Chapter IV -- 18
Let us now summarize
In artificial intelligence, St Thomas has provided the researcher with an algorithm which describes the determination and execution of an action by a robot. He has already also given his reader a considerable measure of advice on how to organize the robot.
The reader might think that we presented this excursion into artificial intelligence with our tongue in cheek. No. We think that a researcher into artificial intelligence would profit greatly by studying St Thomas’ theory of human action. Moreover, the design specifications for a robot are nothing more or less than a formal model of human action, which essentially is what St Thomas has set out to provide. And ‘writing out’ the design specifications of a robot, or even building the robot, are a test not only of our understanding of the formal model of human action, but also of the consistency and completeness of the formal model itself. For we see that in St Thomas’ model of human action there is a problem with the reciprocal interdependence of the intellect and the will: how things get started. Of course, we ourselves are reserved whether such formal models do justice to the reality of human psychology, especially mystical psychology. However, despite the ambiguities and difficulties that his model does have, St Thomas has provided his reader with a very detailed formal model of how it is that humans come to determine and to do a human action.
While we did not present our excursion into artificial intelligence with tongue in cheek, we think, quite soberly, that it can be construed to be the reductio ad absurdum of St Thomas’ approach. St Thomas’ training in logic has led him to construct a model of human psychology that is, as they say in mathematics, very ‘tight’: formally precise, rigorously logical, without gaps or loopholes, logically consistent and complete. These logical requirements on a model are precisely what is needed by the researcher into artificial intelligence so that he can use the model to design a robot that actually works, but we have strong reservations whether such a robot more than mimics human behaviour, and, moreover, we have very strong reservations about the validity in human psychology of this rationalistic aspect of St Thomas’ psychology. While studying St Thomas forces you to think about the issues and to clarify your own thinking about human action, surely human psychology is not a matter of such rigorously formal logical models. In this St Augustine was by far the better psychologist.
Earlier in this work, in Chapter I, we remarked that one problem with the use of algorithms to represent the human mind (nous) was whether a priori the researcher limited the powers of the mind (nous) to those which easily fit the theory of algorithms. While St Thomas surely knew nothing about algorithms, which in any event are simply mechanical procedures, we find this problem even in his own work. For from the beginning, following Aristotle, whom we doubt was as rationalistic as his disciple, St Thomas, St Thomas has restricted human actions to those which are constituted both of a determinate end, over which the man has rationally deliberated and which he has consciously adopted, and of the determinate means to the end, over which the man has rationally deliberated and which he has consciously chosen. Other actions of a man are not human actions. A spontaneous act of love, of charity or of compassion is not a human action.
A. Donagan, in his article, ‘Thomas Aquinas on human action’, Chapter 33 of The Cambridge History of Later Medieval Philosophy, writes as follows:
As for the scarcely less common objection that the elicited acts which according to Aquinas are components of even the simplest complete human acts are too numerous to be credible, and correspond to nothing in our experience of our own acts, the reply—made by action theorists today—must be that the components of simple human acts are ascertained, not by introspecting what happens when we perform them, but by examining various cases in which the act is begun but not completed.…[1]
We challenge this.
Implicit in the method that Donagan is espousing of examining interrupted actions is the notion that there is an underlying procedure for coming to do human actions that the human mind follows, the steps of which can be elucidated by considering cases of interrupted actions. Were we to consider a multitude of interrupted actions, it seems, we would be able to discover all the detailed steps of the underlying procedure of coming to do a human action. That would be true if the mind of man worked by algorithms. But that is precisely what is in question. It is an a priori assumption—perhaps even unspoken and unconscious—on the part of the action theorist.
Let us clarify the connection between human actions and algorithms.
First let us remark that ‘procedure’, ‘mechanical procedure’ and ‘algorithm’ are synonymous.
There are two levels at which a human action can be seen or described in terms of algorithms. The first is the actual process of coming to settle on and to do a human action: St Thomas has provided us with an algorithm that we ourselves have just outlined above: consideration of a possible end, adoption of the end, enjoyment of the end, deliberation over the possible means to the end, consent to the good in all the possible means, choice of the means, intention, command and use. This is human action seen from the point of view of a fixed procedure that humans supposedly follow in deciding upon and doing a human action. Donagan himself has a chart, which clearly manifests, at the level we are discussing, the algorithmic structure of the Thomist model of human action, as Donagan understands it surely.[2]
There is another level at which human action can be seen as an algorithm or mechanical procedure. This is the level of the actual concrete actions that a man comes to do as a result of following the procedure, for which St Thomas has provided the model, of coming to settle on and to do a human action. That is, once I adopt the end and the means to the end, I come to do a sequence of actual, concrete actions. This sequence of actions can also itself be seen as an algorithm. Although we doubt that Donagan was familiar with the theory of algorithms, his section ‘Complete acts and component acts’ strongly hints, although probably inadvertently, at this level of algorithmic analysis. For he says:
Hence many human acts must be complex, in the sense of having other complete human acts as components. … But what may be called organised acts, consisting in a certain number of different complete human acts, the performance of which in a certain temporal order … is designed to bring about a certain result, are more common and more important. Making a cake and tying a complicated knot are homely examples.[3]
But baking a cake—following a recipe—is the classic example of an algorithm.
In discussing the applicability of the concept of algorithm to human action, there are therefore two levels at which the matter must be approached. Let us start with the second level, that of a discrete sequence of actions such as following a recipe to bake a cake. It is true that men often perform such discrete sequences of actions. It is often true that they do so when they are learning a complex action, especially one involving the body: learning to dance and learning to practise the martial arts are examples of the sort of training where one blindly—mechanically—follows a discrete sequence of steps repetitively until one has learned them.
However, when the man has learned the sequence of steps, and the music starts, or the fight is on, is he still following a mechanical procedure, an algorithm? No. This is a matter of human psychology, one that only someone who has learned to dance or learned to practise the martial arts can grasp. Imposing the view that the man has internalized a mechanical procedure or algorithm, so that he practises it ‘automatically’ or ‘by memory’ is the a priori imposition of an algorithmic model on what actually happens psychologically when one learns to dance or to practise the martial arts—or indeed to do any of a multitude of other such actions, even cognitive, such as playing chess. The logical analysis of a sequence of actions that the man does is not equivalent to an analysis of how the man learns the sequence of actions psychologically and even repeats them once he has thoroughly learned them. More is involved: concentration, attention, psychological bonding to the environment, a sense of rhythm (even in the martial arts), non-verbal communication with the partner or opponent (even in a genuine fight), understanding (there is even a non-rational conscious understanding of the body and its movements). A champion in the martial arts once remarked to us that fighting a match was something between Baryshnikov (the great Russian ballet-dancer) and Muhammad Ali (the great American boxer). Surely this is more than a matter of an internalized mechanical procedure. Moreover, the logical analysis of a complex action—which can even produce a sequence of simple actions that the man should do; this is the domain of industrial engineering—can bewitch us that we have uncovered the psychological structure of performing a complex action. But this is an unwarranted assumption. It has not to be proved so much as to be recognized as an assumption that determines our understanding of what it means for us to do a complex action.
It is as we said in Chapter I with the chess master: it is not that his mind is a better computer than the minds of all his opponents, so that it looks ahead more moves than the minds of his opponents, but that the chess master understands the chess game better than his opponents.
It is the same with the martial arts champion: his mind is not a faster computer, his body a better robot, than the minds and bodies of his opponents: he has not internalized a program for unarmed combat that he exercises with greater accuracy and rapidity than all his opponents. He has certainly been trained in the martial arts, but he is also more alert, more attentive, more understanding than his opponents. A Japanese karate master once remarked in our presence concerning an old karate master: ‘He is slow but he beats all his opponents.’
The situation is much the same in the case of the first level, that of the sort of procedure that St Thomas describes in his theory of action for coming to settle on and to do a human action. As we have pointed out, St Thomas’ model of such a decision process is algorithmic even though he most likely did not realize it. One can conduct a logical analysis, based on certain concepts, of how it is that a man comes to settle on and to do a human action. And Donagan has asserted that cases in which the decision process is interrupted can elucidate the structure of the decision process. However, the argument is circular. If the process of coming to settle on and to do a human action is algorithmic in nature, then examining cases of interrupted actions is a valid means to uncover the structure of that algorithm. But that is an assumption, not a fact. For if the human mind works differently from an algorithm, examining cases of interrupted processes of coming to settle on and to do a human act with a view to elucidating the structure of the algorithm will only bewitch us that we have uncovered the inner, algorithmic structure of the human mind. Moreover, since such an elucidation can proceed arbitrarily closely—at the cost of greater and greater logical complexity, surely—to describing the supposed way in which the human mind works, such an approach becomes an argument that the human mind does indeed work by algorithms.
What is in issue is the nature of the human mind, what it means to be a man alive.
The danger is that a priori we will exclude from consideration aspects of the human mind that do not easily fit into an algorithmic model of how it is that humans come to settle on and to do a human action. We have already seen this in St Thomas’ rejection of most intuitive faculties of human cognition and in his limitation of human actions to those actions which are done for a rationally determined end over which the mind has rationally deliberated, which end is attained by the rational selection and execution of means.
Being alive and being human are not matters of algorithms.
The difference is that a man is a person. It is precisely in the consideration of the humanly free person in the context of St Thomas’ theory of the reciprocal interactions of the will and of the intellect in a man’s coming to settle on and do a human action that we see the fundamental problem in St Thomas’ approach. His approach is forced to treat the will and the intellect as so to speak separate ‘sub-persons’ subject to Aristotelian causal analysis, which ‘sub-persons’ together constitute the human person: there is no way to pass from such a causal analysis of the reciprocal relations of discrete faculties of the human soul to the rationally free human person taken as a whole in himself, as an actor, as an agent—as a person. It is precisely here that the difficulties in the reciprocal interactions of the will and the intellect arise which lead St Thomas to posit the residual influence of the one faculty on the other. Since a robot is a machine, it can easily be designed to have separate, independent, discrete subsystems. Such a machine can then mimic a man’s behaviour. There is no question of a unitary consciousness in a robot; no question of a mind (nous), no question of a person, in such a machine. Hence, St Thomas’ analysis is useful for the researcher into artificial intelligence. However, to the extent that the human person is a ‘singularity’, an image of God, and not the mere composite of separate, independent, discrete faculties of the human soul subject to Aristotelian causal analysis, which faculties together produce the behaviour we call human, then the human person cannot be decomposed into the reciprocal interactions of the will and the intellect. It might be objected that St Thomas emphasizes the unity of the human soul. But, as we have seen, St Thomas’ analysis of human action proceeds on the basis of handling the faculties of the human soul as separate logical entities subject to Aristotelian causal analysis.
We can see in St Thomas’ treatment of human action not only his rationalism, but also his naturalism. We find it hard to convey this naturalism with the same precision with which we have just conveyed St Thomas’ rationalism. However, this naturalism of St Thomas is just as important as his rationalism, not only for an understanding of his system taken in and of itself, but also for an understanding of the temper of Roman Catholicism. For we do not doubt that it as Fr McBrien says in the quotation from Catholicism with which we began this chapter: ‘No theologian in the entire history of the [Roman Catholic] Church has had such a decisive impact on Catholic thought and the shaping of the Catholic tradition as St Thomas Aquinas.’
This naturalism finds its expression in the emphasis on the naturalness of the passions when they are subject to the control of reason. We will see this just below when we discuss St Thomas’ doctrine of the passions.
This naturalism also finds its expression in the doctrine of natural law that St Thomas advances in the Summa, although that doctrine is largely beyond the scope of this work. As one might expect, the natural law is equated by St Thomas with conformance to reason; however, there is a very great emphasis in St Thomas’ doctrine of the natural law on the natural impulses of man, more so than in the ascetical tradition of the East that arose with the Egyptian Fathers. St Thomas’ position is what he himself asserts in his discussion of the passions that we will look at just below: the passions are good, not bad, precisely to the extent that they are subject to the command of reason. Here Fr Copleston helps us to grasp what St Thomas means:
The natural law is expressed passively in man’s natural inclinations, while it is promulgated by the light of reason reflecting on those inclinations, so that inasmuch as every man naturally possesses the inclinations to the end of man and possesses also the light of reason, the eternal law is sufficiently promulgated for every man.[4]
Two sayings of St John of Sinai in the Ladder of Divine Ascent seem apposite here: ‘The monastic life is the forcing of nature so as to attain to that which is above nature,’ and, ‘The angel is the light of the monk; the monk of the layman.’ There is a different orientation in the East from the West. The East is oriented to the ‘true light that enlightens every man that comes into the world’;[5] the West, to the grace that builds on nature.
There is nothing of the naturalism of the Summa in the ascetical doctrine of Evagrius Pontikos that we will study in Volume II, nothing in the ascetical doctrine of St Mark the Ascetic that we will refer to in Volume III, nothing in the ascetical and mystical synthesis of St Hesychios that forms the subject of Volume III. St Thomas’ system has a naturalism that is not found in the ascetical tradition of the Orthodox Church. We think that a reading of Plotinus throws the difference into relief: while it would be excessive to suggest that Orthodox asceticism is Neoplatonist, Orthodox asceticism is certainly far more Platonic in its anthropology as concerns the nature of the passions and the proper task of man in regard to them than is the doctrine of St Thomas in the Summa. In this, we think that St Thomas introduced a sea change into the anthropology of the Roman Catholic Church, a sea change which is precisely the introduction of an Aristotelianizing psychology in place of the previous Platonizing psychology of the Roman Catholic Church.
It might be objected that we are comparing apples and oranges, that the Roman Catholic ascetical tradition recognizes those aspects of asceticism that we are emphasizing over and against the naturalism of St Thomas. To a certain extent this is true. However, we consider that notwithstanding that there is a different orientation between the Orthodox Church and the Roman Catholic Church on the matter.
previous | Table of Contents | next
[1] Medieval p. 654.
[2] Ibid. Figure 1, p. 653.
[3] Ibid. p. 647.
[4] Copleston Volume II, p. 409.
[5] John 1, 9.
0 Comments:
Post a Comment
<< Home