(i) Analysing intention

Philosophical perplexity about intention revolves around three things: future intention (“I intend to type up a section of my paper tomorrow”), intention with which I act (“I type with the intention of writing a section of my paper), and intentional action (“I type intentionally.) The principal task of the philosophy of intention, as I understand it, is to uncover the unity between these three things.^{Footnote 8} This part of section (1) has two subparts. The first examines what an intention actually is. The second explores the relationship between intention and autonomy, and questions of agential authority.

This part of the essay is avowedly introductory; it aims to lay the groundwork for later sections. In particular it describes for example how intention functions as a planning tool and how it is future-oriented. That future orientation is something we return to when we examine mispredictions in section 2(A)—predictions (say) that A will happen next week only to find B happens. Intention is for Bratman integral in practical reasoning.^{Footnote 9} He suggests that explains a methodological priority for future-directed intention. While I am eating breakfast therefore it is fair to say that I intend to start my car later to go to work. Once I am in the car, I am starting it; it now seems odd (at least colloquially) to talk of intending to start it.^{Footnote 10} This feeds into Bratman’s discussion of plans. He argues that we are all planning creatures. This is important for a number of reasons, not the least of which is that planning structures help to support forms of unity that are central to what Bratman calls “cross-temporal self-governance”. Essentially planning allows us to co-ordinate with our future selves.^{Footnote 11} My plan to write this paper is constant over time. It allows me to organise myself over time so that I first read Bratman’s articles with a view to understanding his theory and with a view to at the very end of the process tidying up the footnotes for this journal. Another important point is that we never plan everything out completely. Rather we decide to go to Bruges for the weekend say—and then decide how to achieve that at a later date—get the Eurostar perhaps. Our plans are partial and so must be fleshed out in the future.^{Footnote 12}

One question that arises is how cross-temporal self-governance ties into the relationship with doing something intentionally. Bratman takes the view that when I intentionally A I must intend to do something, but I need not necessarily intend to A;^{Footnote 13} he asks us to suppose then that there are cases where I A intentionally, but where I intend to B. In order to explain this he says there must be an account of the types of actions that may be performed intentionally in the course of executing a given intention. In doing so, he examines what he calls the simple view of intention, which is the view that where I intentionally A, I must A with the intention of A-ing. Bratman suggests that it is in fact relatively easy to show that the simple view is incorrect.^{Footnote 14} He takes an example of throwing missiles at targets. There are two targets and it is impossible to hit both. If I hit one the other shuts off too, and if the machine detects I am on target to hit both simultaneously, one shuts off.^{Footnote 15} If I know this, I cannot intend to hit both without being irrational—my intentions and beliefs are not consistent. Clearly, giving both a go is perfectly sensible, yet the simple view tells us that if I succeed in hitting target 1, I do so intentionally. If I did so intentionally I must have intended to hit it, and symmetry demands the same intention apply as regards target 2. However, this seems to place me in a form of irrationality; I must intend to hit whichever one I hit, and therefore at the time I try to hit them I must intend to hit both. Yet I know that is impossible.^{Footnote 16} The simple view is wrong.

This, however, allows us to define the motivational potential of an intention; if I step onto a train in Norwich intending to go to London Liverpool Street, but get muddled and mistakenly step onto the Liverpool Lime Street train, I intentionally step onto the train and given my belief that it will go to London, part of my motivational potential for getting on the train is my intention to go to London. I step on the train intentionally, but the intention “with which” I do so is impaired. It is this intention “with which” we are concerned about in our discussion of mistakes. This is also consistent with the difference between objective and subjective impairment of our intention. If my intention is subjectively impaired, say by duress, the law provides that the impaired transfer is valid, but can be rescinded.^{Footnote 17} If it is objectively impaired, so I did not know there was a transfer, I retain legal title to the asset.^{Footnote 18} We can say that I did not transfer the asset intentionally. Mistake is controversial, although in Pitt v. Holt,^{Footnote 19} Lord Walker said that for the equitable jurisdiction to set aside voluntary disposition to be used the donor must make a “causative mistake of sufficient gravity”, which will normally, in what he said was “additional guidance”, be satisfied only if the matter is basic to the transaction, or as to its legal character or nature.^{Footnote 20}

There are three norms that for Bratman are characteristic of intending, which, as we see later in section C (i), would be affected by a mistake. Our intentions must be means-end coherent; they must be consistent with my beliefs and my intentions must meet the requirements of agglomerativity. I take means-end coherence first. Means-end coherence involves the idea that the means by which I intend to do A must be a method I believe will enable me to do A.^{Footnote 21} I am typing with the intention of writing a subsection on intention, but if I were to paint a landscape with that intention, my intentions would not be means-end coherent, because painting a landscape does not help, nor do I seriously believe it will.^{Footnote 22} To be means-end coherent at any given time means that my intentions must include sub-plans to accomplish them. The second characteristic is intention-belief consistency. A good co-ordinating plan must also be such that it is a plan for the world in that I should be able to realise while all my beliefs are true. The irrationality of intending something one knows to be impossible, to which we have already alluded, flows from this.^{Footnote 23} I might without being irrational want to play football and finish this paper, but know I won’t have time to do both; I cannot, however, rationally, knowing that, intend both.^{Footnote 24} The third criterion is agglomerativity.^{Footnote 25} This requires that if the agent intends A and also intends B then she intends A and B. It must therefore be possible (as she believes) to both A and B. If that is not possible, the agent will have to decide whether to A or B. It should be possible to agglomerate one’s various intentions in a way that creates a larger intention. In other words just as incomplete plans require sub-plans to carry them through, so sub-plans must be capable of forming a larger plan.

(ii) Intention and agential authority

We are used in the philosophy of action to drawing a distinction between two different kinds of action; actions that are governed or directed by the agent—things she does—and are therefore truly attributable to her and those that merely happen to her.^{Footnote 26} It is sometimes said that a policy that has agential authority speaks for the agent. This is jargon, but refers to the fact that the agent directs or governs the action and so the policy must be what Bratman refers to as a self-governing policy. The critical thing is that we have desires about our actions, but also the capacity to step back and reflect on whether to act on them. It is those higher order self-governing policies which say which desires we treat in our deliberation as justifying reasons for action,^{Footnote 27} and which play a role in our practical reasoning as to what to do next and why. For example I might desire a cigarette, but critically reflect that I ought for health reasons to give up. That policy to improve my health has agential authority because it directs and governs my decision to give up smoking.

This idea has importance because our intentions, as a result of their critical role in our lives as planning agents, look not just to the present but to the future as well. We formulate intentions on the basis of assumptions and conditions some of which relate to the future and some to the immediate present. At each point we are concerned with the activity as a whole, including its future and its past components and in a future section we examine the way in which this is constituted philosophically in much more detail. A planning agent’s purposive activity is therefore an interwoven structure—hence the need for means-end consistency and agglomerativity—of partial, interlocking and more or less stable plans.^{Footnote 28} When I start reading articles by Bratman I am not just reading articles, I am preparing for the research process and thinking ahead to which journal to publish in, but I have not necessarily decided which journal. The plan is partial, but it is stable—after all you are now reading the essay. It seems that temporal extendedness is a deep feature of our agency, and the question arises whether there are higher order policies that support such agency.^{Footnote 29}

The answer seems to be yes. For Bratman our higher order policies are critical in helping us structure our intentions over time and he argues this is an important part of an account of agential authority.^{Footnote 30} Such self-governing policies with agential authority are policies that are general in their content and help to structure our intentions over time by keeping our current intentions consistent with those in the recent or not-so-recent past, and those regarding the future.^{Footnote 31} Bratman takes a broadly Lockean view of personal identity, one which is accepted by this paper, and which identifies a person as a thinking being who can know itself as the same thinking being across time.^{Footnote 32} We are not time-slice agents; we do, as indicated earlier, plan partly on the basis of the future, and this provides an important reason for adopting the Lockean view that Bratman does. Our general higher-order policies as to which desires and motivations we treat as important help to constitute that identity over time. My desire to do research is consistent over time, helps constitute my identity as an academic, and, as an appropriate reason for action, provides a justification for the work going into this paper. When I engage in practical reasoning, deciding what to do today for example, my stable higher order belief in the value of research helps to make sure that today I intend to redraft the paper, just as yesterday I intended to read Bratman’s new book, and further that I re-draft on the basis of the ideas in the book, having yesterday read the book with a view to understanding if there were new ideas to incorporate.

It seems likely that the agent must be aware of these higher order attitudes and further be satisfied with them so he is not seeking to change them continually.^{Footnote 33} Those attitudes will not be able to form a stable platform for the development of intentions if they change all the time. That is not to deny that our desires and preferences do in fact change over time, however. They need not therefore be immutable. There is a defeasible presumption against reconsideration of the attitudes, but that presumption against reconsideration still means that I am able to construct stable plans that add to my satisfaction with my life across time; I value not just that this paper get finished, but more precisely that I finish it.^{Footnote 34} Lack of such stability may lead to conflict with prior plans, and unless the conflict is resolved to what Bratman refers to as “cross-temporal incoherence”, so that I intend mutually inconsistent things.^{Footnote 35}

In summary we can model agentially directed practical reasoning as practical reasoning and planning guided by authoritative and stable self-governing policies. There is therefore a hierarchy of norms of differing levels of generality which help govern what I do, as opposed to what just happens. By contrast Bratman suggests that there can be cases where one’s thinking is in the grip of a norm.^{Footnote 36} Here practical reasoning is somewhat attenuated, so I might decide to do whatever the person in charge says. Soldiers in battle often act in the grip of a norm and not really for themselves. Often this type of thinking is simply ignored by the law, but there may be cases where it indicates duress or undue influence where the law does take cognizance.^{Footnote 37}

(i) Internal and external conditions

Conditions may be internal or external to a party’s intention.^{Footnote 38} A condition is internal when it is part of the content of the intention, and external when it is a condition of the formation of the intention. External conditions therefore take the form “If X, Z will intend to Y” and we can put them to one side. As J.P.W. Cartwright rightly argues, a claim that intentions are conditional is generally seen as a claim of internal conditionality.^{Footnote 39} In some cases there are explicit conditions in the actor’s mind; I might think “If it is sunny tomorrow, I will go to the beach”. However, just because I have not formulated explicit conditions does not mean I have an intention to act “come what may”.^{Footnote 40} In practice this type of completely unconditional intention is vanishingly unlikely and in theoretical terms problematic, as Luca Ferrero explains.^{Footnote 41} As circumstances changed and we formed new intentions it would be impossible to keep them in synch so that our plans were rationally capable of being carried out; our intentions would not be agglomerative. Ferrero uses the example of Don Giovanni and suggests that he is set on pure unconditional pursuits; however, if he is set on several of those how does he handle conflicts? Because of their unconditional structure there is no plan for conflicting intentions.^{Footnote 42} What if I unconditionally intend to play football and unconditionally intend to finish this paper but there’s only time for one? Pure unconditional intentions need to be atomistically separated so that each project is isolated. We are not that kind of agent.

There will therefore be explicit and implicit conditions on our intentions all the time. It is impossible to outline a complete list of contingencies to cover any eventuality and this makes for a difficult question as to how to describe my conditional intention. I might say my intention is subject to the condition “if nothing prevents me”, or a condition “if I can”. These are not true conditions though.^{Footnote 43} My capacity or ability to do something is not a reason to do it; it is merely a necessary condition. Even the falsity of the condition does not prevent my having the intention; I might intend to A but find myself unable to in fact A. We might also say I intend to A “if nothing happens that might make me change my mind”, but that would be simply discontinuing the intention.^{Footnote 44} To deal with this problem, Ferrero refers to the deep structure of our intentions. This includes all the conditions that qualify the content of the intention whether generic or specific, express or implied.^{Footnote 45} It might not be immediately clear to an agent that a condition exists, but it may still play a role in her psychic economy and emerge when required even if dormant in the recesses of her mind. Identifying the conditions is a work in progress.^{Footnote 46} It is true that Ferrero introduces questions of feasibility and advisability as generic placeholders for specific conditions that although not immediately operative could pop up, but these are not changes in mind for Ferrero, they are rational pressures integral to the intention.^{Footnote 47} A conditional intention is not suspended. It exists now and the agent is to prepare himself now for doing A when and if the time comes.^{Footnote 48} To claim an intention to go to Bruges unless I change my mind produces no rational pressure to prepare to go. What is the point in buying a Eurostar ticket if I might change my mind?

(ii) Background and foreground conditions

Klass argues that a background condition is one where the party accepts its satisfaction or non-satisfaction as a given, whereas he does not do the same for foreground conditions, which the agent recognises are open questions.^{Footnote 49} Once a qualification is taken as a given, it is not erased from the content of the intention. The intention does not become unconditional. Rather the condition is pushed into the cognitive background, and the agent need no longer entertain it in his practical reasoning.^{Footnote 50} However, if the intention were to become purely unconditional on the agent’s coming to believe the condition satisfied he is committed to acting even if he later discovers he was mistaken.^{Footnote 51} That is unlikely. It is important that the conditions remain. If we are talking about a future intention the agent will need to keep them under review in order to decide whether revisions to her plans are required.^{Footnote 52} The deep structure of the intention does not change. However, what does change is its epistemic version, which reflects the changing beliefs and acceptances of the agent concerning the conditions. Let assume I think I owe John $10. I know I will see him on Monday, so I form the intention to pay him on Monday. It is now a background condition of my intention that I owe the money. However, if I discover on Sunday that in fact I do not owe the money, I will reconsider the decision to pay the following day, something I would not do if the intention became purely unconditional.

Before moving on, we also need to recognise that the distinction between belief and acceptance is important—and one way in which it is so we see when examining collective intention in 2(B)(i). It is possible for us to accept things, for the purposes of our practical reasoning, that we do not necessarily believe 100%.^{Footnote 53} For the purposes of buying a house I accept it will not burn down tomorrow (although if pressed I would concede that an arsonist might pass by), but for the purposes of getting insurance I do not simply accept that it will not burn down. My acceptance that it might burn down is precisely why I buy insurance.

(i) Mistakes of fact

If it is a background condition of my intention to pay John $10 that I owe the money, I cannot deliberate on what I might do if I were free of this obligation. Of course I might think, “It’s such a shame I have to pay John $10; I could go to the cinema if I did not”. Yet what I cannot do—consistently with maintaining the intention to pay John—is spend the $10 bill. I have to un-make the intention first; I must change my mind.^{Footnote 54}

The justification of a mistake claim is sometimes put in terms of my intention being vitiated, and we have referred to this before and put the word in inverted commas.^{Footnote 55} We can now see that there is nothing in fact wrong with my intention. The inverted commas were appropriate. I do indeed actually intend to make the payment; it is the motivation that is the issue. As Charlie Webb argues, the justification for relief is that the intention (with which) simply does not cover the case that actually arises.^{Footnote 56} This matters because if the law does not permit recovery of the money I paid to John on the basis I owed him $10 when in fact I did not, it disrespects me as a self-legislating autonomous agent. Autonomy refers to my capacity to act in accordance with, what Aruna Nair refers to as, my own “rationally determined norms”.^{Footnote 57} As Nair puts it, a mistake as to the relevant facts may cause (indeed rationally compel) a person to act in a way that violates his own norms.^{Footnote 58} Following the intention leads me to paying money I do not owe, an action inconsistent with the higher order policy only to pay what I owe and no more, a policy which has agential authority for me. In the same way, we can analyse the facts of Barclays Bank v. Simms ^{Footnote 59} as follows. Barclays paid out on a cheque. They had not appreciated that the cheque had been stopped. Had it not been stopped, the cheque would have created an obligation on the bank to pay. It was therefore a background condition of the bank’s payment that they had an obligation to pay. There was no obligation to pay, but the bank’s ‘intention “with which”’ only covered the case where the cheque was payable.^{Footnote 60} Had the bank realised this, its intention to pay—had it been maintained—would have been means-end incoherent, because of intention-belief inconsistency.

Bratman’s view therefore elucidates the structure of the claim further and is consistent with the justification offered, but we may still think it insufficient as an explanation. The puzzle is this: does Nair really explain why a defendant who has done nothing at all should be liable? In other words even if the agent has not acted autonomously why should the recipient care? Even if we could not think of an answer to this, Bratman would still enable us to better understand the structure of the psychic states leading to a mistaken payment and we would still have made some progress. To answer the puzzle fully would require another paper but some comment is in order. The normal answer is that the defendant is no worse off and since no harm is done to him he cannot really have any objection,^{Footnote 61} but this may simply beg the question of which baseline we use to decide that the defendant has not been harmed and why such a baseline is appropriate. Certainly there is an almost universal intuitive sense that the payment must be returned, but the admitted difficulty remains in articulating a justification for that intuition. It may be entirely pragmatic. Once we have decided that something in the payor’s intention is awry, this is simply the easiest way to sort it out, and any harm to the recipient is worth the cost. But again many will find this unsatisfactory.

(ii) Mistakes of law

Nair criticises my definition of mistakes of law, which characterises them as mistakes as to the interpretation of the authorities. I know (sometimes at least) what the relevant cases and statutes are, but I misinterpret them to require something they, properly understood, do not.^{Footnote 62} Nair says that that makes a mistake of law what she terms “a normative mistake”. She argues on such a view when I pay under a mistake of law, I am mistaken as to the norms I should apply. If the court requires the recipient to repay it substitutes its own norms for my norms. It thereby violates my autonomous right to decide what those norms should be.^{Footnote 63} She uses the example of a person who believes anyone who asks for money should be given it. If a court reversed a payment on the ground the norm was wrong, it would violate his right to choose on what basis to give out money, and Nair seems right that no recovery should be allowed.

We saw earlier, however, that there is a hierarchy of norms. This is part of our account of agential authority, and Nair does not deal adequately, or indeed at all, with the hierarchy of norms in her account of mistaken payments. Bratman’s views help us integrate that hierarchy into our account of the claim and therefore better understand what it means in this context to talk about acting in accordance with our own rationally determined norms. To recap, I have higher order policies which have agential authority for me; these tell me which of my desires or motivations are appropriate reasons for action in my practical deliberations. There may well be several layers of such norms, of differing levels of detail and specificity. One of my higher order norms is that I will act in accordance with the law. That higher order norm then leads to two lower norms. The first is that I pay my creditors what I owe them and no more, which on interpreting the law I decide is $1000; I believe the contract to be valid when it is void. There are three norms in play therefore. There is the normative proposition of law (which is incorrect) that I owe $1000. There is the policy to pay what I owe and no more and the still higher order policy to obedience to the law. My reasoning process here is different from the case that Nair introduces. In the charitable giving scenario there is no disconnect between my higher order desires and my immediate intention. I have a higher order policy of generosity, which tells me that when John asks for $1000, that is a justifying reason for forming an intention to pay. Here by contrast there is such a disconnect. My intention to pay is, as Nair correctly says, based on a normative mistake, but if I pay I am not acting in accordance with the higher order policy in favour of paying what I owe (and no more) that treats my misinterpretation as a justifying reason for forming an intention to pay and acting on it. In this case, but not in a charity case, my payment puts me in violation of a higher order policy with agential authority, and does so in the same way as a factual error.

Nair’s example is one of a normative moral error—the court would consider it a moral error to be so generous.^{Footnote 64} It is possible to conceive of moral mistakes if there is an objective measure of moral obligation. Ronald Dworkin argues there is.^{Footnote 65} He must do so; there is an element to his theory of law which examines the moral justification of the law. Nonetheless we are intuitively less happy with relief for moral mistakes and therefore with Nair’s example above. This may indicate no more than that one of our higher order policies is that we act in accordance with externally generated legal norms, and this may derive from the presence of a legitimate law-determining authority,^{Footnote 66} but the absence of an equivalent for morals, which is what for Dworkin delineates legal from other political rights.^{Footnote 67} The importance of this is that it allows the court to substitute its norms for mine, because one of my higher order policies, ones which have agential authority for me, is that I will submit to the views of such a legitimate legal norm-determining and adjudicative authority.

(i) The Nature of temporally extended agency

We have already seen that Bratman argues that the importance of the notion of intention lies in part at least in our ability to make plans which allow for cross-temporal co-ordination with our future selves. This ability is predicated on general self-governing policies which have agential authority for us and these policies constitute us as thinking beings over time. This Lockean view of personal identity seems, as we said earlier, very plausible, and we try to expand on the idea in this section to understand the effect of different kinds of condition on our future-oriented intention.

There are typically said to be two aspects of the will—the reflective and diachronic. J David Velleman raises the former and argues that by making up our minds before we act we ensure that our performance is neither automatic nor unreflective.^{Footnote 68} In order to make up our mind we reflect on the reasons to act in a particular way as to any other way. Having predicted that we will act in a particular way, our desire for self-knowledge takes over and we intend to act in that way and that intention is a reason to act because otherwise we do not increase our self-knowledge.^{Footnote 69} This is reflective and it also reveals a cognitivist view of intention.^{Footnote 70} The diachronic dimension relates to future-directed intentions, and is critical for the argument made in this paper.

Ferrero addresses the issue.^{Footnote 71} For Ferrero, Velleman ignores the diachronic (future-oriented) dimension in that he does not appreciate that there are aspects to the will that cannot be explained through his reflective model. In short if I decide to go to buy mince pies, then at each moment I might be said to be reflectively acting to gain self-knowledge, by putting my coat on, getting in the car, checking I have my wallet etc. But this type of moment to moment agential governance is simply inadequate for Ferrero^{Footnote 72}—and rightly so. We are not, as we have seen before, that type of time-slice agent. Ferrero argues “at each moment of the activity’s unfolding, one sees oneself as autonomously and continuously engaged in the activity as a temporally integrated unit”.^{Footnote 73} Indeed he claims that without the diachronic will we could not even conceive of the goal of getting the mince pies in the future.^{Footnote 74} Ferrero goes on to discuss what he calls the internal unity of the agent. He suggests that the diachronic will is necessary not merely to engage in temporally unified activities, but a temporally unified existence. To be a temporally integrated agent, one’s existence over time must be more than an uninterrupted line of physical connections. There must be an overall arrangement of some sort in which the agent at all times accepts the global constraints on their action in terms of their activities and the coherence of their intentions with each other over time.^{Footnote 75} If we accept constraints over time on our actions it would be unreal to suppose that we never act on the assumption of something to occur in the future. It is unreal to think therefore that when I start reading articles by Bratman, knowing that one of the constraints on my finishing this essay will be that the teaching allocation respects the 40% of time for research rule, I do not assume that the Head of School will work on that basis, even though I realise he has not yet begun the teaching allocation.

Ferrero’s discussion of the diachronic will is reflected in Bratman’s insistence on intention as critical to our existence as planning agents. If intention has as its aim to make its content true, we can see this in two ways. Either we can see intentions atomistically. That is we can see intention in such a way that each intention aims to make its aim true. Or, as Bratman does,^{Footnote 76} we could see matters more holistically. Each intention would aim to make its content true as a part of a coordinated realization of one’s planning system, in the world as one believes it to be. Bratman explains this in terms of a projected unity of agency, which provides a rational force towards the agglomerativity of intentions. As well as valuing research, I value teaching, and so planning my future work involves dividing time between this paper and my teaching preparation, so that both get done and done properly.

We see our entire project of activities as a whole not just as time-sliced bits and pieces. The importance of the diachronic will can be established by demonstrating that there are activities that can only be pursued in this mode.^{Footnote 77} Rational discourse is one of Ferrero’s examples. Where an agent makes a claim, lays out an objection or an argument, or offers further explanation, she is not merely engaged in saying things, which just happen to come out of her mouth one after the other in no particular order. She must make sure they are consistent and coherent with things said in the past and also prepare rebuttals if her interlocutor puts a counter-objection to her.^{Footnote 78}

For Fererro agents like us can only achieve the aim by engaging in internally unified activities. The diachronic will is indispensable to overcome the difficulties of our being temporal beings with limited resources and rationality. There are differences with Bratman then but they are not huge. Ferrero argues, however, the norm of agglomerativity might be stricter once we take temporal unity into account. This is because the diachronic agent sees both the activity and his agency as extending over time into the future; there is pressure for greater stability of intentions over time, and therefore also of self-governing policies.^{Footnote 79} Ferrero argues that a structure such as Bratman’s requires only very short time spans for the cross-temporal ties between parts of specific plans. He believes that the stability of the higher order policies is at best fragile on Bratman’s view, but if we are temporally integrated the stability of the policies is greater.^{Footnote 80} Whether Ferrero’s criticism of Bratman is justified or not, we can conclude that the broadly Lockean view of identity accepted here requires that the agent maintain stable policies and certainly if the agent is continually looking into the future the plans he makes will be reconsidered frequently if those policies are insufficiently stable, and will be reconsidered frequently if the agent does not assume or accept certain conditions as to the future will be satisfied.

If this is true, we can say that in acting to buy building supplies now as part of a plan to build an extension to my house, for example, which can only be successful if a builder can be found, in buying the bricks I must act background conditionally on being able to find a builder. This is all part of our ability to co-ordinate with ourselves over time. If we were unable to accept things as to the future, we could not do many of the things we do. Planning agency is always future-oriented. On this basis it is not easy to see why accepting a future condition as satisfied for the purpose of putting our plans into action should not be treated the same as accepting a present condition as satisfied.

(ii) Unilateral mispredictions

How should we treat mispredictions, relating to a possible future event as opposed to a present or past matter of fact or law?^{Footnote 81} If from a philosophical standpoint it seems to make little difference because our intentions as part of wider plans always have an element of futurity to them it is unclear why there should be no relief. After all if the bank is able to recover in Barclays Bank v. Simms because the background condition that the money was owed was in fact unsatisfied because the cheque had been stopped before the payment was made, and therefore the bank’s intention did not cover the case as it turns out to be, the same could be said, for example, about an assumption that liability will accrue in the future.

The case law muddies the waters somewhat. In Re Griffiths,^{Footnote 82} for example, Griffiths made a disposition of property and died the next year. Lewison J. held that the gift could be revoked on the ground of Griffiths’ mistake that he was not suffering from cancer.^{Footnote 83} This characterisation can be criticised—Lloyd L.J. in the English Court of Appeal in Pitt v. Holt thought it was a misprediction as to how long he would live for.^{Footnote 84} The orthodoxy suggests that if characterised as misprediction, no relief is possible. Yet given that mispredictions are based on beliefs about the present—here a mistake as to whether Griffiths was ill led to a misprediction as to his life expectancy—it is easy to re-characterise mispredictions in this way. That said assumed future obligations are, or seem to be, treated differently. In Kerrison v. Glynn Mills Currie and Co ^{Footnote 85} the claimant was accustomed to putting the Kessler & Co in funds to honour bills of exchange drawn by a Mexican mine (Bote Mining Company). The New York bank went insolvent and the claimant had already deposited a sum of money to their credit with the defendant bank on the assumption that liability would arise. The money was recoverable.

Weeliem Seah argues that relief for mispredictions is barred for two reasons. First, the decision making of the claimant is not impaired and secondly he is a risk taker. As for the first, we have already seen that the planning function of intention and the way our agency spans past present and future entails that we must, in order to plan properly, accept (in context) certain conditions as bound to be fulfilled. It is not obvious therefore that (in context) the claimant’s decision-making was unimpaired, although I will discuss this below. As for the second, Seah says the future is always uncertain and therefore the claimant must realise it might not turn out as he predicts.^{Footnote 86} I will be subjectively taking a risk that the event or condition will not take place. But I need not do so subjectively for the rule to kick in. Maybe the claimant is an objective risk-taker. There is a risk of things not turning out that way and adverted to or not that bars relief. The law has been unwilling to accept objective risk-taking in mistake cases.^{Footnote 87} Nonetheless Paul Davies argues an objective approach is intuitively more appropriate.^{Footnote 88} He uses the example of a busker. The busker wants to be paid but takes the risk of getting nothing. Yes, but a misprediction as to the behaviour of others gives no relief because they have no opportunity to refuse the busking. The example is flawed; talk of risk is conclusory; the busker (or mispredictor more generally) only takes the risk because we have already decided there is no right to be (re-) paid.

The type of risk-taking about future events I have described here (if it is risk taking) is, however, simply endemic; planning would be impossible without it. It may be precisely the endemic nature of this type of risk-taking that justifies non-relief. Further what keeps us to our plans is anticipated future regret; that is we know that if we do not act now we will regret it later. I might know that if I have a second glass of wine and am unable to work afterwards that I will regret not finishing this paper. Despite my really wanting more wine, that future regret stops me.^{Footnote 89} Anticipated future regret seems more easily discountable because there is always the possibility, given that our intentions and higher order self-governing policies are not immutable, that those policies will change and we withdraw from the plan. Secondly, this tends to explain why a misprediction as to future legal liability might more easily justify relief; the court simply takes it as read that our policy in favour of paying what we are obliged by law to pay will not change, yet it does not take it as read that when I pay on the basis of some other misprediction that I would not have withdrawn from the plan before the predicted event happens (or does not). This argument may, however, be somewhat tenuous as I complain only after the predicted event fails to happen so we might say that in point of fact I did not withdraw from the plan; at the point at which I acted that could not, however, have been said for certain and that may be of some importance.

(i) What is collective intention?

This is rather difficult. There are different views on this. In particular we can contrast the work of Michael Bratman with that of Margaret Gilbert and others. Gilbert rests her idea of collective intention on a plural subject and the group’s emulating the action of a single subject.^{Footnote 91} For Bratman collective intention arises in those scenarios where we can say “We intend to J”.^{Footnote 92} That—for our purposes—involves two interlocking individual intentions, such that I intend that we J, and you intend that we J and we each intend to do our part in J-ing knowing, or at least reliably predicting the other also intends that we J and will intend to do and actually do their part in the plan. In most cases of restitution for failure of basis we will be concerned with some form of shared goal: an interest rate swaps agreement, building a house, arranging the purchase of a company.

At no point in his account of, what he describes as, modest sociality does Bratman resort to what Margaret Gilbert has called a plural subject. Gilbert rests her idea of collective intention on a plural subject and the group’s emulating the action of a single subject.^{Footnote 93} For Bratman this violates what he calls the “own-action condition”^{Footnote 94} which is that the subject of the intending should be the agent of the intended activity. The idea behind the own-action condition is simple; I cannot intend your action. There is no metaphysical “we” separate from the two distinct “I’s”. This seems intuitively plausible; however, Bratman’s own view also violates that condition but does so in an acceptable way. It does so via a mechanism that Bratman calls “other agent conditional mediation”. That describes the mechanism by which I intend that we do something. In intending that we J, I believe that my intention will lead to our J-ing, and that we would not J if I did not so intend. And, normally, in intending that we J you believe likewise. We can mesh these by saying that I believe that my intending that we J is a cause or reason of your intending that we J and therefore of your doing your part in J-ing. Consequently Bratman is able to draw on what we know of ordinary intentions in order to understood apparently shared intentions. His view then is that

• – We each intend that we J;

• – For each of us the persistence of the intention that we J depends on continued knowledge (or acceptance) that the other also intends that we J, thus leading to its being common knowledge that we each intend that we J; and

• – if and only if we both so intend, then will we intend to J and so in fact J.^{Footnote 95}

A bilateral transaction, this paper suggests, is one where this holds good; failure of consideration can then operate whenever there is such a putative collective intention and only in such cases, because only in those cases is the other party’s autonomy engaged. Importantly for Bratman any constraints in terms of my being forced to mesh sub-plans with yours or find ways to do my part of J are purely rational norms, but your intentions and plans pertaining to our J-ing have agential authority for me because of what Abraham Roth describes as my bridge intention to mesh my J-related plans and intentions with yours.^{Footnote 96} As far as possible, Bratman argues that collective intention is subject to the same rational constraints as personal. For there to be a shared intention Bratman suggests we have the normal rational pressure in favour of means-end coherence, consistency and agglomerativity mentioned earlier.^{Footnote 97}

Your intention that we J has some agential authority for me in that it helps settle what I am to do; in its absence I become free to decide what to do. What I am to do is no longer settled. Consequently there must be a connection such that each of us is responsive to the other in ways that track the fulfilment of the end goal.^{Footnote 98} Roth disputes this; he argues that all I need to do is to stop intending. Because your commitment to our J-ing is partly dependent on me, you must stop intending that we J. If you’re not committed, neither am I.^{Footnote 99} Further Roth argues that I never was committed. He argues this by referring to a peculiar type of self-referential intention—“I intend to A on condition that I intend to A”. He suggests quite correctly that this does not settle the question of whether I am to A. Effectively this explanation of collective intention falls into the same trap. My intention is referential on your intention which is referential on mine (and I know this); consequently I intend to A if (on condition that) you intend to A if (on condition that) I intend to A. Everything is circular and there was no collective intention in the first place.

Bratman would reject this and it tells us something interesting about how he would view a failure of basis claim. For Bratman in order for me to intend the joint activity I must believe, or be able to reliably predict that the other participants intend it or will do so in due course.^{Footnote 100} I may therefore intend that we J even if you do not; my intention that we J is not the same as our intending to J, although it is a component. I can, he suggests, settle for myself whether we J by making an appropriate prediction, one of which might be that my intention that we J will lead you to intend the same. This is an aspect of the idea of “other-agent conditional mediation”, which we have encountered already. I intend that we J, because I believe or accept that my intending that we J is a cause or reason of your intending that we J and therefore of your doing your part in J-ing. Let’s say I want to dance the tango at a party with Maria.^{Footnote 101} Maria is absurdly indecisive and will not say that she will dance with me when I ask before the party. I rely on her intending or coming to intend to dance with me. I also need to use my reliance as an element in my practical reasoning in deciding for example whether to wear my dancing shoes to the party.

This might make it appear as if Bratman’s collective intentions are extremely fragile. Yet they may not be as they appear. First, although Bratman does not require that there be moral norms between the parties for there to be a shared or collective intention, he accepts that such assurance-based moral obligations are useful in order to increase the stability of the shared intention over time. This gives rise to what Bratman calls obligation-based interdependence.^{Footnote 102} In most cases with which unjust enrichment deals we have an agreed shared plan, personal intentions to perform one’s own part in that plan, which are derived from the joint intention, and strong intersubjective obligations which are capable of surviving a change of mind.^{Footnote 103} Those obligations may be purely moral, but they may also be contractual and legally binding. In either case, Raimo Tuomela argues that acceptances of the joint plan must be communicated so that everyone knows that the others accept the plan.^{Footnote 104} That mutual knowledge must be out in the open; we have seen above that Bratman’s analysis of joint intention involves common knowledge or acceptance that the other party intends that we J, and the other party knows I believe this. Communication is one—the only according to Tuomela—way in which such knowledge can arise.^{Footnote 105}

To explain where such moral obligations come from, Bratman appeals to Scanlon’s moral principle of fidelity, which was developed in the context of explaining why promises are not exclusive in generating such moral obligations.^{Footnote 106} These moral obligations to keep to the plan are generated via the other party’s intentionally induced reliance,^{Footnote 107} and this is consistent with at least some views of the basis for contract. Dori Kimel suggests that contract and promises fulfil the same instrumental function of facilitating a form of reliance.^{Footnote 108} Randy Barnett posits that the obligations of contract are generated through the party’s consent,^{Footnote 109} but the objectively apparent consent is what is important precisely because it gives the other party a reliable meaning of the statements to base their own conduct on. Secondly, Bratman posits a type of constructivism,^{Footnote 110} whereby individual participants might be driven by norms of individual planning agency, but the content of the intentions and their inter-relationship bring with it certain social norms which the parties internalise and use in their practical reasoning. In other words each of us use the fact we accept the other intends that we J as a factor in our own future planning as to how we J and what we do after J.

(ii) Failure of Bratmanian basis

At last we get to the point where we can ask what impact the non-satisfaction of different background conditions has on a collective intention and how that can—if at all—be seen as a legally effective failure of consideration. It is worth stressing again that we do not need to subscribe to any particular justificatory theory to accept Bratman’s contribution to our understanding of the claim’s structure, although an explanation based on respect for our autonomy in the sense outlined above is a plausible one, which we accept. The failure of basis might be the failure of a contingent agreed (non-promissory) future condition, as in Chillingworth v. Esche.^{Footnote 111} It is always possible in Bratman’s terms to see J (the thing intended) as that the defendant be enriched. We may cavil that this is unlikely, but in Chillingworth itself the claimant agreed to purchase land subject to contract and paid a deposit. A contract was drawn up but the claimant did not sign and claimed the deposit back—the intention that the defendant retain (be enriched by) the deposit was explicitly lacking. The claim succeeded as the contingent condition of the contract being signed had failed. If so, we can say that

1. a. They each intended that the deposit be paid and the defendant thereby be enriched if X—a signed contract resulted;

2. b. For each party the persistence of the intention that the defendant be enriched depended on continued acceptance that the other also intended the deposit be received (and the defendant thereby enriched) as part of the plan;

3. c. If and only if they both intended that there be a contract, payment of a deposit and the defendant be enriched would they intend as much; and

4. d. The purchaser knew that not-X—there was (and would be) no signed contract—so his intention that the defendant be enriched by the payment of the deposit no longer held. The vendor’s intention could have held on Bratman’s view, but only so long as he continued to believe or accept that the other party still intended the contract and so the enrichment. Since in Chillingworth both parties were aware on the vendor’s tendering of the contract that not-X the collective intention that the defendant be enriched collapsed.

It might be the failure of a promissory condition as to the future. In Benedetti v. Sawiris ^{Footnote 112} Benedetti had made an agreement (the Acquisition Agreement) with Sawiris (and his associated companies) whereby they would jointly acquire Wind, a telecommunications company, from the Italian energy firm, Enel, with the aid of outside investors. The method by which the acquisition was to take place changed considerably. The services that Benedetti actually provided were not those contracted for. As Arden L.J. put it in the Court of Appeal, the agreement simply ran out.^{Footnote 113} If so we can say (simplifying somewhat) that:

1. a. They each intended that Sawiris be enriched by the services (as part of a wider plan);

2. b. For each of them the persistence of the intention that Sawiris be enriched depended on continued belief or acceptance that the other intended that he be enriched (as part of the plan);

3. c. Benedetti intended to perform the services as his part of the plan (part of which was he would be paid) and believed or accepted that that would be a cause of Sawiris’ paying;

4. d. Sawiris did not pay; and

5. e. Benedetti’s acceptance that Sawiris intended to receive the services as part of the wider plan failed, and therefore his intention that Sawiris be enriched failed. Sawiris could hardly complain about this because his failure was the problem. Parenthetically we might note the flipside of this, which is that if you do intend that we J and intend to Y as part of our J-ing, you can say “I am ready and willing to perform”,^{Footnote 114} which negates the cause of action, because it negates the failure of the promissory condition.

It is rare, if not unheard of, for an intention to be wholly unconditional.^{Footnote 115} For Bratman there is rational normative pressure to make our sub-plans mesh. This entails ideally that we collectively plan for the success of the activity, and that we plan for the different contingencies. Klass contends that a foreground necessary condition—that is a condition non-fulfilment of which will make the intention impossible to go through with, but which might yet still not be fulfilled)—reduces the rational pressure to intend means to the end^{Footnote 116} and it reduces the chances of performance and agreeing ways to fill in the gaps.^{Footnote 117} Basically it may still turn out to be a waste of my time to prepare to do something when it is not yet clear that I will be able or allowed to do it. However, given the rational pressure to mesh sub-plans we should try to agree on contingencies. The difficulty comes with background conditions. These can (but need not) be agreed—in which case it is a collective background condition and Chillingworth is at least arguably an example of this. If the background condition is mine alone the question of the effect of a unilateral mistake emerges.

Let us take the swaps cases as an example—although in many cases there would in fact have been a mistake on both sides. The interest rate swaps agreements made by a large number of local authorities in England in the 1980s were deemed void and ultra vires, and Kleinwort Benson v. Lincoln City Council ^{Footnote 118} is authority for the proposition that the losing party may sue in mistake. In an open swap, where payments remained to be made when the swap was ended, failure of consideration was also (uncontroversially) a valid cause of action. We concentrate on the effect of the mistake here:

1. a. I intend that we engage in the swap and that you be enriched by my payments if the swaps agreement is valid at law; you intend that we engage in the swap and that you thereby be enriched simpliciter. We each therefore intend that there be performance of the swap.

2. b. The persistence of your intention that you yourself be enriched by my payments in pursuance of the agreement depends on your acceptance that I intend that you be enriched via the swap and vice versa. It is therefore a condition of your intention that we engage in the swap and that you be enriched thereby that I intend it also.

3. c. If and only if we both intend to engage in the swap, will we carry the swap through.

4. d. Given that my background condition of the validity of the agreement is not met, I do not intend that you be enriched because my payments were on the basis of an intention conditional on facts other than those that turned out to be the case. You can continue to intend that the swap go ahead (i.e. you can continue to intend that we J) until you are aware of my lack of intention, but we no longer intend to engage in the swap and we no longer intend that you be enriched.

I should therefore recover in mistake as there is a background condition (the validity of the obligation) to my payment that was in fact unfulfilled, even though I believed or accepted it was fulfilled. The fact that my interlocking personal intention falls away means that the collective intention falls away too because ultimately your intention that we J is mediated via mine and therefore also fails. When you know that I will not be paying, your expectation of counter-performance fails. There is a failure of consideration, and likewise for me when it becomes clear you will not be performing.

(iii) Implications for the common law

The importance of Bratman’s view lies in three insights. The first, as we have just seen, is that it allows us to see why the two causes of action—mistake and failure of consideration—can be concurrent. Subject to any super-added obligations, which, if of legal effect, will need to be rescinded, mistake and failure of consideration should on this view be able to lie concurrently.^{Footnote 119} Edelman—now Edelman J. of the Supreme Court of Western Australia—argues that in bilateral cases the defendant’s autonomy is also engaged and therefore communication between the parties is required for relief to be available,^{Footnote 120} which entails that the causes of action cannot lie concurrently. This is the wrong way round. The real position is that it is only because of the communication that there is a collective intention and bilaterality. As we have seen mutual knowledge of each other’s beliefs and intentions is vital for a shared collective intention and therefore a shared basis. Some of the cases which Edelman criticises are therefore not, in point of fact, bilateral transactions. He argues for example that Dextra Bank v. Bank of Jamaica ^{Footnote 121} is a bilateral transaction. The belief that Dextra had made as to the BoJ’s taking a loan was causative of the payment, but relief was barred because they had not communicated that belief there was a loan.^{Footnote 122} Elise Bant and Peter Creighton argue that the decision would be defensible as barring a claim in failure of consideration, but relief in mistake should not be so restricted.^{Footnote 123} The BoJ thought there was a currency purchase; Dextra thought it was a loan. It is reminiscent of the case of Raffles v. Wichelhaus ^{Footnote 124} where there was no contract because the parties were, unbeknownst to each other, referring to entirely different ships with the same name. There was no contract or collective intention there, and likewise in Dextra, because they were talking at cross-purposes. In such a case a mistake claim must still be possible.

The second insight therefore that Bratman gives us is that understanding failure of consideration as a failure of collective intention allows us to distinguish between cases which are failure of consideration and cases where we should give relief, if at all, on a different basis. It gives us a means of controlling failure of basis’ ambit and preventing its uncontrolled imperial expansion. An example of this is Felicity Maher’s suggestion^{Footnote 125} that Woolwich Building Society v. IRC, ^{Footnote 126} where the building society was granted repayment of money paid under regulations that turned out to be void, be seen in terms of failure of basis. It should not be. It does not matter whether the Inland Revenue Commissioners knew or agreed with the building society that they (the IRC) take any risk of payment not being due. It was not due and in the context of overpayments of tax, this is all we need.

Related to this is the third insight to which we have already alluded several times. Risk is irrelevant or at least superfluous—despite its enduring popularity in the cases and literature. Risk is intuitive, which probably explains that popularity, but if I take the risk, all that means is that the collective intention has not failed.

Let us make one final point about the ambit of the claim, related to our second insight, as a postscript. Robert Chambers has argued that automatic resulting trusts are based on failure of consideration.^{Footnote 127} In Vandervell v. IRC ^{Footnote 128} for instance Vandervell wanted to endow a chair at the Royal College of Surgeons. He arranged to transfer to them a number of shares, with an option for his trustee company to buy them back. The Inland Revenue claimed that he had not divested himself fully of the beneficial interest in the shares. The RCS had, as part of the agreement with Vandervell, granted an option to buy the shares back to the trustee. An option is itself a proprietary right in the thing subject to it. The trustee company retained a relationship with Vandervell in that he retained the right to decide on what trusts the shares would be held after the option was exercised. He failed to name the trust beneficiaries. There was therefore a resulting trust. The recipient of the money is enriched at the expense of the claimant and the condition for the transfer, a trust, failed. The important point is that the trustee was never intended to hold the asset outright.^{Footnote 129}

The trust might fail for a number of reasons. It might fail because no beneficiary was named as in Vandervell. It might fail because the objects are uncertain—my “favourite undergraduate” say,^{Footnote 130} but in these cases it will not fail because a collective intention to create a trust fails. We can demonstrate this relatively easily. It is trite that a trust will not fail for want of a trustee;^{Footnote 131} the court has a statutory power to appoint a trustee where it is expedient to do so.^{Footnote 132} Consequently if the trust is valid and the trustee has no interest in acting he can be replaced. In initial failure cases a collective intention of the settlor and trustee that there be a trust is neither necessary nor sufficient. Alternatively where the trustee has agreed to act, but the basis—the marriage of the beneficiaries—fails to take place there has been a failure of a collective intention,^{Footnote 133} but this is not a necessary condition. It seems therefore that resulting trusts are not really based on a failure of collective basis but individual basis. This of course does not mean that we ought not to still see these as unjust enrichment cases. Cases like Vandervell would after all be a condictio sine causa specialis in South African law,^{Footnote 134} but the failure of a future condition—the parties’ marriage in Essery v. Coulard ^{Footnote 135}—a condictio causa data causa non secuta, indicating the different underlying rationales of the two cases. The fact that they are both labelled as automatic resulting trusts and the fact that this is now deeply entrenched in the common law does, however, suggest that we must accept failure of basis as having stretched (a little) too far and that it cannot be easily pulled back. Nonetheless, as I have argued before in this context it makes little practical difference, except as to limitation and private international law consequences.^{Footnote 136}