The luck of ancestry

A letter by Maurice Winter in Cambridge Alumni Magazine, issue 67, Michaelmas 2012, commenting on an article in issue 66, includes the following: "Luck is of fundamental importance to all sportspeople. Take Usain Bolt. It is pure luck that his parents provided him with the genes that built him in the way they did".

One reading of that, which may very well be the reading that the author intended, makes it straightforwardly true. We have different genetic endowments, and the result is that many of us would never be able to keep up with the best in sports, however hard we trained. But there is another reading, and it is one that draws our attention to the question of what kind of luck may be said to be associated with ancestry.

The reading is that the future Usain Bolt was there, before conception, waiting to be assigned genes, and perhaps waiting to be assigned parents. On this reading, it was pure luck that the genes that made great sporting achievement possible fell to him, rather than to someone else, or to nobody.

It is certainly pure luck that this particular combination of genes came together in a human being at all. When two people have a child, with a particular genetic make-up, there are many other genetic make-ups that could have been embodied in a child of the same parents, but that were not so embodied.

But this does not mean that we should accept the picture of a future Usain Bolt, waiting to be given a genetic endowment. While the term "Usain Bolt" now has a referent, it does not have any referent in temporal slices of the world that come before his conception. There is a connection with Kripke's thoughts on a person's being essentially the person made from that particular sperm and egg. They could not be that person until the relevant sperm and egg had come together. It is the coming together that makes the operation of a Kripke-type criterion of identity possible. Quine's slogan, "No entity without identity", has a use at this point.

If we are not to accept the picture of a pre-conception Usain Bolt, it was not luck that Usain Bolt had his genetic endowment. It was luck that this genetic endowment was embodied, but not luck that he got it. And we should not be distracted by the (probable) fact that only his parents could have given precisely that endowment to a child. Again, it was not luck that he was conceived in that family, rather than in any other. He did not exist, in any sense, until conceived.

So much may look obvious, once it has been said. But it is worth thinking it through, in order to understand the nature of the luck that is involved in inheritance.

I would not draw any conclusions about what to do. These thoughts might make it illegitimate to say things like, "It is most unfortunate that X got saddled with those genes", in a sense that would imply that X could have had different genes. Someone with different genes would not have been X. But these thoughts would not make it illegitimate to say things like, "Someone with X's genes finds life tougher than the rest of us, so we should make special provision for X, offer changes to X's genes (when that becomes possible), and so on". That is, there is no ground here for a fatalistic conclusion that people should simply live with their genetic inheritances. Nor is there any ground for an anti-welfare-state conclusion that we should not care for people with genetically based difficulties because those difficulties are so bound up with what makes them who they are.

Now let us consider an area in which we are free to vary a basic mechanism that determines a significant consequence of parentage, in a way that we are not free to alter the mechanism of genetic inheritance. There are still some hereditary monarchies left around the world. The next monarch is whoever happens to have been born to the right couple, and in the right place in the order of their children. No-one from any other family has a chance. Let us assume for the sake of argument that it is appropriate to have a lifetime head of state, and let us confine ourselves to constitutional monarchies on the modern European model, to avoid having to discuss the evils of absolutism and of arbitrary rule. Is a hereditary system unfair? Would it be fairer to have an elected lifetime head of state, a modern-day Doge of Venice?

There is a sense in which it is not unfair to have a hereditary monarchy. No-one can say, "I was excluded from that position by the luck of my parentage", because no person could have had different parents from those that they actually had. Anyone with different parents would have been a different person.

On the other hand, it is not possible to say, "The hereditary principle is as good as any other, because we need to pick someone, and a lottery would be fair, and picking whoever has the right parents is equivalent to a lottery". One cannot say that, because it is not as if we were all there, pre-conception, waiting to be allocated our parents at random. That is, we were not participants in a lottery, so the analogy with a lottery would break down. Anyone who wanted to maintain that the hereditary principle was as good as any other, would have to find some other grounds for that claim.

Finally, the lack of a certain sense of unfairness about the hereditary principle would not prevent one from putting forward other arguments against its use.

Turing's secret papers

This year, we mark the hundredth anniversary of Alan Turing's birth. One celebration, among the less publicised but also among the most fitting, has been the release into the public domain of two of his wartime papers on mathematical aspects of cryptography. The release was announced by GCHQ, here:

http://www.gchq.gov.uk/Press/Pages/turing-papers-released.aspx

It seems that the papers were not released earlier, because their contents were still considered to be sensitive. It was thought to benefit the UK to have the information while others did not have it, and (presumably) while others did not even know that the UK had information of this precise nature.

This raises an interesting general question. If a country's authorities have scientific or technical information, and feel that the country can gain a significant advantage over other countries by keeping it secret, should the information be kept secret?

An argument for release of the information would be that if the information is useful to someone, it is likely to be useful to others too, and its release might well promote economic development. It is not hard to think of examples. Encryption systems help commerce, by facilitating financial transactions where there would otherwise be a risk of diversion of funds to criminals. Knowledge of the vulnerabilities of encryption systems helps developers to make the systems more secure. Algorithms for the management and distribution of military supplies could be useful in commercial logistics, allowing goods to be stored and distributed at the lowest possible cost. If the detailed technology of radar and of jet engines, both largely developed in the run-up to war and during wartime, had been kept as military secrets, long-distance travel would not have become as convenient as it did in the 1950s and 1960s. And so on.

It is tempting to say that decisions must be made on the facts of each case, and that general principles would be very difficult to find, and of far less importance than specific facts. But we should not give up quite so quickly.

One could, for example, adopt the reasonably general principle that when the usefulness of information to others would be purely military, there would be a strong presumption in favour of keeping it secret. It may be desirable to help the world's economy, but when it comes to potential military uses, a country could be argued to be fully entitled to put its own interests first. One difficulty in applying this principle would be that it is all too easy for the military, and the security services, to see only the military uses of information, and to be unaware of commercial possibilities.

Alternatively, one could adopt a utilitarianism that gave no special weight to one's own country. Then, if the consequences of secrecy and of disclosure could be computed (which they could not be), or plausibly estimated (which might be possible), the decision on whether to disclose each piece of information could be reached by a mechanical procedure.

One could row back a little bit from this generous principle, and adopt a utilitarianism under which the interests of the citizens of one's own country were weighted more highly than the interests of others. But it would be hard to make a case for a precise difference in weighting. There would only be a plausible range. Counting each of one's own citizens as twice as important than each non-citizen might be defensible, particularly when one's own citizens formed a small minority of the world's population, so that the total weight assigned to one's own citizens was still considerably less than the total weight assigned to all non-citizens. Counting each of one's own citizens as 100 times more important as each non-citizen would not be defensible, regardless of population. One would also have to balance the potentially incommensurable economic and security interests of one's own citizens (and of non-citizens, to the extent that their security would be affected by the release of the information). And one would have to bear in mind that a boost to the world economy from an invention that new information makes possible can lead to a boost to the economies of many nations, including those that do not directly exploit the invention.

It is never going to be easy to make the right decisions, even assuming that there are right decisions to be made. One reason why this is so is that the specific facts of each case are bound to have a substantial influence on the decision, and it is in the nature of the case that those facts must be kept secret unless the decision is to publish. Only general principles can be debated publicly, and those general principles can only take us so far.

Geographical sensitivities

At the start of the British Museum's temporary exhibition, Ritual and revelry: the art of drinking in Asia, there is a map of a large part of Asia. At the bottom of the map, there is the following statement: "The names and designations used on this map do not imply official endorsement or acceptance by the British Museum". The intention is presumably not to upset anyone who might think that the map implied territorial claims that they rejected, or who rejected certain place names because of their historical or political associations.

This strikes me as silly over-sensitivity. No boundary lines are drawn on the map. The only remotely contentious labels on the map are "Tibetan Plateau" (instead of "Tibet"), "Korea" (not distinguishing North from South), and "Burma" (instead of "Myanmar"). Taiwan is not labelled at all. I cannot see any reason why anyone could reasonably take offence at what is on the map. And if someone took offence at the absence of names for some countries, or at the absence of boundary lines, that would be equally unreasonable. It is not just Taiwan and Tibet that are not labelled. Several countries are not labelled, simply because there is no need for the map to show that much detail.

It is not, however, simply a matter of over-sensitivity to people's political and geographical sensibilities. There is a second-order issue. By adding the statement, the British Museum has conceded that it is reasonable for people to make a fuss about maps which are published by bodies that have nothing to do with any governments or aspiring governments, and where the bodies clearly do not have any intention to make political waves in the regions mapped.

I am concerned at a possible consequence of making that concession, a consequence that is by no means certain to ensue, but that would nonetheless be serious. It is this. Such a concession would put us on the road to allowing scholarship to be constrained by political, cultural and religious sensitivities: "Don't present that result, or that theory, it would upset such and such a group". Over the past few centuries, we have gradually shed such constraints, although we have not got rid of them completely. Their return would be an intellectual disaster.

Are satyrs human?

The star of the excellent exhibition of bronze sculptures that has just opened at the Royal Academy in London is the truly spectacular Dancing Satyr:

http://en.wikipedia.org/wiki/Dancing_Satyr_of_Mazara_del_Vallo

There are other, smaller, early modern satyrs and satyresses too. These reflect the merger of the ideas of the satyr and the faun, in that the legs are distinctively goat-like.

Now suppose that satyrs and satyresses with such distinctive goat-like features really existed. Would we regard them as human?

It is tempting to make the answer depend entirely on genetics. If they were a separate line, with any common ancestor with us being very remote, and if they did not interbreed with human beings, we would be inclined to say no. If, on the other hand, they were born of human beings by a strange mutation, that prevented interbreeding with non-mutated human beings and that was generally passed on to their offspring, we might well say they were human. We would be particularly likely to do so if a few children of satyrs and satyresses did not have the mutation and could merge back into the main line of human beings, interbreed with non-mutated human beings, and have children who did not have the mutation.

Even if we decided to rely solely on genetics, this would not answer every question. Suppose that some satyrs and satyresses had been born of human beings five thousand years ago, and no-one from their line had ever merged back into the main line of human beings. Would we still say that they were human beings? One thing that might hold us back from saying that they were not human, would be the possibility of their having children without the mutation, who could merge back into the main line of human beings.

If, however, that possibility would carry weight, how strong would the possibility need to be? If a remote possibility would carry weight, we might have to say that chimps were human because at some time in the future, they could evolve into alternative versions of homo sapiens sapiens who could merge into the human line. We would probably rule out attaching weight to that possibility in relation to chimps, on the ground that even if it might happen, it would certainly not happen for many thousands of years, whereas with the satyrs and satyresses, we envisage that it might happen at any moment. That is, remoteness in likelihood and undoubted remoteness in time can have different effects on our attitudes.

Alternatively, we could take a view that was not purely genetic. We could say that social interaction mattered too. The more we interacted with satyrs and satyresses, in the same ways that we interact with human beings, both at work and at play, the more likely we would be to regard them as human. But we could not let that become the only criterion. If it did, then those with whom we cannot interact, either because they are inhuman in their conduct or because of their restricted faculties, would not count as human, and that would be quite wrong.

I have repeatedly referred to satyrs and satyresses together. When we see them as male and female, as they are portrayed in some of the sculptures at the exhibition, and not just as male, that makes it much easier to see them as human beings. And so I would, were they to move in next door.

Atlas Shrugged by Ayn Rand

Ayn Rand's writings polarize opinion. There are those who think she has all the answers, and those who think she has none. Both attitudes are mistaken, but it is interesting to ask why the attitudes arise.

One reason that I shall leave to one side is that the support for Rand's views which she offers in the form of academic philosophy, is not adequate. But this does not prevent some of her economic and ethical claims from being true. And it should not lead us to discard all of her work, and only to seek those true claims elsewhere. The fact that her explicitly philosophical works do not stand up well to academic examination, does not mean that we cannot learn from her novels. (We can also extract some useful ideas from her explicitly philosophical works. There is, for example, something to be made of her analysis of perception.)

One reason why Atlas Shrugged invites polarized responses is that it contains some propositions that are very easy to accept as obviously true, and some that are very easy to reject as obviously false. People can easily pick on one set or the other, and then jump to opinions about the whole work.

One obvious truth is that planned economies, in which government agencies dictate who is to produce how much steel or what trains are to run, are uniformly disastrous. Likewise, crony capitalism, under which some businesses get favours because they have friends in high places, is disastrous. Moreover, Rand sets out, in exquisite detail, the mechanisms of perverse incentives, laziness, incompetence and corruption that lead planned economies and economies that run on cronyism to disaster.

One mechanism with contemporary relevance is set out when Hank Rearden is trapped by Floyd Ferris, because he has sold metal to a customer in breach of regulations (part 2, chapter 3; pages 433-434 of the 2007 Penguin edition). Ferris admits that this was the point of the regulations. Set up lots of rules, let people break them, and then you have a hold on them. It would not be difficult to see those who appoint our modern zero-tolerance authority figures, the street wardens who catch you drinking alcohol in a park where that is forbidden or who find that you have put your dustbin out on the wrong day, as thinking in the same way, albeit on a much more trivial scale.

Another obvious truth is that we must face facts and live by reason. It is both disastrous and pathetic to live by superstition, or to think that wishing will lead to the results that one desires. The specifically political dangers are set out in John Galt's speech (part 3, chapter 7, pages 1,009-1,069). The speech is not only an attack on totalitarianism that is just as strong as George Orwell's attacks, although a good deal less subtle. It is also a powerful attack on mysticism in all of its forms, not least the form in which it may by invoked by totalitarians who claim to embody a truth that all must accept (pages 1,042-1,046).

The leading obvious falsehood, and the one that repels those who believe in a welfare state, is the claim that we have to go to the opposite extreme, that a government has no business doing anything more than maintaining law and order. Income tax is deemed to be theft, when the pirate Ragnar Danneskjöld includes it in the sums that are to be restored to Hank Rearden (part 2, chapter 7, page 579).

It is true that a possible consequence of taxation is a "from each according to his ability, to each according to his need" scheme, such as the one that ruined the Twentieth Century Motor Company (part 2, chapter 10, pages 660-670), but one could guard against that by having a firm limit to the ratio of public spending to GDP, ensuring that most resources were distributed in accordance with market forces. One third of GDP might, for example, be a sensible limit. One would also need to require an economically neutral tax system, not one that was riddled with special reliefs for particular activities, in order not to veer towards cronyism or central planning.

Moreover, while deploring central planning in general, we may recognize the need for some economic regulation, or some taxation of specific activities, for example to control externalities. That would merely amount to correcting for costs which the market fails to price into activities. We may also recognize that the market tends to under-provide public goods (in the economists' sense of non-excludable and non-rivalrous goods).

Another reason why Atlas Shrugged invites polarized responses is that the arguments are put in dramatic form, in which it is easy to be carried along by the story. This encourages those who think that the novel has all the answers not to look too critically. It also gives those who doubt the merits of the arguments an opportunity to slow down, to examine the dramas, to note that the characters on the wrong sides of the arguments make implausibly weak cases for their points of view, and then to suspect the quality of all of the arguments in the book. We can see this by looking at comments made by Wesley Mouch and by those around him to justify central control, and at the pleas by members of Hank Rearden's family for his financial support.

While the cases made by some of the characters may be implausibly weak, the portraits of some of the professional politicians, who have done nothing but climb the greasy pole, are delightful, and worth pasting up on the mirrors of some of our own ministers and mandarins. Another character whom we may spot in real life is the philosopher Simon Pritchett, who denies not only the significance of human life, but the power of reason, the validity of our concepts and our ability to know things (part 1, chapter 6, pages 131-133). Some of those who waffle under the banners of postmodernism and deconstruction are just as bad.

A third reason why Atlas Shrugged invites polarized responses is that it is easy to concentrate on one aspect or another of Rand's ethic itself, and to feel exalted or repelled, depending on the focus of one's attention.

To start with the exaltation, the ethic of setting yourself challenging and productive goals, and then striving with all your might to achieve them, only relying on support from others if they freely choose to co-operate with you, and never coercing nor allowing yourself to be coerced, is undoubtedly a noble one. Likewise, the call to live by fact and reason, not by superstition, is noble.

On the other hand, it is easy to be repelled by the lack of any requirement to help others: "You have no duty to anyone but yourself" (part 3, chapter 2, page 802). We need to look at what lies behind this view.

In his speech, John Galt says that we should not help people who claim our help as a right or as a moral duty that we owe them, but that it is fine to help someone on account of his virtues. The latter would be a trade, and the virtue of the person helped would be your reward. But helping those without virtue, even at no cost to yourself, would be "treason to life" (part 3, chapter 7, page 1,060). The ground for this last claim seems to be that helping those without virtue would corrode the foundations of society. We must insist that others live with the consequences of their own lack of virtue, otherwise people in general will sink into vice.

In the same speech, John Galt says, "A morality that dares to tell you to find happiness in the renunciation of your happiness - to value the failure of your values - is an insolent negation of morality" (page 1,014). In these words, he appears to expose an outright contradiction at the heart of the traditional morality of personal sacrifice that he attacks, in both its religious and its socialist forms. And he does indeed expose a contradiction, if and only if the first reference to happiness is, like the second one, a reference to one's own happiness. But that is a large point at issue. Should one's goal be one's own happiness, or the happiness of a wider class of people?

Rand could support the former option, by reference to her ethic of life. Random altruism would clash directly with that ethic. But we might not find her arguments for that ethic, in the form that it would need to take in order to yield opposition to altruism, satisfactory. And we should not fear that altruism would rob us of the admirable implications of that ethic, in particular the implications that we should strive, should achieve and should live by fact and reason. Those implications could be supported by other means.

It is rather disturbing that in the Valley, people do not simply give of their time and skill, even though the inhabitants are virtuous, and know one another to be virtuous. There is, for example, a charge for attendance at lectures and concerts, offered by some members of the community to others (part 3, chapter 2, pages 773-774). It is a point of principle not to give. One can see the sense of one's not being required to give, but a policy of not giving is decidedly odd. If it is based on anything, it seems to be on the attitude expressed by John Galt "that the unearned cannot be had, that the undeserved cannot be given" (part 3, chapter 2, page 798). If this were a claim that nothing could be consumed that had not been produced, it would be correct. But if it is meant as a claim about relationships between people, and not merely about the unforgiving relationship that nature has with humanity, then it is false.

Finally, even those who do not concentrate on selected aspects of Ayn Rand's philosophy, but who take all elements into account, may still be encouraged to take up an extreme position on its merits. The reason is that her philosophy is presented and perceived as a package. It is easy to think that one must accept it all, or reject it all. But that is not so. We are free to accept some propositions within her philosophy and reject others, so long as we find support for the accepted propositions other than the supports that are so bound up with the package that their use would commit us to accepting the whole package.

Computer-generated characters in films

The film Ted, which will appeal to all arctophiles, is about to be released in the UK. The star, a talking and beer-swilling teddy bear, is a piece of computer-generated animation. The human actors had to act as if he were there, then he was inserted after the human actors had been filmed.

Now suppose that human actors will be replaced in the same way, and that we will watch films with wholly realistic animated characters, such that human perceptual apparatus cannot tell the difference between them and human actors. That would take a lot of computing power, and some sophisticated programming, but it is perfectly possible that we shall see this development within the next 50 years, and perhaps sooner.

Philosophers of perception might find this an interesting new source of problems. It seems that there ought to be some difference in how we should describe the perception of the viewer, but it is not obvious how to characterize the difference. Disjunctivists, for example, could not capture the difference merely in terms of seeing human actors versus having an impression that was qualitatively indistinguishable from seeing them, because the stage of generating a mere impression that would be at issue would take place outside the perceiver's head. The fork that reflected the disjunction would have to be placed outside any specific perceiver, at the point of creation of the film, with one prong running back to real actors and the other to a computer that generated images. One might prevent this from being a problem for the philosophy of perception by regarding viewing a film as a species of seeing the objects filmed, but that would be a challenging course to take. However, it does not seem that this sort of problem would be specific to the portrayal of human beings, as distinct from the portrayal of other entities, such as mountains,  by computer-generated images.

Another question would arise in connection with the paradox of fiction. We get emotionally involved in films, even though we know that the characters are not real. If we knew that the characters were not even portrayed by real actors, would that affect our degree of emotional involvement? If the theory that we suspend our disbelief is straightforwardly correct, our degree of emotional involvement should be unaffected. If we suspend disbelief, we do not see the characters as portrayed by actors: we see them as real. Then the fact that actors had been replaced by animations should not matter. We might learn something from the extent of our capacity to get emotionally involved in cartoons that are obviously cartoons. The greater that capacity, the more likely it is that we would get fully involved in films where the characters seemed to be played by real actors, but were in fact computer-generated.

Finally, what would such developments do to the film industry? There would be many more films. They would be cheaper to make, and cheaper to amend after release in order to make small improvements, or to correct continuity errors. Films might be more finely honed to the preferences of audiences, because it would be cheaper to make extracts that could be tried on sample audiences before the final content of the film was determined. Such developments would be disruptive, but I think that on balance, they would be welcome, just as the word processor and electronic publishing have been disruptive, but have also brought great benefits.

Tax avoidance and the problem of establishing purposes

Tax avoidance is the reduction of tax liabilities by the use of contrived schemes, which are such that when the law is applied to them, the resulting tax liabilities are less than one might have expected. It differs from tax evasion, which is the reduction of tax liabilities by not disclosing the full facts, or by misrepresenting the facts.

The tax code of the UK is likely to have a general anti-avoidance rule, alternatively known as a general anti-abuse rule, or a GAAR, in the near future. The Government proposes to introduce one in 2013. Its consultation document on the proposal, published on 12 June 2012, and related documents, are available here:

http://www.hm-treasury.gov.uk/tax_avoidance_gaar.htm

(The link to the June 2012 document is about half way down the page, following the heading "GAAR consultation".)

The purpose of a GAAR is to remove the tax advantages that would be obtained by the use of contrived tax avoidance schemes,when the schemes are not defeated by specific provisions (for example, a specific provision that says that when an asset is sold between connected parties for less than its market value, the transaction is to be treated as taking place at market value). Governments are interested in GAARs because the counteraction of avoidance by specific provisions is an endless game of cat and mouse. Each year, new provisions are introduced to defeat known schemes. And each year, new schemes are devised to get round the existing specific provisions.

I shall not discuss the merits of GAARs here. Instead, I shall set out how some problems concerning the identification of the purposes of arrangements that have been devised by human beings arise in the context of GAARs.

It is common for anti-avoidance legislation to include a purpose test of some sort. The legislation may, for example, only take effect if a reduction in tax is one of the main purposes of the arrangements that the taxpayer has chosen. "One of the main purposes" is likely to be the phrase, rather than "the main purpose", so as to catch schemes in which tax avoidance is a significant purpose, but there is some greater commercial purpose that has nothing to do with reducing tax liabilities. For example, clause 2(1) of the proposed GAAR, discussed on page 13 of the June 2012 document, reads:

'Arrangements are "tax arrangements" if, having regard to all the circumstances, it would be reasonable to conclude that the obtaining of a tax advantage was the main purpose, or one of the main purposes, of the arrangements.'

The problem is this: how should we determine whether obtaining a tax advantage was a main purpose of some given arrangements?

The first point to make is that the challenge is not to find a way to look inside the actual taxpayer's head, and establish his or her subjective motives. The question is, "What is the point of these arrangements?", not "What was this taxpayer trying to achieve?". Some of the points that follow would have parallels in a discussion of the problem of establishing motives, but that is not the topic here.

Having said that, we cannot ignore the human element. Purposes are purposes of creatures that have goals. We can reconcile this point with the irrelevance of the motives of the particular taxpayer who in fact entered into the arrangements, by regarding the purposes in question as those of a hypothetical rational taxpayer who entered into the arrangements. Abstracting from the actual taxpayer's psychology, what would we say was going on, if all we knew was that some taxpayer had entered into the arrangements, and we were told nothing more about the taxpayer apart from his or her income, assets, and existing family or financial links to other people who were affected financially, along with any other facts that we would need to know in order to understand his or her tax affairs?

It is not enough for arrangements to be such that tax liabilities are lower than one would expect, for example when a profit is made, but no tax arises on it, nor will arise in the future. That is not enough because the lack of tax might be a pure accident. While the motives of the actual taxpayer may be irrelevant, the achievement of accidental consequences could not be seen as a purpose at all, not even a purpose of a hypothetical taxpayer, about whom nothing psychological was known. It is, however, most unlikely that such accidents would occur. Their occurrence would indicate that the tax system had been very badly designed.

It seems that we must look for indications that tax-saving features were included in the arrangements by design. One sign would be that a feature took some work to include, but that the other likely purposes of the arrangements (such as the transfer of an asset from grandparent to grandchild, or the purchase of some land followed by its sale at a profit) would have been achieved just as well without the feature, and without its replacement by another feature that would have taken a comparable amount of work to include.

This does seem to be the appropriate sort of way to proceed, if we are to abstract from the psychology of the actual taxpayer. We should compare what happened with other things that might have happened. The basis on which we would establish purposes could then be a principle of sufficient reason, something like this, where B is what the taxpayer actually did:

'If some commercial result was achieved by doing B rather than C or D, where all of B, C and D were possible, take it that the purposes of the agent must include enough purposes to explain the preference for B over C, and the preference for B over D, and identify enough purposes accordingly.'

Three problems with this principle as it stands are immediately apparent.

First, the purposes that would explain the preference for B over C might be incompatible with the purposes that would explain the preference for B over D. That would not, however, affect the workability of a GAAR, so long as the purpose of achieving a reduction in tax featured in both sets of purposes. We can imagine one hypothetical taxpayer preferring B to C, and another one preferring B to D.

Second, there might be some differences between B and C, or between B and D, about which the hypothetical taxpayer would be indifferent, making it inappropriate to seek explanatory purposes. If there were other differences which meant that the preference for B could only reasonably be explained by a purpose of achieving a reduction in tax, that would not affect the workability of a GAAR. But if all of the differences were ones about which the hypothetical taxpayer would be indifferent, while B was nonetheless the choice that happened to reduce tax, legislation which ensured that a GAAR still applied would have the same effect as legislation that required people deliberately to arrange their affairs so as not to make tax savings. Such legislation could reasonably be argued to be excessive.

Third, the principle would not always be enough to allow us to work out whether a GAAR should apply to take away a particular tax advantage. There might be a range of possible sets of purposes to impute to our hypothetical taxpayer. Each set would be sufficient to explain his or her preference for B, but some sets would include a purpose of achieving a reduction in tax, and others would not include such a purpose. This problem would become more acute if we were required to determine, not only what the hypothetical taxpayer's purposes were, but which ones were main purposes.

These difficulties may not be insuperable. But if they are not addressed explicitly in the guidance that we are promised will accompany the UK's forthcoming GAAR, it may not be easy to discern a consistent line of thought behind the decisions that judges will make, once cases start to be litigated.