Daniel Dennett

Intentional Systems

The three stances of intentional systems and their prerequisites

Analytical philosophers often contemplate the question of what is going on in our mind. Do special states in our brain correlate one to one to physical states or is this nonsense and there is no such thing as mental processes but mere behaviour? In “Intentional Systems” Daniel C. Dennett wants to show that this is not the essential issue. It is more important to deal with shown behaviour – so to explain and predict it. To possess these possibilities we only have to accept that we have somehow intentional states regardless of how they are realized in us, whether as a combination of neurons or in our fiction. Dennett therefore introduces the concept of an intentional system, which I will elucidate in the following. But first of all I have to explicate his theory of the three stances so that you get an idea what the prerequisites for such an intentional system are.

According to Dennett we can describe and predict the behaviour of organisms or other complex systems such as computers with the help of three different stances.

The lowest abstract level is the physical stance. In this stance you explain and forecast the behaviour of systems on the basis of their physical and chemical components. You know what structure and properties they have. So you are concerned with things like weight, velocity or chemical molecules. Moreover you know that they obey the laws of nature. Taking all these facts into consideration you should be able to make good predictions about what behaviour will be shown, if their components interact. You can use this stance for example to foretell how long a candle will need to burn down or that it will running out, if you expose it to the wind.

Only in this stance it is possible to detect malfunctions of a system, because you can detect mistakes in the structure. That leads to a very important prerequisite for the next two stances – the optimal design. If you want to predict something on given rules, a malfunction will falsify the result and will make it impossible to forecast anything, respectively. I will go into that more deeply later on.

Sometimes it is not very convenient to take the physical stance. There are many complex systems like organisms and computers, which are too complicated to get to know all the important facts you need to predict what they are going to do next. You should rather take a look how the computer or organism works. The design stance deals with this level of function. It helps you to analyse its components functionally and is hence concerned with the design of the system and its purpose. For instance if you predict a bird to fly, because you see how it stretches its wings, you do that on the basis that wings are made for flying. To sum up, with the functional stance you are able to forecast behaviour only with the knowledge of the function of parts of a system. You do not need to know what they exactly are made of. But the predictions will only succeed, if all parts function as they should.

Mostly, living organisms and highly developed computers are even too complex for the functional stance. It would take too much time to discover all their functions and take them into consideration. Thus, if you really want to have a good behaviour prediction of such complex systems, you have to take the third stance, the intentional stance. You have to consider that they have certain information like beliefs and desires and that they act towards goals. In addition you presume that these systems show rational behaviour. If you take these facts into account, you will be able to explain and predict the behaviour of complex organisms and other systems. To illustrate that once again with the help of an example, you can imagine a mouse that runs away because it knows that the cat is coming. You know the mouse knows that the cat is its enemy and that the mouse has the desire to survive and so you can hence conclude that it want to escape from the cat and will run away. So this stance is an enhancement to the functional stance. It considers not only the functions of a system but also their intents and ways of acting towards personal goals from which you can infer how they will behave in their environment.

Now that we know what the three stances are, we can have a closer look at what Dennett calls an intentional system. He claims that everyone and everything has intentional states, if their behaviour is predictable with the help of the intentional stance. This is true for nearly all living organisms and also for many computers. If their behaviour is reliably explainable, we can describe them as intentional systems. But they are only intentional systems in relation to such an observer who makes predictions, so in relation to somebody who treats a system as an intentional one. That is very important. An organism or computer is not for himself an intentional system.

Now the question comes up again what prerequisites we need to treat something as an intentional system. I have already interspersed some facts due to that point, but I want to summarize them now explicitly and discuss them a little bit further. As we have seen in the first stance, there is no problem which we are concerned with. If something wrong happens, we are able to find the mistake in the basic physical structure of the system we deal with. In the second and especially in the third stance we have to assume some things for that we can make true predictions.

First of all the system has to act rational. It must have the beliefs which it should have in its environment and personal situation. Otherwise we have no general rules from which we could conclude anything. Furthermore it has to have the desires which it should have according to its needs. To sum up briefly, the system behaves rational, if it is acting towards its desires and beliefs in a rational way. That rationality gives us our basis to estimate what the system will do next. Without rationality there would be no rule, which we could take into consideration and hence we would be completely clueless as to how the system could perhaps behave. To give an example: If you play chess against a computer, you act on the assumption that it makes its move straight forward to the victory and that it does not set its figures randomly onto the chess board. In the last case you have no possibility to predict which move it will make next. Therefore it will be very difficult to plan strategies, if you do not know at any time what your opponent’s intention and goals are in this game.

Secondly, we have to assume that the system works properly as it was constructed. If there is a malfunction our predictions will turn out false in all likelihood. So we have to presuppose optimal design as well as rationality. Concerning living beings the optimal design results from evolution. With the help of natural selection only those creatures should have survived and reproduced, which were better adapted to their environment than others. As a result animals and humans today should show the optimal designed behaviour towards their environment. So we can expect a limited number of behaviours in a special situation because of the optimal adaptation of our observed organism to the environment.

I want to conclude once more. We can describe complex systems as intentional systems, if we can explain and predict their behaviour with the help of the intentional stance. The more rational this system behaves, the more efficient our forecast will be. If this system was not optimal designed, it leads to malfunctions, which falsifies our evaluation of it, too.

In fact this theory is a bit idealized. Neither do living beings behave rational at any time nor are all systems optimal constructed. So this leads inherently to false prediction of other’s behaviour sometimes. But what is important is the pragmatic strategy we apply to other systems. We need to predict the behaviour of other organisms or even computers to understand their beliefs, desires and intentional goals so that we can react in an appropriate way. Thereby it is not important whether our beliefs and desires are somehow physically realized or if computers really have such feelings. It plainly is the basis for social interaction to be able to observe others in an intentional stance. For this goal one can accept that there are exceptions in reality and that sometimes one’s predictions will fail. But in most cases they succeed and that is the main thing.