Agnes Callard getting Akrasia wrong

Agnes Callard writes in the Harvard Review-

He goes to the party, though he thinks he should stay home and
study. What explains a weak-willed (or akratic) action like this one?

The explanation of a weak-willed action is that the dude wot done it was weak willed- d'uh. Similarly an angry action is explained by anger. A drunken action by drunkeness. And so on and so forth.

We need a Structural Causal Model of the agent in order to be able to explain his actions in an alethic manner. Lacking that, we should say nothing.

One answer is that he is a pseudo-akratic: he doesn’t really believe he
ought to study.

Another answer is that he is genuinely akratic. He gets off on telling everybody how weak-willed he is till someone pressures him to put out.

Which answer should we pick? Well, that depends. Are we talking about a lack of will power or is there some other problem? Perhaps the guy is dyslexic and knows, deep down, that it won't make any difference if he stays home studying. Here the fix may involve getting him specialist educational help. A different problem would be that the guy actually has a substance abuse problem. He says he is going to a party- because that is socially acceptable. Actually, he just wants to get high.

Suppose the problem really is that he has a weak will. What makes it weak? Perhaps, he was sexually abused as a child and this damaged his sense of self. This means lots of therapy. Alternatively, he may have ADHD and needs medication. If he can't afford either therapy or medication, perhaps he could buy a cheap book on auto hypnotism or Neuro Linguistic Programming or something of that sort. If all else fails, he could turn to Religion. Perhaps the ibbur of a great soul will endow him with a strong will and a good moral character.

There is little point in analyzing a situation where something is defective in a merely analytical way. If you have a heart attack, you don't want learned Doctors to stand around asking- 'what reason could his heart have for attacking him?'- only for their discussion to ruefully conclude with Pascal's bromide- 'The heart has its reasons which reason knows nothing of. '

The truth is our will may be weak because we have no strong reason for it to be otherwise. The same may be true of our muscles and our brain power.

It appears that academic philosophy takes a different view-

 Concluding his critical survey article on akrasia, Arthur Walker (1989) writes With the notable exception of Wiggins, all the [28] philosophers we have discussed accept (i) that whatever reason the akrates has for performing his akratic action is overridden by his better judgment and (ii) that the akrates fully recognizes that his reason is thus overridden. Yet they maintain that the akrates acts on this reason. (p. 670) 

The problem here is that it is being assumed that the Will is univocal or, alternatively, that Judgement is a weak preference ordering of an Arrow Debreu type. That's fine if Knightian Uncertainty did not exist or we lived in a Occassionalist Universe or Evolution is just a silly theory. Sadly that it isn't the case. Judgment can't be 'Utility maximising'. It must be 'Regret Minimizing'. But Regret is closely associated with Will. Utility is not related to Will in any way. A person with no Will power who maximized utility would not be an akrates though he may be a despicable human being.

The notion of Will is associated with the capacity to keep a vow or resist a temptation. In this case, it may be that the young man identified studying with exercising will power and partying with failing to do so. In this case the reason he partied is because his will was weak. We don't know that partying was his best option, all things considered, after studying was ruled out.

Callard dismisses the 'weaker reason' thesis for a different reason

Let’s take a closer look at the case of Will. His weaker reason, that partying is fun, is a pro tanto normative reason to go to the party.

It may be- but equally it may not. We have all gone to parties which weren't fun simply because we didn't want to go home and do our taxes.

As such, it is counted in the agent’s deliberations about what to do, deliberations which lead him to conclude that, all things considered, he should stay home and study. The difference between the agent’s representation of the stronger reason (I ought to stay home and study for the sake of academic success) and the final reason, the one connected to his all-things-considered judgment, (I ought to stay home and study for the sake of academic success even though this means missing the party, which would be fun to attend) reflects the agent’s mental act of counting the weaker reason. It is at this point that the weaker reasons theorist wants to claim that, even after it has been outweighed by the stronger reason, the weaker reason counts as a reason to go to the party. He needs to say this because he takes the akratic to be motivated by the normative force of this reason after it has been outweighed. The weaker reasons theorist is thus saying more than that the party is still fun after Will has decided not to attend it. He is saying that this thought, that the party is fun, still counts as a reason to go to the party.

But why say that? Why not say the reason Will went to the party was because he was weak willed? This is also the reason he didn't stay at home and study.

Which is to say that the reason has rational force over and above its role in the deliberations concluding in the all-things-considered judgment.

The all things considered judgment was 'I should assert my willpower and stay home and study'.  The outcome was that he went to a party because his willpower was too weak to conquer temptation. In other words, he lacked a quality needed to perform the judgment. I may say 'I should jump 60 feet into the air in the same manner that Superheroes do.' However, I can only jump one foot into the air. Thus, because I am not a Superhero, I can't comply with a normative consideration which does not apply to me because I can't abide by norms of a superior type of being.

I claim that this is not so: once the weaker reason has been counted in the deliberations leading to the all-things-considered judgment, it cannot (in a sense to be explored below) be re-counted.

This is because the judgement was what I have described not what Collard says it is.  We may judge an action to be wrong and do it anyway because we are wrongdoers. We don't say that we changed our judgment as to what is wrong when we did that wrong thing. We may say we didn't know we were wrongdoers or that we would become wrongdoers. Yet that is the outcome. The agent changed. The judgment didn't.

To see why, let us consider the following financial analogue: you deposit $5 into your bank account. All else being equal, this means that you are now in a position to withdraw $5: the $5 that you deposited counts as $5 the bank owes you. But assume other things are not equal: before making the deposit, you had owed the bank $3. Now it no longer remains true that a deposit of $5 allows for a withdrawal of $5. When you deposit the $5, your balance reads not $5 but $2. The fact that you deposited $5 is not always a basis upon which to demand that the bank give you $5.

So what? The fact is, people do demand something similar all the time. Indeed they may demand eight dollars given that the Bank already allowed them to overdraw by three dollars.

The weaker reasons theorist is, in effect, acknowledging that the balance is $2 (the all-things-considered judgment), but claiming you still have, somewhere, $5 in your account (the weaker reason) to draw upon.

Because the Bank seems to have allowed you a three dollar overdraft limit. This is quite normal practice.

The insistence that the bank still owes you something for that $5 deposit even after it has been calculated into the $2 balance exhibits the fallacy that undermines the weaker reasons thesis: I will call it the fallacy of double-counting. In applying this to the case of explaining action, it is helpful to distinguish between a net and a component factor. If C and N stand to one another in the relation of component and net (i.e., C is a component of N), then C can only figure into an explanation insofar as N has not already been deployed. The component cannot play any role independent of the net into which it figures, otherwise it would be invoked twice in the same explanation.

Explanations are not enumerations. Double-counting is only a problem for the latter. In any case, there is only one reason for everything here- Will was weak-willed. That's why he partied instead of studying. That's also why he may have got drunk or hit on his best friend's girl.

A good reason is one which features multiple times- e.g. 'I didn't commit the crime because I was not there. Because I wasn't there I could not have inserted a knife into my enemy's chest. I could not dispose of the murder weapon because I wasn't there to use it in the first place'.

The weaker reasons theorist asks for the weaker reason to first be counted as a component of the all-things-considered judgment, and then as an independent consideration motivating the agent to act against his all-things-considered judgment.

So what? That judgment was not implemented because of weakness of the will. The actual outcome too arose for the same reason. Suppose the party really was fun. Then we may say 'weakness of the will caused a fun activity to appear enticing. That's why Will went to the party.'

A Court of Law may not be able to implement a harsher judgement because of some obstacle- e.g. the criminal can only be extradited if the death penalty is off the table- this does not mean they have to set aside the more lenient judgment. If we can't get the guy on Murder, get him on manslaughter.

Suppose a rational investor should buy Dollars not Sterling. But the investor gets drunk and irrationally buys Sterling. Does the investor have to hand back the Sterling and keep nothing? We don't say, 'X had no reason to buy Sterling.' We say he had a better reason to buy Dollars but was too drunk to act accordingly.

The all-things-considered judgment is the agent’s net normative reason: what he has reason to do, given all his component (pro tanto) reasons.

A normative reason may cease to apply to an agent who loses a particular property- e.g. being alive- and thus no longer subscribes or is subject to that norm. In this scenario the loss is of will-power. This means his component reasons have changed.

Suppose I stop going to Church and attend a Mosque instead. It is likely that this is because I have ceased to be a Christian and have become a Muslim instead. I may say 'my reason for going to Church was to get closer to God'. I may even say that I admit that going to Church does bring one closer to God and this is desirable. I further stipulate that I don't feel closer to God in the mosque. I go there for a different reason- viz. to comply with the norms of my new Faith. It may be there is evidence that I changed religion because I wished to escape persecution or to gain a worldly reward. I may even say that if my Will had been stronger I would have resisted the temptation to convert. This does not change the fact that the reason I go to the mosque, not the Church, is because I am now a Muslim, not a Christian.

Unless revised, rejected, superseded or forgotten—as, ex hypothesi, the akratic’s isn’t—an all-things-considered judgment is, to continue the analogy from above, the balance in the agent’s rational bank account.

But the agent has changed. If he had will-power and self confidence he would demand eight dollars, insisting that 3 dollars was his overdraft limit. But he lacks will power. His attitude is supine. He cringes in front of the cashier and accepts two dollars.

It is foolish to say 'ex hypothesi' nothing in the agent has changed. It would be like saying a person who converts to another religion, hasn't actually done so- ex hypothesi.

Which means that it reflects all his (normative) reasons, in the way that a bank balance reflects all one’s money.

The point about bank balances is that they represent Credit. Banking is about trust. What matters is whether the person has become more or less bankable. How much you can withdraw depends on that. I may have a large sum of money in the bank. Yet my business is failing. I go to withdraw what I have. The Bank Manager warns me that I may be committing an illegal action. I could be prosecuted as a willful defaulter. Indeed the Bank may be within its rights to forbid the transaction depending on the jurisdiction.

The point can be transferred from finance to physics: if multiple forces act on a body, the net force created by summing up all forces in question does all the work; the components don’t have independent force.

But if the mass of the object being acted upon changes magically because of a component, then the outcome is quite different. That is what is happening here. Will's will weakened.

The all-things-considered judgment is the net (normative) rational force, and once you’ve cited a net factor, its component cannot play any further rational role.

Yes it can if mass magically changes. Conatus is like inertia. A change in will-power changes conatus.

If the weaker reason moved the agent, it would have to be in virtue of the fact that it had already been reflected in the all-things-considered reason,

or else in the agent himself

just as a force on an object can move it only via the net force.

not if a portion of the mass of the object is turned into energy by a 'component' force.

It is natural to claim that the akratic partier, even after he has decided against it, still has a (normative) reason to go to the party. I am not denying this claim, but rather the weaker reasons theorist’s inference from it to the conclusion that the akratic acts on this reason: “In one sense, it is easy to say why he acted as he did, since he had reasons for his action” (Davidson, quoted on p. 69 above).

There is no great scandal here. We understand that temptation saps the Will.

Consider, again, the financial analogy: there is a sense in which the agent with the $2 balance still has the $5 that he deposited, since that $5 is working against the debt of $3 to produce the $2 in his balance.

It is also 'working' to provide an argument for a higher overdraft limit.

But this is not a sense which supports an attempt to withdraw $5 from the account.

Why not? A forceful personality and a confident manner might enable him to withdraw twice that.

It isn’t as though the $5 deposit disappeared—on the contrary, there is an entry of $5 on the credit column of the account which goes into the figuring of total assets. The $5 that you deposited counts, and is counted, in figuring your balance; but it does not, in this case, count towards your being owed $5 by the bank.

What you are owed by the bank, or what you are owed it, is by no means a cut and dried affair. You may claim to have had an understanding with the Bank. They may have acted in a manner such that a reasonable person would feel they had, for consideration, agreed to an elastic arrangement. This is an idiographic matter for lawyers and accountants.

Similarly, it is possible to make sense of the claim that the akratic still had a reason to go to the party, even if I am right that he cannot act on this reason.

If Collard is right, then the concept of opportunity cost is irrelevant to decision making. Either you do what's best, all things considered, or you have no reason to do anything. You must put all your eggs in the superior basket. Never put any eggs in the inferior basket. Furthermore, you must suffer no change in your conatus. Be your best self now and continue to be that no matter how the dynamics plays out. If currently you paint well- for an eight year old- you must continue to paint like an eight year old till your are Eighty years old.

Here are some ways to interpret the claim that “fun is still a reason to party” so that the later comes out true: 

(1) Even after having been decided against, partying is still fun.

Is it though? Will you really enjoy the party knowing you are a weak-willed creature?

(2) It would be rational for this agent to party if he did not have to study.

We don't know that. The opportunity cost may consist of working out instead of studying. After all, working out takes more will-power than mindlessly getting drunk.

(3) The fact that partying is fun is a consideration that it is rational for people in a position to go to parties to weigh and consider; the agent was not wrong to weigh it in his initial deliberations. That is to say, its rational force has not been undermined: it is a pro tanto rather than a prima facie reason. 

How do we know? This is not an 'interpretation', it is an unfounded assertion. Further, it is clearly false. It is not a fact that partying is fun. It may be or it may not be. It has no rational force.

(4) The all-things-considered judgment endorses studying more weakly than the stronger reason, since if he didn’t have a party to go to he would have even more reason to study.

No. Opportunity cost can rise without Choice being affected. All that matters is that the next best alternative is less- no matter how slightly.

But none of (l)–(4) supports the claim that the fun-ness of partying can, under akratic circumstances, be a motivating normative reason. For that, the weaker reasons theorist needs the following claim: the fact that partying is fun can be counted by the akratic agent as a reason to go to the party, even after it has been outweighed. But this is not true, since it has already been counted.

(1)-(4) were false or foolish statements. Under akrastic circumstances, an agent changes such that a previously binding norm ceases to apply whereas a new norm- which may previously have been anathema- becomes binding. The strong willed stay home and study just as the strong willed Christian goes to Church. Being discovered to be weak willed, or choosing to be so, entails not studying but partying or changing religion for a worldly reason.

I think Collard has written nonsense because of a fundamental confusion as to what reasons are. I think they correspond to predictions of the correct Structural Causal Model. Thus if we know a person is weak willed we predict that they take the easier path- this may be partying instead of studying or it may be converting to another reason for worldly reasons. Here Will power is a paramater in a model which predicts the agent's behavior. This is standard Economic thinking

Collard, however, is an academic philosopher. She writes-

'To say that an intentional action is done for a reason is just to say that an explanatory reason can also be a normative reason.

I don't think this is true. Saying something silly may or may not be an intentional act. If there were reasons for all intentional acts, intentionality would be irrelevant. We say 'I'm sure x had a good reason for his actions' if we don't feel it worthwhile to inquire into the licitness of his actions on the grounds of our beliefs about his character, competence and motivations. On the other hand, if we are suspicious of him, we may well interrogate him or otherwise seek to understand the reasons for his actions. It may be that we fear he acted capriciously or in a nihilistic manner. In a legal context, this may prompt us to issue a 'show cause' notice. Why? It is because an 'explanatory reason' may be antinomian, antaganomic or radically perverse. They may be the opposite of whatever is meant by normative reasons.

The idea of acting on a reason is the idea of an explanatory reason which explains in virtue of the fact that it is a normative reason.

This is a stupid idea. Reasons are either linked to Structural Causal Models of a useful type or else they are 'anything goes'. An explanatory reason either points to a SCM and the behavior in questions is regulated by a parameter in the Model- e.g. weakness of Will, or 'Christian' or 'Muslim'- which has a normative aspect, or else the explanatory reason 'explains' in virtue of some protocol established by the one who receives it. Such protocols may vary widely depending upon the purpose they serve. Suppose I seek to enter the country. The Immigration officer asks me the purpose of my visit. I reply 'because I'm a citizen with a right of domicile'. That explanation is sufficient though it doesn't really explain anything. The Visa Officer may request me to attend a police interview where I am questioned as to what I was doing in Northern Pakistan. I may say 'I have dual Nationality. I'm Pakistani same as I am British.' But this is not an adequate explanation if the Police suspect that I waged war against British forces while staying in Pakistan and that this was the true purpose of my visit there. Though my right to enter Britain would not be endangered, I may be detained by the British Justice System unless I can provide a good explanation of why I was in a particular locale. At some point I may appeal to a SCM. I may say 'look, I'm a Christian Bishop. It is absurd to think I was helping the terrorists. They massacre Christians.'

My point is that reasons matter if something can be done about them.Will power can be enhanced or reduced. Religions can be changed. Citizenship and other Legal rights can be modified or revoked. So long as there is a SCM it is worthwhile inquiring after reasons. But such reasons can't be the subject of philosophy. They are purely instrumental and of a provisional or heuristic nature.

Consider the following-

To sum up: Will’s weaker reason appears twice in his deliberations: in the original weighing, and right before he acts.

How do we know? There may have been no deliberations or there may have been thousands. 'Deliberation' is not well defined. What stops him changing his mind one second after he acts? What about two seconds? What about five minutes later when he realizes it is cold outside?

The first time, it’s a normative reason but not an explanatory reason (because there is, as yet, no action to explain).

It can't be a normative reason because he didn't act on it. An explanation doesn't have to occur after the act. It can be proffered before hand. This is what happens when you have a medical procedure. They explain the thing before they do it. After it is over, you couldn't care less.

The second time, it’s an explanatory reason but not a normative one (it can no longer be counted in favor of the action).

It is a normative one. The guy counts himself as a weak willed party animal, not a studious striver and acts according to the norm governing people of his type. It is also the correct explanation of his behavior- he has a weak character and likes getting high with people like himself. 

David Wiggins attacks accounts of rational choice that take all deliberation to consist in the maximizing, or ranking, of goods in terms of one property.

So, a defective theory of Rational Choice is being attacked. Surely, the best way to do so is by looking at a more Rational theory? Regret minimization is one such. The future being uncertain, we hedge our bets.

He says that on such a theory, which he labels “commensurabilist,” “it will be harder than it ought to be to understand weakness of the will as having reasons of its own” . He is thus proposing that the concept of incommensurability can be of service in saving the weaker reasons thesis.

This is foolish. A wrong theory- e.g Aliens want to fuck us over- should not be replaced by its opposite- Aliens want to help us- to explain why I can't find the remote. Aliens have nothing to do with the fact that that my remote slipped under the sofa.

Wiggins thinks that incommensurability is manifested in the failure of our better judgment to compensate us for the loss of the goods featured in our “weaker” reason; akrasia, he asserts, is the occasional result of this incommensurability.

This is crazy. Nothing can compensate us for our opportunity cost. We can't have our cake and eat it too.
If akrasia is a real thing- as 'Vishaad' is a real thing in the Mahabharata- then a person with it is governed by different norms. The Bhagvad Gita is about dispelling Vishaad- removing akrasia- so that the agent recovers his previous status and the old norms apply.

Because success is no compensation for the loss of fun (they cannot, he insists, both be represented as quantities of some good, e.g. pleasure), fun remains a motivating consideration even after we have decided it’s best to  succeed.

The reverse is the case. To the strong-willed, partying isn't fun- it is beastliness.

We can see this as an attempt to respond to the double-counting problem: if success cannot compensate one for the loss of fun, then maybe fun was not fully ‘counted’ in the all-things-considered judgment –and if fun was not counted the first time around, it can rationally motivate after the all-things-considered judgment is formed.

The only double counting here is the notion that we have to be compensated for the loss of cake when we eat it. Choice under scarcity doesn't work that way. Nor does Reason.

There are several problems with this approach, but the one that interests me is this: ex hypothesi, the akratic agent has taken the fact that partying is fun into account when forming his all-things-considered judgment.

No. The non akratic agent has done that. To be akrastic, the agent must say something like 'If I have a strong will, this is good, that is wrong. But, I have a weak will. So I do wrong'.  Similarly a sick person may say 'If I were healthy, I'd go to work. Because I'm sick, I can't get out of bed.'

Someone with a knowledge of the right SCM could help them do what they say they want to do.

He compared the goods in question (say, success and fun), and decided that he ought (at least, for now) to pursue one and not the other.

If that was his decision, where is the weakness of the will?

It may be true that success and fun are incommensurable, in the sense that the goodness of fun is not reducible to some other kind of goodness (say, pleasure) to which success is also reducible; thus it may be that the skipping the party entails an uncompensated-for loss of fun, and the akratic agent knew this would be the case.

So what? Apples and oranges may not be 'commensurable', but we choose between them all the time. They become comparable by reason of their being on our choice menu.

But how is this relevant to explaining akrasia?

It isn't.

Presumably these are all things he took into consideration, before deciding that, all things considered, he ought to stay home and study. Here’s the problem: the issue is not commensurability but comparability. Fun may not be commensurable with success, but it is certainly comparable, as is illustrated by the akratic agent, who compares them. The diference between commensurability and comparability is best illustrated by the arena in which these concepts are at home, geometry. The hypotenuse of a right triangle is incommensurable with the side, meaning that one cannot find a unit of measurement small enough that it will ft without remainder into both line segments.

The hypotenuse is simply longer than either of the sides. Both commensurability and comparability obtain because the method of measurement is the same.

But the two line segments are, of course, comparable: the hypotenuse is longer. The corresponding ethical distinction is between whether or not one good is greater than another (comparability) and whether or not the agent who chooses the greater over the smaller is compensated in kind for the loss of the smaller (commensurability).

Rubbish. Whatever you chose is your compensation for what you didn't chose.
Wikipedia says- 'In ethics, two values (or norms, reasons, or goods) are incommensurable (or incommensurate, or incomparable) when they do not share a common standard of measurement or cannot be compared to each other in a certain way.' Comparability is commensurability pro tanto. No agent is ever compensated for his opportunity cost. If you eat the cake, you can't also keep it for the next day. Why is Collard 'double counting' opportunity cost? Just look at the crazy shit this forces her to write!-

The question on which the double-counting problem focuses our attention is whether the akratic has taken the reason in favor of partying (fun) into account when deciding to study, and this is the question of comparison, not measurement.

Cardinal Utility is comparison turned into measurement. We know that for any Revealed Preference there must be a Cardinal Utility measure.

The fact that A is incommensurable with B is not grounds for claiming that our reasons for pursuing A were not taken in to account in our all-things-considered judgment to pursue B. This suggests a way of modifying Wiggins’ position in order to save it: perhaps we could say that fun and success are incomparable goods.

No we can't as you will soon admit.

Two goods are incomparable if a determination of relative value cannot be made between them. But in the standard case of akrasia it is part of the setup of the scenario that there is comparability of fun and success—they are not only comparable, but in fact compared. Some agents are stymied, unable to compare their available options (“Which is better? Perhaps I’ll never know!”), but the akratic is not one of them.

To qualify as akrastic it must be the case that there is a true SCM such that Will is a parameter and it governs which norms he will subscribe to.

How can incomparability be useful to explaining akrasia if, ex hypothesi, the akratic has compared fun and success?

It can't. Get over it.

Perhaps we need to look outside the fun/success dichotomy to explain what motivates Will. Let us suppose there is some good or value other than fun that we might take his akratic action to be aimed at. The job of this value, call it X, is to be incomparable with success, so as to be able to serve as the basis for the Will’s reason to go to the party without running afoul of the double-counting problem. But now, let us ask of X: is it relevant to the action of going to the party? It must be, or it cannot be the basis of a (normative) reason to go to the party. In this case, Will must have weighed X, when he was coming up with his all-things-considered judgment, and decided that he had less reason to pursue X than success. How did he decide this, if success and X are incomparable? Perhaps he decided to flip a coin to decide between them, and then had a procedural reason, of the sort explored by Bratman (1987, ch. 5), to follow through on the plan and select whichever of the goods “wins” the toss.

If it wins, then a stochastic comparison mechanism has decided the issue. So there was comparison.

Perhaps he decided that, given some facts about the circumstances he is in, it was not the time and place to worry about X (this is not the same thing as saying success has more value than X). However he made the decision, he must have done it somehow if he formed an all-things-considered judgment that he had most reason to go to the party. But in this case the reason which advocates partying in the name of the value X has been counted in the deliberations that produced the conclusion that he had most reason, under these circumstances, to pursue a different value, success, by staying home and studying. If we try to describe Will as acting on the reason to party for the sake of X, we will be double-counting once more.

No we won't. We will simply be talking nonsense.

The point is this: incomparable values can give rise to comparable reasons.

Only if one believes a Value can give rise to a state of the World. But, in that case, why not demand an even better Value so as to live in a nicer World?

There seems to be one move remaining for the incomparabilistic explainer of akrasia: claim that the reason motivating the akratic action was subjectively incomparable with the reason the akratic thinks he ought to act on. But at this point we hit a wall: for even if we could make sense of incomparable reasons (and I doubt we can) seeing oneself as torn between such reasons would preclude the formation of an all-things-considered judgment.

Not necessarily. If cognition is costly or information is asymmetric, and Knightian Uncertainty obtains, there is a regret minimizing procedure which corresponds to the all-things-considered Judgment. Why? Because the costs and constraints of judging have been internalized.

An all-things-considered judgment just is the judgment that, given all the considerations I take to be relevant, this is what I ought to do. The agent faced with incomparable reasons relevant to her decision would have to conclude that it was impossible to take all relevant considerations into account and give up on forming an all-things-considered judgment.  

Her own cognitive costs and constraints are part of an all-things-considered judgment. In this case, it may be regret minimizing to suspend judgment but, then again, it may not.

I conclude that whatever it is we are doing when we act akratically, it is not acting on an acknowledgedly weaker reason.

When we say 'I lack the will to do what is best' we may be acting on a 'weaker reason'. Why? Because there may be something which allows you to do something even better and which does not require any will-power whatsoever. Suppose you want to give up ciggies and lack the will power to do so. You then use a nicotine spray. This causes you not just to quit smoking but also cut down on your drinking. Useful disciplines identify alethic Structural Causal Models and find effective solutions. Analytical philosophy is wholly worthless. It leads to people saying stupider things than they would otherwise do.