systems; we would not have believers or belief-avowers. The norm for belief is evidential well-foundedness (assuring truth in the long run), and the norm for avowal of belief is accuracy (which includes sincerity).”

There appears the “phenomenon of incorrigibility” in this theory as well: the incorrigibility between rationality and accuracy of belief. The problem is that the rationality (intentionality) of the system’s behavior is relative in the short-term examination – but it appears that in the long-term examination beliefs and views of a system are the basis of its actions; and though they depend on the environment, they perform a back influence on the situation and are (in the long run) influencing it as well.

If an agent (a system) performs actions that do not correspond with the suggestion of rationality, this may happen because of 2 factors:
1) the suggestion on the system’s views and beliefs that we have made, is not appropriate with the environment, and does not correspond to the situation;
2) the system is not intentional in our understanding.

We can surely make mistakes when trying to predict the system’s views and beliefs, but there are some factors and goals, the absence of which inevitable means that the system is not intentional in our understanding. First of all, intentional system is not supposed to self-destruction. Secondly, the system is tending to survive and interacts with the environment according to its goals. If the system does not correspond to these suggestions, it can not be regarded as intentional; so systematically irrational agents (in the current understanding of rationality) do not have an intentional behavior.

How to find out that the set of beliefs is not contradictional and the system that possesses the set of beliefs can be called and treated as intentional? Sennett recalls J. Vickers on this topic: “Consider a set T of transformations that take beliefs into beliefs. The problem is to determine the set Ts for each intentional system S, so that if we know that S believes p, we will be able to determine other things that S believes by seeing what the transformations of p are for Ts. If S were ideally rational, every valid transformation would be in Ts; S would believe every logical consequence of every belief (and, ideally, S would have no false beliefs). Now we know that no actual intentional system will be ideally rational; so we must suppose any actual system will have a T with less in it. But we also know that, to qualify as an intentional system at all, S must have a T with some integrity; T cannot be empty.”

To make a conclusion, intentional theory is meant to connect what we used to consider “rational” and what we used to consider “irrational”. Realization of these questions is very important for the attempts to analyze the processes of interaction between intelligent systems and understand the mechanisms of such interaction; with the help of these statements we can distinguish whether we can predict the behavior of a particular agent.

Custom written research papers

Custom Essays

research paper

Article source: Expert Articles