CONSTRAINT, EMPOWERMENT, AND GUIDANCE:

A CONJECTURAL CLASSIFICATION OF LAWS OF NATURE

(Article first published in 2001 Philosophy 76, 341-70: link to online edition)

© The Royal Institute of Philosophy 2001. Reprinted with permission

DAVID HODGSON

This paper introduces a conjecture that laws of nature may be of different kinds, in particular that there may, in addition to laws which constrain outcomes (C-laws), be laws which empower systems to direct or select outcomes (E-laws) and laws which guide systems in such selections (G-laws). The paper defends this conjecture by suggesting that it is not excluded by anything we know, is plausible, and is potentially of great explanatory power.

…………………………………….

1 Introduction

It is widely accepted that there are regularities in our universe which accord with laws of nature; and that a primary goal of science is to discover these laws. It is also widely accepted that the hypotheses and explanations provided by science do approximate to these laws, with increasing accuracy and assurance. Sometimes it is suggested that the laws of nature are merely descriptive of regularities that just happen, rather than being caused to happen by the laws and initial conditions; but I believe this is a minority view, and in this essay I will proceed on the assumption that laws of nature are prescriptive or otherwise efficacious in bringing about or contributing to what occurs in the world.

Now, it seems to be assumed that laws of nature are all of one particular type, a type that I call laws of constraint, or C-laws: laws which do no more and no less than constrain or compel or circumscribe what actually happens. In this paper, I introduce a conjecture that there may be other classes of laws of nature, in particular two classes that I call laws of empowerment, or E-laws, and laws of guidance, or G-laws. E-laws are conjectured laws that empower or enable certain emergent systems (E-systems) to direct or select outcomes, within the constraints of C-laws; and G-laws are conjectured laws that provide guidance to E-systems, by grounding non-conclusive reasons on the basis of which they direct or select outcomes.

As will become clear, I take human beings to be the clearest case of E-systems, and conscious human choice to be the clearest case of something empowered by E-laws and guided by G-laws. But I will suggest that animals of other species on our Earth are also E-systems, and I do not rule out the possibility that there are or could be E-systems which are not animals or even living things.

I will begin my discussion by considering C-laws and what I see as their limitations, as regards both their constraint of outcomes and also their capacity to engage with features of the world.

I will argue that it is reasonable to accept that initial conditions and C-laws can leave open a range or spectrum of outcomes, and that only one of these outcomes ever actually occurs. This gives rise to the question, how is it settled which outcome occurs? I will refer to three possibilities: (1) that the particular outcome occurs randomly, or randomly within probability parameters established by the initial conditions and the C-laws; (2) that it is uniquely pre-determined by some constraint which is not a C-law; or (3) that it is selected on the basis of some guiding non-conclusive considerations - so that neither is the selection pre-determined by any constraint, nor is it random. I conjecture that the third possibility does obtain in some cases, in accordance with E-laws and G-laws.

Then I will elaborate on how I think E-laws and G-laws could operate. And I will go on to suggest that the conjecture that there are such laws is reasonable, and should be pursued. Although this conjecture is difficult to test, it is not excluded by anything we know, it is plausible, and it is potentially of enormous explanatory power.

Finally in this introduction, I note that the idea that laws of nature may be of different types has some analogy in legal theory.

The jurist Wesley Hohfeld drew distinctions between (1) laws that impose duties on some persons, and give correlative rights to others; (2) laws that allow liberties for some persons, and give others no-rights to interfere; and (3) laws that give powers to some persons to change duties, rights, etc., and subject others to liabilities to have their duties, rights, etc. changed. Laws of the first class, for example a law prohibiting theft, circumscribe what conduct is permitted, and thus may be compared to C-laws. Laws of the third class, for example a law empowering a judge to impose a term of imprisonment, specify what entities, in what circumstances, and to what extent, can change what is permitted for others, and thus may be compared to E-laws. And insofar as a judge has a discretion as to what penalty to impose, non-conclusive principles applied by the judge in making this decision may be compared to G-laws.

Another analogy for C-laws, E-laws, and G-laws is provided by the actual practice of courts of law in deciding what the legal result should be in particular cases. Established laws often do not specify a clear result for the facts of particular cases, so that courts have to apply non-conclusive methods and principles in order to reach a decision as to which result, out of those permitted by the established laws, should be adopted. The established laws can be likened to C-laws, the laws empowering courts to decide cases can be likened to E-laws, and the non-conclusive methods and principles can be likened to G-laws. I do not suggest that the likenesses here are absolute: for example, in common law systems the established laws are incrementally changed by decisions made by courts in individual cases, and I don’t suggest C-laws are changed in any similar way. But the analogy is still quite close and suggestive.

2 C-laws: the Laws of Constraint

It is generally assumed that the world changes over time in regular ways, in accordance with laws of nature that are in principle discoverable. Since the time of Newton, some people have believed that the laws of nature pre-determine unequivocally all changes that occur in the world - that the laws of nature always fix a single course of development from any particular state of the world, and a single outcome at any particular time thereafter.

Newton’s laws themselves suggested that, given any distribution in the whole universe of quantities and motions of masses and of other physical properties that affect and/or are affected by forces, then any later (or indeed earlier) distribution of such quantities and motions is unequivocally fixed by the given distribution and the laws of nature. This suggestion received some confirmation in the nineteenth century from Maxwell’s equations of electromagnetism, according to which electromagnetic forces are unequivocally determined by the distribution of quantities and motions of a physical property, namely electric charge; just as, according to Newton’s law of gravitation, the force of gravity in any circumstance is unequivocally determined by the distribution of another physical quantity, namely mass. And the suggestion received further confirmation in the early twentieth century from Einstein’s equations of relativity, which resolved an inconsistency between Newton’s and Maxwell’s equations, and gave a deeper explanation of gravity.

Plainly, the laws of nature involved in these theories are what I call C-laws - laws which constrain or compel or circumscribe what actually happens. And according to these theories, the C-laws constrain a single outcome in all circumstances. However, in the 1920s, quantum mechanics (QM) was devised; and QM suggested that outcomes are not always unequivocally fixed by initial conditions and laws of nature. Rather, QM suggested that, while a physical system may change over time in a way fixed by laws of nature so long as it remains unobserved, it will upon observation be found to be in just one of a range or spectrum of possible states; and the laws of QM gave probabilities for each of these possible states.

But the laws of QM are still C-laws: they constrain outcomes to the extent of limiting them to a range or spectrum of possibilities, with actual outcomes occurring randomly within probability parameters determined by the C-laws. And the invention of QM can in fact be seen as extending the domain of C-laws. QM showed how events at the atomic level could be explained by C-laws, while at the same time accounting for the apparent accuracy of Newtonian physics in the everyday world; and it also subsequently provided a basis for showing how chemistry can be explained in terms of the very same C-laws as physics. And in more recent times, the discovery of DNA has meant that biology also may be explained in terms of the same C-laws as physics and chemistry.

Thus it has become orthodox to think that there are a finite number of C-laws which together explain all changes that occur in the world, so far as they can be explained - indeed, perhaps just one basic C-law, that could be written on a T-shirt! There are opposing views,¹ to the effect that Nature and its laws are of infinite depth and complexity, but these are I think minority views. Consistently with the orthodox view, I will assume in this essay that all C-laws ultimately reduce to the laws of physics.

In accordance with this approach, I state my C-laws hypothesis as follows:

There are a finite number of C-laws such that, given any possible state of the universe (the initial conditions, which I call SU₁), in the region of time T₁, the state of the universe SU₂, in any later region of time T₂, will be constrained by SU₁ and the C-laws either to be one uniquely-determined state, or else to be within a range or spectrum of possible states.2

In other words, the initial conditions and C-laws will exclude the occurrence of all but a limited range or spectrum of outcomes, at the extreme excluding all but one single outcome.

3 Limitations of C-laws

My C-laws hypothesis leaves open a number of alternatives as to the ultimate determination of outcomes. One is that the initial conditions and C-laws uniquely determine a single outcome in all cases, consistently with the rule-based determinism of Newtonian physics. This now seems unlikely, in the light what QM tells us about the world, and I will not consider it further in this essay. Another is that where the initial conditions and C-laws do not determine a single outcome, but leave open a range or spectrum of possible outcomes, all such outcomes actually occur: this is suggested in particular by what has been called the many-worlds or many-minds interpretation of QM. I believe this interpretation of QM is untenable,³ and I will not consider it in this essay. In fact, I will adopt in this essay what I call the one-outcome hypothesis, as follows:

Where initial conditions and C-laws leave open a range or spectrum of possible outcomes, only one of such outcomes actually occurs.

On that approach, the C-laws have the limitation that they do not always uniquely determine outcomes: something more is required. And shortly I will discuss what this something more could be.

First however, I will look at another limitation of C-laws, namely a limitation as to the features of the world with which they can engage.

C-laws are by hypothesis general, and they must engage with features of any SU₁ so as to constrain features that SU₂ will have. Now some features of SU₁ may be considered causally efficacious in bringing about SU₂ without themselves having to engage with laws of nature, namely those features that remain constant or else are modified by the operation of laws of nature on other features; but causation through the operation of C-laws does require that all features of SU₁ which are causally efficacious in combination with C-laws in actually bringing about the transition to SU₂ must have a kind of generality, so that these features can engage with general laws.

The laws of physics, to which I have assumed all C-laws reduce, apply to a finite number of types of particles or waves or fields of matter and radiation, and to quantities of properties such as mass, electric charge, energy, momentum, spatial and temporal separation, direction, duration, and the like; so that the features of states and objects and events that can engage with C-laws, and thus be causally efficacious in combination with the C-laws in bringing about changes in the world, are just their general and quantitative features of that type. It would be fallacious to say that this implies that objects and events of physical systems are not themselves causally efficacious, because only mass, energy, and so on engage with C-laws; but my assumption does imply that states and objects and events have their causal efficacy, in bringing about changes in combination with C-laws, only through their features of mass, energy, and so on.

Nor does this mean that the features of SU₁ which are unique, in the sense that they have never occurred before and will never occur again, cannot be causally efficacious in bringing about changes in combination with C-laws. The laws of physics apply to any quantities of mass, energy, distance, direction, and so on, and to any combinations of these quantities; so the circumstance that any particular combination of these quantities in SU₁ is unique would not prevent the application of C-laws to these features of SU₁ so as to constrain the outcome SU₂. But this is precisely because the features in question are quantitative, in such a way as to engage general laws. And thus, if there were any features of SU₁ which were unique, but which were not quantitative in such a way as to engage general laws, then on this approach those features could not be causally efficacious in bringing about changes: all that could be efficacious in bringing about changes would be the general and quantitative features of the type I have mentioned.

The point can be illustrated by reference to the Game of Life. This game was devised in about 1970 by John Conway, a Cambridge mathematician. Its rules can be stated shortly:

Life occurs on a virtual [and potentially infinite] checkerboard. The squares are called cells. They are in one of two states: alive or dead. Each cell has eight possible neighbours, the cells of which touch its sides or corners.

If a cell on the checkerboard is alive, it will survive in the next time step (or generation) if there are either two or three neighbours also alive. It will die of overcrowding if there are more than three live neighbours, and it will die of exposure if there are fewer than two.

If a cell on the checkerboard is dead, it will remain dead unless exactly three of its eight neighbours are alive. In that case, the cell will be ‘born’ in the next generation.⁴

Given an initial configuration of live and dead cells, everything that happens in the game is determined unequivocally by these basic rules, which may be considered as analogous to the laws of physics.

Features on a larger scale than the one-cell-and-eight-neighbours scale dealt with by the rules may be causally efficacious in the sense that they are modified by the rules engaging with individual elements of these features, so as to produce further larger-scale features; but they do not themselves engage with the rules so as to bring about changes. This is true even of features that appear to unfold in accordance with larger-scale rules; since their causal efficacy in bringing about changes is only through their properties at the one-cell-and-eight-neighbours scale. For example, there is a five-cell pattern called a glider which, after four generations in which no other live cells are encountered, results in an identical pattern displaced diagonally by one cell.

The rules of the game dictate that the state of any cell in any generation is wholly determined by the state of that cell and the eight adjoining cells in the preceding generation; so the rules do not in fact engage with the glider pattern as such. Of course, the glider pattern is itself causally efficacious in the sense that it is modified by the rules engaging with individual elements of the pattern so as to produce further glider patterns; but there is nothing in the game itself that recognises or responds to a glider as such, either as cause or effect. An outside observer may recognise a glider, and may construct a rule that a glider will continually move diagonally across the checkerboard, progressing by one cell every four generations, unless it encounters any other live cells. This chunking of the five live cells of the glider, and their adjacent dead cells, provides an observer with a useful description of what is happening, and may help the observer to understand and predict the game’s unfolding; but the glider pattern itself is not causally efficacious in bringing about any changes, because efficacious causation of this kind is entirely at the one-cell-and-eight-neighbours scale.

It might be said that the rules of the Game of Life could be restated in such a way that the glider pattern engaged with the rules of the Game, and that according to this restatement the glider as a whole would be causally efficacious in bringing about changes. Of course, that argument would not apply to any pattern which, unlike the glider, was unsuitable for application of general rules because of its complexity and the rarity of its occurrence. And in any event, any rule in this restatement which concerned a larger-scale pattern such as a glider would have to be qualified so as to ensure that in no situation did it conflict with the two basic rules of the Game, and these two rules would have to be included in any event to cover situations not covered by rules concerning larger-scale patterns.

And in fact, however the rules were stated, their true effect would be that, if the five live cells of a glider were those numbered 1 to 5 in the above diagram, then the live state of cell 1 would have absolutely nothing to do with the state, in the next generation, of the cells numbered 3, 4 and 5, because it is not adjacent to them.

It would be different if, instead of just restating the rules of the Game of Life, one could modify them so that, for example, looking at the above diagram, the state of cell A in the next generation was not wholly determined by its eight neighbour cells, but was affected by the complete glider pattern. Suppose, for example, that cell A is dead and that the rules of the modified game are such that (1) if just three of A’s eight adjacent cells are alive, then it may or may not be alive in the next generation; and (2) in that event, it will be alive if and only if those three live cells are themselves part of a glider. Only by that kind of modification to the rules would a ‘gestalt’ such as the glider pattern be made truly efficacious in bringing about changes.

So, I have discussed two limitations of C-laws. The first of them gives rise to the question, how is it determined which one out of the range or spectrum of possible outcomes actually occurs? If, as suggested by standard QM, the single outcome occurs randomly, or randomly within probability parameters established by the initial conditions and the C-laws, this would mean, in terms of the second limitation of C-laws, that all the other features of the world, apart from those general and quantitative features that engage with C-laws, have no contribution to make in determining what changes occur in the world. If such features are to make some contribution in at least some cases, then in such cases either:

(1) the single outcome is somehow pre-determined by some constraint other than the initial conditions and the C-laws, a constraint which would have to be particular to the features in question, because if it were of general application it would itself qualify as a C-law; or

(2) the single outcome is neither random nor pre-determined, but is somehow selected by something and/or some process occurring in the initial conditions and/or during the time between the initial conditions and the occurrence of the single outcome.

Possibility (1) seems to have little to recommend it: since the constraint which pre-determines the outcome is not a general law of any kind, it seems that it would have to be arbitrary and incomprehensible. What I want to explore in this essay is possibility (2).

4 E-laws and G-laws

If, in accordance with possibility (2), there is sometimes a process of selection involved in the determination of a single outcome, this gives rise to two questions:

(1) What systems and/or processes make such selections?

(2) How do they make them?

If the selections are to be other than random or arbitrary, then they must be made by something or some process on the basis of some kind of reasons; yet these reasons cannot involve rules or laws that pre-determine the outcome of the selection, because then the outcome would be fixed by C-laws. And it seems plausible that any selection that a system can make will take effect through a selection of some future state of itself. So what I suggest is that laws of nature are not confined to C-laws, but extend to the kinds of laws proposed in the following hypothesis, which I call my E-laws and G-laws hypothesis:

Laws of nature include E-laws, which specify what kinds of systems are emergent systems (E-systems) having power to make selections as to their own future states; and also specify what kinds of selections E-systems can make in what kinds of circumstances, and what kinds of reasons E-systems have for making the selections; and

Laws of nature also include G-laws, which guide but do not constrain E-systems in making such selections.

In the remainder of this essay, I will argue that this hypothesis does not conflict with anything we know, is very plausible, and is potentially of enormous explanatory power.

E-laws could specify that a system with features of the kinds A(x₁, x₂, x₃, …) in circumstances of the kinds B(y₁, y₂, y₃, …) (all being part of initial conditions SU₁):

(1) has power to make a difference, in respects of the kinds X(a₁, a₂, a₃, …), in relation to its own future states, within the range of possible outcomes permitted by SU₁ and C-laws;

(2) knows-how to do so;

(3) has reasons of the kinds Y(b₁, b₂, b₃, …) relevant to the exercise of the power, reasons which generally consist of information and/or feelings, and which the system can grasp as a whole;

(4) where there are conflicts within the reasons as to how the power should be exercised, can resolve those conflicts in exercising the power.

The E-laws would not be enough to enable reasoned selection of an outcome in cases where reasons conflicted, unless they included or were associated with principles guiding the resolution of these conflicts. Just as judges do not toss coins to decide cases within the leeways left by established laws, but rather proceed in accordance with non-conclusive but guiding methods and principles, so also E-systems would need something similar: otherwise the outcome would be merely arbitrary. Accordingly, I postulate G-laws, which might possibly be considered either as an important sub-class of E-laws or alternatively as a distinct class of laws. G-laws must have generality, so as not to be arbitrary, yet must also be non-conclusive, because if they were conclusive they would themselves be C-laws, and the E-laws would be pointless.

In order to explain what I think G-laws could be, and how they could work, I will start by looking at a system that is not quite an E-system, and discussing what changes might convert it to a primitive E-system. Such a system would be constrained in its development by C-laws, and where the C-laws left open alternative future states of the system, the state which actually occurred would occur randomly within probability parameters established by the C-laws. From the point of view of the system itself, if and in so far as it had one, any outcome required by the C-laws would simply be the outcome to which the system was disposed or ‘motivated’; and it would occur ‘without question’. Where there was some fuzziness or internal conflict in the system’s disposition or motivation, the system would have no capacity and no reason to prefer one outcome over any other, except to the extent that it was favoured by the probability parameters established by the C-laws; and the single outcome would occur randomly within those parameters.

Now suppose an otherwise similar system did have some marginal capacity and reason to prefer and bring about one future state of itself, among those left open by the C-laws, over another such state, in circumstances where there was fuzziness or conflict in its disposition or motivation as to what state should occur. For example, suppose that it felt, in a primitive way, something like what we would feel as motivation to minimise the pain of an injury, and also something like what we would feel as motivation to satisfy hunger; and that it felt it could follow one feeling or the other, but that following one would preclude following the other (getting the food would exacerbate the pain). Or suppose that what it felt could be described as an urge to mate, and also a conflicting fear of severe injury (let’s say the system is a small male spider, faced with a possibility of mating with a much larger female spider which tends to kill or dismember nearby males). The system, having these conflicting feelings, and feeling itself motivated by them towards differing future states of itself, both of which were open to selection by it, could I suggest also feel something like a requirement to resolve them ‘rightly’, and to bring about one state of itself (that is, to act) in accordance with that resolution.

This suggests the most basic G-law, which would to some degree guide and be felt by even such a primitive E-system, a law which I call ‘act rightly’:

Act so as ‘rightly’ to resolve fuzziness or conflicts of motivation.

I use ‘rightly’ at this stage without any moral implications, so that the law here simply means, do whichever of the conflicting possibilities is apt or fitting or appropriate or ‘to be done’. I say this law would be felt, because its guidance would, to some extent at least, take effect through its influence being felt and acted upon by the system itself. So I suggest that the system would feel a requirement to select the alternative which appeared to it to be appropriate, and that it could thereby produce a result neither determined by C-laws nor simply random within the constraints of C-laws.

Of course, I recognise the possibility that apparent conflicts of motivation like this may in fact be resolved by computation-like processes which proceed in accordance with C-laws; but what I’m suggesting is that an alternative possibility should also be recognised and considered. And I’m aware of the view expressed by Hume to the effect that action is determined by preponderance of desire; but I believe it would be a serious fallacy to assume uncritically that desires are commensurable and capable of determining outcomes in the same way as forces in Newtonian physics.⁵ I do accept that, on the approach I am suggesting, there would have to be some kind of primitive judgment made by the system as to which motivating feeling was to be given effect to in preference to any conflicting feeling, a judgment which would be indeterministic and therefore fallible, and I am unable at present to suggest any further analysis or explanation of such a judgment. Indeed, as will become apparent, I contend that even human plausible reasoning involves primitive judgments of that general kind, although clearly in that kind of reasoning the judgments can be subject to checking and testing, which can be the subject of analyses and explanations.

In more complex E-systems, the basic G-law could come to separate out into two distinct aspects or sub-laws, which I call ‘decide rightly’ and ‘carry out’:

Decide what act would rightly resolve fuzziness or conflicts of motivation; and

Carry out that decision.

That is, there could arise the possibility that the system would not act in accordance with a judgment of rightness which it had made. And in such systems, selections could be assisted by further G-laws and/or by principles associated with them. In particular, these systems could feel and apply a G-law, which I call ‘find out’:

Optimise the reasons (including information and feelings) on the basis of which to act.

In moderately primitive E-systems, this could be felt as requiring attention to relevant information-and-feelings, as delivered by the senses and emotions. In more elaborate E-systems, it could be felt as requiring such things as exploration of relevant information-and-feelings, appropriate verification by checking, looking for coherence and consistency, attending to analogies (relevant similarities and differences as compared with other information-and-feelings), and seeking a grasp or understanding of issues facing the system. Some of these steps, such as attending to analogies, could involve primitive judgments as to what factors are to be preferred over other factors, of the same general type as referred to above.

Even in E-systems with the self-conscious rationality of human beings, the application of G-laws would be largely instinctive, a matter of feelings and knowing-how rather than consciously following explicitly-known rules or principles. Again looking at one of my legal analogies, different judges appear to apply similar methods and principles in deciding cases where there is no clear precedent, yet they cannot state explicitly what those methods and principles are: it is very difficult indeed to set out principles for determining which is the most appropriate analogy or which new rule would ‘fit best’ with existing rules. Indeed, the same is true of human plausible reasoning generally: as I will discuss shortly, no one has ever been able to set out explicitly principles and rules which ground satisfactory plausible reasoning.

In E-systems without the self-conscious rationality of human beings, the application of G-laws would not be a matter of morality, although analogies with human moral issues could be drawn. For example, some conflicts of motivation could be analogous to human moral conflicts, such as a conflict between an animal’s motivation to minimise its own pain and its motivation to protect its offspring; and some actions by animals may display ‘virtues’ of courage and determination in carrying out decisions as to what act would be ‘right’.

However, in E-systems with self-conscious rationality, G-laws could have central moral significance. The basic G-law ‘act rightly’, as it applies to these systems, could be the fundamental moral prescription. Moral laws such as ‘do no harm’, ‘be fair’, ‘be honest’, ‘fulfil commitments’, and ‘do good’ could be further G-laws felt by these systems. Plainly, these moral laws could conflict with each other, and they could also conflict with a system’s basic motivating feelings. In such cases, the basic G-law would require the E-systems to resolve conflicts rightly, having regard to all relevant G-laws. However, these E-systems could nevertheless act wrongly by following motivational feelings in disregard of the requirements of G-laws, or through weakness of will failing to ‘carry out’ an action selected as the right one; and they could even be evil in that they took pleasure in doing harm contrary to the requirements of G-laws.

One of the G-laws that a rational self-conscious E-system could feel is the following moral law, which I call ‘improve oneself’:

Enhance one’s own ability to find out, decide rightly, carry out, and do good.

This law would require the cultivation of virtues associated with the seeking of truth, particularly in so far as the truth was relevant to one’s own actions; and of virtues associated with readiness, willingness, and ability to put decisions rightly made into effect, and also to enlarge one’s opportunities to do good. Thus it would require a rational self-conscious system, in its finding-out, to try to be both interested (inquisitive, enthusiastic, self-reliant, diligent, tenacious, and so on) and disinterested (balanced, fair-minded, self-critical, aware of its own strengths and limitations, concerned for truth not self-promotion or self-justification, willing to admit error, duly respectful of other opinions, and so on). And it would require the system to enhance its capacity to give effect to its decisions and to do good, for example by improving its practical skills and courage and sympathy, combating laziness, guarding against giving undue weight to its own interests, and so on.

Thus, in E-systems with self-conscious rationality, the G-laws, which I conjecture are laws of nature that are to some degree felt by all E-systems, could come to be felt as being or including a system of guiding moral laws, which are truly existing features of the universe, ascertainable, and to be respected by us whether we like it or not. There could be disagreement as to what G-laws require in particular circumstances, and as to what, among rules which people claim to be moral laws, are G-laws or rightly derived from G-laws, and what are merely fallible human inventions; although I don’t think there would be much room for disagreement about the very general G-laws which I have identified. And as well as supporting the values of truth and goodness, as I have discussed, G-laws could also be associated with aesthetic values relevant to the decision-making of rational E-systems.

Finally in my account of what G-laws could be, it will be seen that I have not treated basic motivational feelings such as pain and hunger as being, or being supported by, G-laws. It might have been possible to do so, just as it might on the other hand have been possible to treat what I have called G-laws as being no more than motivational feelings like pain and hunger and so on. I have said that G-laws might be considered a sub-class of E-laws, but I do suggest that at least very general G-laws like those I have identified have an existence and validity such that feelings associated with them are more than evolutionary artifacts:⁶ indeed, this contributes to the explanatory power of my hypothesis, in particular in relation to plausible reasoning and morality. Furthermore, although basic motivational feelings do in a sense guide E-systems, and do so non-conclusively, it seems to me that they do so wholly because of their character as raw feeling, rather than because they are in any sense apprehended as law-like requirements. They are just felt as incentives, whereas, as I see it, G-laws are in a sense known as requirements. Even the most primitive E-systems must in some sense come to know that they can make selections and come to know how to do so; and at the same time, I suggest, they must in some sense come to know the basic G-law ‘act rightly’, as a requirement and not merely as a felt incentive.

5 Plausible Reasoning

The notion that there are laws of nature which guide but do not constrain, and that they do so at least partly through being felt and in some sense known, may seem an odd one, which unnecessarily introduces strange and mysterious entities and ideas. But I suggest that something like this notion is necessary if we are to understand even such a familiar thing as human plausible reasoning; and that it also is a very promising basis for solving many other problems associated with human and animal consciousness.

In this section I will consider plausible reasoning and the possible role of G-laws; and in the next I will look more generally at the possible explanatory power of the E-laws and G-laws hypothesis.

(a) Human plausible reasoning

One important aspect of human rationality, in particular in theoretical reasoning, is that we reach and justify conclusions on the basis of reasons, some of which serve as premisses or data. In some cases, the relationship between the conclusion and the data can be expressed in terms of definite rules, so that the conclusion can be reached and justified by the mechanical application of rules to the data. This kind of reasoning is called formal reasoning; and it can be performed by a machine such as a computer whose operation is fully explicable in terms of C-laws. The most obvious examples of it are to be found in the fields of logic, probability theory, and mathematics. (I do not suggest that all reasoning in these fields is formal reasoning: in mathematics in particular there is great scope for creative thinking that is not a matter of applying rules to data.)

However, most of our reasoning is not of this type, on the surface at least. Generally, the most important parts of any process of human reasoning require judgement: the conclusion generally does not follow from the data on the basis of definite rules, such as the rules of logic or probability theory. When we are trying to decide what to believe on a particular point, very often the reasons for and against alternative beliefs seem inconclusive, and also seem to be immeasurable and incommensurable. A decision then cannot be made by calculation or computation or any other way involving the mechanical application of conclusive rules: a decision can only be made by the exercise of judgement. This kind of reasoning is called informal or plausible reasoning; and most of the reasoning we encounter in everyday life, as well as in scientific and philosophical writings, is of this kind.

Now, over the centuries there have been arguments to the effect that plausible reasoning can be formalised, and that, in so far as it cannot be, it has no value: either a conclusion is supported by premisses on the basis of conclusive rules (logic, mathematics, etc.) or it is not supported at all.

One important strand in this debate is the problem of induction: the problem of how general statements or theories about the world can be supported or justified by particular observations. Significant contributions to this issue have been made by writers such as Hume, Popper, Hempel, Goodman, and Putnam, and I will not recapitulate them here: the general point is clear enough. No amount of particular observations (or statements reporting them) can possibly entail a general statement about the world: no amount of observations of black ravens (short of observations known to be of all ravens) can entail the conclusion that all ravens are black, or indeed that any unobserved raven is black. No conclusion which contains more information than the premisses on which it is based can be justified logically by those premisses.

Yet most of us accept that science and philosophy give us good reason to believe that certain general statements about the world are probably at least approximately true; and that the progress of science delivers closer approximations and greater probabilities. If this were to be justified on the basis of rules, which can be applied to data without the necessity of any exercise of judgement, then it is clear that those rules would have to concern how conclusions may follow from data as a matter of probability rather than as a matter of certainty. And the basis of all significant attempts to formulate rules to justify probabilistic conclusions is Bayes’ theorem: an equation of probability theory devised in the eighteenth century by the Reverend Thomas Bayes. However, application of this theorem itself requires judgements of prior probabilities, and so cannot obviate the necessity of the exercise of judgement that goes beyond both observation and the application of rules of logic or probability.⁷

It is for reasons such as these that Hilary Putnam has argued that human rationality cannot be completely formalised, without formalising complete human psychology - and possibly not even then.⁸

Putnam’s reference to the formalising of complete human psychology seems to involve the assumptions that human rationality is really a matter of useful computations or computation-like procedures selected by evolution; and that this rationality, apart from so much of it as involves the application of rules of mathematics, logic, and probability, does not have any justification from logic or probability theory or any other generally-applicable rules, but only the justification that it consists of computation-like procedures which have proved useful in promoting the survival and reproduction of our distant ancestors. If it is the case that human rationality does come down to computation-like procedures selected by evolution, then, while it may not be a practical possibility to discover and set out these procedures, this would be consistent with full explanation in terms of C-laws.

This to my mind is the least implausible suggestion as to how our informal rationality could be consistent with full explanation in terms of C-laws. One difficulty with it is that it assumes that the problems which faced our evolutionary ancestors, during the millions of years in which our computational capabilities evolved, were such that the features relevant to our computational capabilities, which were selected to deal with these problems, are suited to all the purposes to which we apply our informal rationality today. As Thomas Nagel puts it:

But the capacity to form cosmological and subatomic theories takes us so far from the circumstances in which our ability to think would have had to pass its evolutionary tests that there would be no reason whatever, stemming from the theory of evolution, to rely on it in extension to those subjects. In fact if, per impossible, we came to believe that our capacity for objective theory were the product of natural selection, that would warrant serious skepticism about its results beyond a very limited and familiar range. An evolutionary explanation of our theorizing faculty would provide absolutely no confirmation of its capacity to get at the truth. Something else must be going on if the process is really taking us toward a truer and more detached understanding of the world.⁹

Furthermore, in order to reach and to justify the view that our computational procedures were selected by evolution and can be relied on for that reason, we have already relied on a vast amount of informal plausible reasoning: so we can’t be justified in having confidence in the belief that our informal plausible reasoning consists of computational procedures with the warrant of evolutionary fitness unless we are justified in having confidence in the informal plausible reasoning on which that belief is based. And if our only justification for confidence in our informal plausible reasoning is confidence in the belief that this consists of computational procedures selected by evolution, then we have a vicious circle.¹⁰

Now if human plausible reasoning has value but cannot be formalized, as this argument suggests, then it would seem that plausible reasoning may need to be supported by primitive judgments of the kind mentioned earlier, and also that the E-laws and G-laws hypothesis provides a promising approach to understanding it better. And the hypothesis has a particular advantage in relation to the place of patterns or ‘gestalts’ in plausible reasoning.

(b) Gestalts

An important feature of human plausible reasoning is that it appears to take account of unique totalities, in their full particularity; and I suggest that, if this is so, then it cannot be through the operation of C-laws alone.

Let us suppose that one feature of SU₁ is that a person is experiencing something as a gestalt, a pattern which the person grasps as a whole. Let us suppose, for example, that the person is judging a piano competition and is just now attending to one of the competitors playing the piano transcription of Bach’s ‘Jesu, joy of man’s desiring’. It seems clear that the totality of the adjudicator’s appreciative experience at this moment is unique: the adjudicator is a unique person, the competitor’s playing is not exactly the same as anyone else’s or even the competitor’s on another occasion, the circumstances are unique, and so, we may take it, the adjudicator’s appreciative experience is unique. It also seems clear that the appreciative experience is causally efficacious, at least in that it is causally relevant to the assessment of the performance and its comparison with other performances. If all causal efficacy is through C-laws, any efficacy in bringing about the transition to the adjudicator’s assessment of the performance must be through general and quantitative features, either of the experience itself or (perhaps more plausibly) of the brain events supporting the experience, the experience’s neural correlates: the unique totality of the particular experience could not be causally efficacious in bringing about this transition (although of course it still could be causally efficacious in the sense of being part of the initial conditions which are modified by C-laws engaging with features that are general and quantitative).

Even a person’s experience of a melody alone would appear to be a totality of a kind which does not, as a totality, have a generality such that it could engage with C-laws. Bach’s famous melody has of course been heard and grasped and responded to many times by many people, but it did not exist until it was composed in the 18th century; and it would be bizarre to suppose there was a C-law waiting around until then so that it could engage with that melody, or a gestalt associated with it, if and when that melody should come into existence. If a person’s experience of a melody is causally efficacious in bringing about the transition to an emotional response of the person to that melody, then it cannot, if all causal efficacy is through C-laws, be through the particular overall quality of the melody or the experience; but rather it must be through other features of the sounds or experience or associated brain events which do have the generality appropriate to engage with the general laws.

If all causal efficacy is through C-laws, then, an experience of a Bach melody would be like the glider in the (unmodified) Game of Life, in that it can bring about changes in the world, not through its particular overall quality, but only through features having a generality appropriate to engage the general laws. This would pose no problem for identification of the melody: plainly a melody can be identified by purely quantitative processes applied to the order, pitch, duration, and spacing of the notes. However, the grasping of a melody, along with any associated appreciative response to it, seems to be an experience of an indivisible whole; and it does seem that wholes or gestalts of this kind have effects on how we think and on what we do. Of course, such ‘seems’ can be misleading; but I believe it is implausible to suggest that an experience of this kind can be causally efficacious to bring about changes, not through its particular overall quality, but only through features which have the generality appropriate to engage with general laws.

However, if there can be causal efficacy through E-laws and G-laws, the transition to an appreciative response to a Bach melody could be caused by the melody as a whole and its associated gestalt experience - just because the person having the experience could contribute to the creation of the response and to the causal processes flowing from it. At least on the first occasion when the person’s appreciative response to the melody occurred, this response itself could involve selection: the physical processes and the C-laws could leave open a range of physical outcomes, just one of which corresponds to the particular appreciative response that actually occurs. And the person’s attentive participation in the process could, through E-laws and G-laws, have a role in the selection of the outcome (just as it would have had in the selection of an alternative, if an alternative had occurred). Even if on subsequent occasions an appreciative response occurs automatically, this response could be specific to the particular gestalt because it was made specific by the initial and partly creative response.

On this approach, the adjudicator of the piano competition would apply, to the unique experiences of the various performances, aesthetic standards which could be, or be based upon, G-laws. Plainly these standards would not be conclusive of the adjudicator’s assessment. As we have seen, general rules cannot be conclusive in relation to unique wholes where, in addition to having effects via those of their features that have rule-engaging generality, the wholes as such enter into the causal process: the assessment would require the resolution of considerations that are inconclusive in their application to the particular case.

This argument in support of my hypothesis depends upon my contention that the unique whole of a gestalt cannot engage with general laws of nature, and that accordingly something like a selection is required if gestalts are to enter into the causal process in such a way as to bring about changes in the world. It might be objected that gestalts must arise without a selection being made: even if there may be some measure of selection in a response to a melody, the melody has to be heard and grasped as a whole before there can be any question of a response to it; and so the gestalt experience must somehow be caused by general laws, and to that extent must engage with them. What I say to this is that E-laws would specify that certain types of events give rise to conscious experiences which have general features corresponding to general and quantitative features of the substrate events, and also specify that these features, whatever they may be, are chunked or bracketed in the conscious experience. A particular and unique gestalt or total experience could thus arise through this chunking or bracketing, without any need for its unique totality to have engaged with C-laws or E-laws, either as cause or effect. However, this totality could then operate as a cause, as the E-system proceeds to apply G-laws.

(c) Pre-human plausible reasoning

Although I do not suggest that other animals are rational in the same way as human beings, I do suggest that they engage in something like plausible reasoning.

I have already referred to systems motivated in different directions by feelings of pain and hunger, or sex and fear. It is reasonable to think that such feelings are correlated in a regular (and therefore lawful and general) way with physical features: indeed, that is part of what I suggest E-laws would do. But no C-laws have been formulated which engage with the feelings as such; suggesting that, if the outcomes in such cases are always determined wholly by initial conditions and C-laws, then the feelings have no role other than to accompany the physical correlates. If the feelings themselves are to play a causal role over and above that of their physical correlates, something more than initial conditions and C-laws seems to be required.

Furthermore, as we have seen, phenomenal features including feelings are characteristically chunked into the unique wholes of consciousness; and these unique wholes can play no part in bringing about changes through C-laws. I suggest it is reasonable to think that, even in quite primitive animals, the feelings and unique wholes of consciousness do play a causal role in bringing about changes; and that this must be through the operation of E-laws and G-laws. This is supported by the consideration that this could have advantages for an organism’s survival and reproduction, and thus be favoured by evolution.

Survival and reproduction of animals is enhanced by such things as obtaining food and shelter, mating and producing offspring, and avoiding predators. These things would certainly be assisted by a central nervous system that can perform elaborate unconscious computations, as ours can. But such computations, depending entirely on C-laws, could only deal with general and quantitative features with which C-laws engage, and could do so reliably only in respect of circumstances which can be related by computation-like processes to circumstances previously encountered by the organism in question and its evolutionary ancestors. What I suggest is that there would be advantages in being able to deal in a reasoned way with circumstances which are not exactly the same as those previously encountered, and which cannot be related to those previously encountered in terms of the general and quantitative features with which C-laws engage.

Suppose that an organism’s evolution-selected computational procedures indicate that action A is appropriate in circumstances X (let’s say these are circumstances suggestive of the presence of food), and action B is appropriate in circumstances Y (let’s say these are circumstances suggestive of the presence of predators). The organism encounters unprecedented circumstances Z, for which the computational procedures can determine no appropriate action, other than to indicate that Z is like X except in a respect which the procedures can only represent by some quantity which has no meaning for the system, and that Z is like Y except in a respect which the procedures can only represent by some other quantity which likewise has no meaning for the system. In that situation, the organism might have a better chance of making the right decision whether to do A or B if it could make some kind of holistic comparison of Z with X and with Y, in which account could be taken, not only of the general features of each of the circumstances, but also of the particular way in which these general features are combined in each of the three circumstances; that is, if the organism could engage in a kind of reasoning by analogy, looking for relevant but unanalysable similarities and differences between whole sets of circumstances.

That kind of holistic comparison, taking into account particular and perhaps unique features of whole combinations of circumstances, could not be achieved by C-laws, but could be achieved by E-laws and G-laws. And feelings such as pain and hunger, which appear superfluous if any determination of what happens is by initial conditions and C-laws, can be understood as motivating the organism to make selections empowered by E-laws and guided by G-laws.

6 Explanatory Power

I will now very briefly outline how I believe my hypothesis could contribute to the explanation of many baffling features of our world.

(a) Consciousness

I believe that, whatever else consciousness involves, it does at least involve the capacity of a system to grasp information as a whole; and I believe it is reasonable to suppose that a system with this capacity does influence its own behaviour on the basis of that grasp, as postulated by my hypothesis.

One enormous shortcoming of all physicalist theories of consciousness is that they are unable to give a plausible account of what consciousness is for. They refer to such notions as representations and the intentional stance; but have no explanation as to why the representations are not unconscious, like those used by computers, or why the intentional stance requires actual rather than assumed or virtual consciousness.

Our brains have prodigious unconscious computing capabilities, for example in the pre-conscious processes that give us stereoscopic vision of an apparently stable scene, despite movements of our eyes, head, and body. These computations are extremely reliable, and beyond the capacity of today’s computers. Compared with these non-conscious feats, our conscious efforts may seem puny indeed: conscious performance of such tasks as simple mental arithmetic is clumsy and prone to error, conscious reasoning generally is riddled with fallacies and biases, and conscious awareness of a crisis can bring on irrational panic.

One might therefore have expected that evolution, which has given us both our prodigious non-conscious computing capabilities and our fallible conscious processes, would have made sure that, for really important decisions, and especially in a crisis, our consciousness would be shut off, in order that our non-conscious processes could work away without interference to find the solution and give effect to it. Yet, as we know, the reverse is true. When faced with an important decision, and particularly in a crisis, our conscious attention is automatically brought to bear; so that we cannot help addressing the problem with full consciousness. This surely indicates that there is something about our conscious processes which gives them an advantage over wholly non-conscious computations, and some explanation is required of what this advantage could be.

My contention is that the great advantage of consciousness for survival and reproduction is that it enables plausible reasoning of the kind considered in the previous section, making use of gestalts and analogies. E-laws would explain how consciousness arises: they would include the bridging laws being sought in current efforts to correlate kinds of brain events with kinds of conscious experiences. Perhaps more importantly, E-laws and G-laws would explain why systems with consciousness can do things that can’t be done by systems without consciousness. Because G-laws do not constrain, but merely give guidance, they need to be applied, and thus presuppose something like consciousness. In order that a selection which is not constrained by laws be other than arbitrary, it needs to be grounded on something that has both generality and validity, thus presupposing some kind of valid but non-conclusive principle or principles - that is, something like G-laws.

(b) Qualia

The application of G-laws would require that a system have information about circumstances facing it, in a form that is comprehensive and comprehensible, so that the system can make a selection as to its future which is based on a grasp of those circumstances.

Now unconscious computation works best with the numbers and symbols of computer codes. Conscious reasoning is very clumsy with numbers and symbols; but works very effectively with the sensory modalities, which present a vast amount of information to a conscious system in a graphic way, so as to make it possible to take in this information all-at-once, with salience given to important features. If it is the case that we really do make selections on the basis of non-conclusive G-laws, it becomes understandable why we have colours and sounds and so on, making the reasons that base our selections vivid, and aiding the use of analogy.

(c) Pain and emotional feelings

We accept that pain has useful functions which explain why evolution has selected in its favour: namely, it draws our attention to possible damage to ourselves and gives us a strong motive to take steps to remedy it and to avoid damage in the future.

However, if our actions were based on the inevitable operation of C-laws, that is, upon computation-like procedures, then the pain would be a superfluity. It would be madness to suggest that, in order to ensure a computer did not deviate from operating in accordance with its program, it would be a good idea to try to set it up so that it felt pain whenever it did deviate!

Similarly with emotional feelings. It is now recognised that emotion is an important part of our rationality; and it seems that it contributes to our rational decision-making inter alia by eliminating from consideration some alternatives for action, and also by drawing our attention to the need to give careful consideration to relevant matters.¹¹ But this presupposes the value of conscious appraisal of alternatives, which in turn confirms the inability of a science limited to C-laws to explain our informal rationality.

(d) Intentionality and meaning

If E-systems can grasp gestalts as totalities, and act upon the basis of that grasp, then meaning could consist in the relationship of a symbol to a gestalt that it represents; and the symbol, the gestalt, and the relationship between them, could constitute a further gestalt which an E-system could grasp and act upon.

(e) Personal identity

If one accepts physicalist views of consciousness, it is very hard to explain personal identity as being anything other than a kind of physical continuity and/or psychological connectedness.¹² Of course, even with such a view, the special concern one has for one’s own future, and also notions of deserved praise or blame, can be given explanations in evolutionary terms; but such a view makes it very difficult to regard these ideas as anything other than evolutionary artifacts with no rational justification whatsoever.

The idea of E-laws and G-laws opens up the possibility of a criterion for personal identity that transcends physical continuity and/or psychological connectedness. If E-laws empower a system to make selections as to its own future states, and G-laws guide the system in making such selections, then, turning this around, one can say that one thing that manifests an E-system’s identity over time is the susceptibility of its own future states to be affected by this process of selection. That is, the relationship of the initial (selecting) state and the outcome (selected) state is one of a continuity of a kind that could ground personal identity.

This relationship will in normal circumstances coincide with the physical continuity and/or psychological connectedness that are the orthodox criteria for personal identity; but it suggests a different and more basic criterion, on the basis of which one’s special concern for one’s own future and notions of deserved praise and blame can be rationally justified.

(f) Passage of time

On physicalist views of consciousness, the passage of time is often presented as an illusion. This is partly because of the concept of a four-dimensional block universe suggested by relativity theory, and partly because the only significance that physics can give to the passage of time is through the second law of thermodynamics.

If all laws of nature are C-laws, then it becomes problematic to hold that the present determines the future in a way that it does not determine the past. Newtonian determinism would suggest that the present uniquely fixes the past just as much as it uniquely fixes the future. If there are just C-laws and randomness, then the only difference between the past and the future is that the past is fixed because it has occurred, whereas, unless the block universe view is correct, the future may not be. (This would not sit well with the assumption I made in the first paragraph of this essay, to the effect that the laws of nature are efficacious in bringing about or contributing to what happens - but this assumption is widely held and should not lightly be rejected.)

On the other hand, if there are E-laws and G-laws, then a direction of time is manifested in that it is only future states of an E-system that can be affected by selections it makes.

(g) Free will and responsibility

If rational self-conscious systems can make selections as to their own future states, that are not determined by C-laws on the one hand and not random on the other, then the plainly there is the possibility of a satisfactory account of free will. And if the personal identity of such systems is tied to this selection-making capacity, then sense can be made of a notion of responsibility that can result in a system deserving praise and blame for a selection, at a time later than the time of the selection in question.

Such an account would not fall foul of Hume’s claim that, if a decision is not caused by a person’s character, the person cannot be responsible for it. A person’s character would through C-laws limit the alternatives available for choice, and would through E-laws be instrumental in determining the operative reasons; and the person (a whole system, including the person’s character) would make a selection guided by G-laws. Thus the decision would in a real sense be caused by the person’s character, though not pre-determined by it or by anything else.¹³

(h) Morality

The existence of moral ideas and attitudes can be given an evolutionary explanation.¹⁴ But such explanations cannot support the validity, let alone the truth, of moral beliefs. Common sense tells us that a belief that it is wrong, for example, to torture people for amusement is a belief which is true, but a thesis such as that of Gibbard precludes him from saying this: indeed, it requires him to say that such a belief is not true, because there are no normative facts. The most he can say is that anyone who disagrees is not a ‘competent normative judge’.¹⁵ Gibbard d