Tuesday, April 14, 2009

Richard Carrier's Satisfaction Theory of Value

A member of the studio audience invited philosopher Richard Carrier to look over some of my work and to provide comment.

I thought I owed my readers a response to what a professional philosopher might say with respect to the ideas that I present here.

The first thing to note is that there are two types of responses to any theory.

Now, this is going to take a bit of effort, I'm afraid. However, the effort is worth it in establishing a more solid foundation for a morality without gods.

One type of response is to propose an alternative theory. An alternative theory suggests that there are some differences in what would be predicted by the theory being proposed and the theory being criticized. Specifically, it offers some alternative set of predictions. In many cases, we can then go out and see which theory makes the most accurate predictions. That will tell us which theory to adopt.

Another type of response proposes what amounts to a different version of the same theory. In this case, the two theories produce exactly the same results. There is no experiment that can be performed or observation to be made that would show a difference between them. Yet, we can still judge the merits of each theory by their complexity, and we have reason to go with whichever theory is the simpler.

An example of the latter distinction is the distinction between Ptolemy’s earth-centered theory of the solar system versus the Copernican sun-centered theory. Actually, the two theories do not give different predictions as the observed motions of the planets. The difference is that the former required epicycles upon epicycles to get the same results. It was judged best, in this case, to go with the easiest theory.

I do not see anywhere in Carrier’s response where he identifies a place where the two theories might deviate, allowing us to adopt the better of the two predictors. We have only the option of looking over the two theories to determine which is the simpler.

The focus of Carrier’s response concerns a case where I hold that happiness theory and desire fulfillment theory produce different results:

Ask individuals what they will do under the following circumstances:

Option 1: "You will be made to believe that your child is living a healthy and successful life while, in fact, your child is being tortured mercilessly."

Option 2: "You will be made to believe that your child is being tortured mercilessly while, in fact, your child is living a healthy and successful life."

For the parent with a desire that P, where P = "my child is living a healthy and successful life", Option 2 is the option that fulfills that desire. It creates a state of affairs S in which P is true. Whereas, if happiness were the goal, then Option 1 would be the better choice because Option 1 will produce the most happiness for the agent. Many people asked this question would choose Option 2 – the desire fulfillment option. Thus, we have a result that demonstrates that desire fulfillment is the better of the two theories.

Against this, Carrier wishes to deny the assertion that happiness theory requires that the agent choose Option 1 – thus denying that the two theories yield different predictions. This means that the two theories are thrown into a competition of which of the two is simpler. Which of the two theories actually answers questions, and which of the two theories seems to add epicycles upon epicycles in order to continue to get results that accurately match real-world observations?

Carrier responds:

If it gave the decider no happiness to "make or keep the proposition true," then by definition they wouldn't have the desire to "make or keep the proposition true" (indeed I suspect it is physiological impossible to possess a desire the fulfillment of which produces no satisfaction).

My first question is: Why would he suspect that? There is no argument here as to why satisfaction is necessary – merely an assertion that it is necessary. Carrier, for all practical purposes, is simply declaring that he has trouble imagining a person seeking the fulfillment of a desire that brings no satisfaction.

Yet, the answer to this is merely to assert that the fault does not lie with the theory, it lies instead with Carrier's imagination. What Carrier needs to provide is an account of what satisfaction is, and why satisfaction and only satisfaction – nothing else has the power to motivate action.

As for the possibility of motivating action through systems other than satisfaction, we are surrounded by examples in which this is possible. Many computer programs use a method where the computer is programmed with the ability to evaluate particular end states, and then evaluate options according to which is most likely to bring about the most highly valued end states. In creating these programs, we do not tend to believe that we are programming computers with a sense of satisfaction over various results. It is simply programmed to give each end state a value.

What is it about happiness and only happiness that gives it the power to motivate actions?

Consider the claim being made here.

There is a state of affairs called "satisfaction" (S) and we are motivated to pursue S. However, for all other states of affairs T that are not equal to S, not only is it the case that we have no motivation to pursue T, but we cannot have any motivation to pursue T, according to Carrier.

Why not?

This is actually a remarkable claim. It is possible for the mother in the above case to be motivated to bring about some state of personal satisfaction, but it is not possible for the mother to be motivated by the state of her child being healthy and successful except insofar as the child's wellbeing is a means (a tool, an instrument) for bringing about a state of satisfaction.

I think we need an account of what satisfaction is and how it is that such a thing can motivate us to action, when nothing else in the universe can even possibly motivate us to action.

Without such an account, I would suggest that the desire fulfillment theory leaves fewer mysteries then the satisfaction theories. If it is true that both theories yield the same predictions, the fact that the desire fulfillment theory does not leave these types of questions unanswered gives it an advantage.

To be continued . . .

13 comments:

Martin Freedman said...

Hi I follow Richard's blog too but have seen no posts on this. Does he have a post on this and if not could he post on this or could you include his full reply here, if he does not mind?

BTW I also pointed Tom Clark of naturalism.org to your site and he approves of your stance with respect to free will. He has not commented (to me anyway) with respect to DU, which I think is better than his ethically natural moral model.

Larry Hamelin said...
This comment has been removed by the author.
Larry Hamelin said...

I thought I owed my readers a response to what a professional philosopher might say with respect to the ideas that I present here.Richard Carrier is a professional historian, not a professional philosopher. At least his PhD is in history.

Martin Freedman said...

Carrier has written books and papers on naturalism and defending metaphysical naturalism. Is he a professional philosopher, well he seems to divide himself between history and philosophy and is working as a post grad academic (his Jesus book is privately funded though). Does it matter, surely it is the strength of his arguments which is at issue?

Richard Carrier, Ph.D. said...

I assume from the first comment my actual remarks weren't actually published anywhere. Fyfe is referring to a forwarded email I wrote privately to another colleague, though with full permission to forward to Fyfe, and I certainly expected he might comment on it here.

It just would have been more helpful to start with quoting or paraphrasing the relevant section of the email more fully, so people would know what this is all about, and what in fact our differences even are (especially as they are so few and so technically minute). Granted, what I wrote in that email was quite rough. I would have spent time to craft a more intelligible version if I were to have blogged it, I just didn't have the time. Instead it was just typed off the cuff in a quick email. But that only pertains to its readability or intelligibility. As to its content, I stand by it. But it is too long to fit in a comment box here. I'll cut to the chase more directly in a following comment.

As to the other comments, they are correct: I am professionally trained as a historian (albeit a historian of intellectual history--ancient science, philosophy, and religion) but am self-trained and published as a philosopher (in fact I have made more money in philosophy, which ironically by some accounts would sooner make me a professional philosopher). I believe everyone who has the means should be a competently trained philosopher (as I argue in my book, Sense and Goodness without God: A Defense of Metaphysical Naturalism). And I take the mission seriously.

Now to the point...

Richard Carrier said...

I want to start by eliminating a mistaken analogy:

"An example of the latter distinction is the distinction between Ptolemy’s earth-centered theory of the solar system versus the Copernican sun-centered theory. Actually, the two theories do not give different predictions as the observed motions of the planets. The difference is that the former required epicycles upon epicycles to get the same results. It was judged best, in this case, to go with the easiest theory."Actually, this is wrong in two respects:

First, the Ptolemaic and Copernican systems were not observationally equivalent. Ptolemy's system was more predictively accurate than Copernicus's system, so in fact Copernicus should have been rejected on observations alone, and correctly was, until Kepler restored the elements of the Ptolemaic system that Copernicus had dropped (non-circular orbits and inconstant velocities) and made a slight improvement (changing Ptolemy's law of equal angles in equal times to the now-familiar equal areas in equal times), which resulted in a system more predictively accurate than both Ptolemy and Copernicus, but no longer as simple as the Copernican system. And notably, no longer explicable in the way theirs had been (both Ptolemy and Copernicus could explain their respective systems with proposed symmetry laws), begging for an explanation (which was provisionally provided by Galileo but not proved until Newton).

Second, even if the two systems had made the same positional predictions for the planets, they still gave different predictions overall, and some of those differences were in fact explicitly known to both Ptolemy and Copernicus: heliocentrism predicts stellar parallax (which has since been observed and thus empirically confirmed, though it was too small to be detected by the instruments available to Ptolemy and Copernicus, or even in fact Galileo and Newton) and inertial physics (the observational point drummed by Galileo ad nauseum, and extensively confirmed by him observationally), once combined with a theory of universal gravitation to replace the theory of stellar symmetry (a replacement confirmed by Newton to be correct, but already understood as an observable prediction of heliocentric theory before Ptolemy, as revealed in the writings of Plutarch, and taken too far by Galileo in his erroneous tidal theory).

The reason I digress on all this is that the analogy fails because it is wrong: you assume it is possible to tell the difference between two theories (geocentrism and heliocentrism) that make identical predictions based on parsimony, but your only example refutes that very assumption, twice over.

First, we only ever knew heliocentrism was correct by making differential observations, not by judging from parsimony. Thus your analogy sneaks in the very premise it is supposed to deny: the decision between them on parsimony can only be considered valid because the final decision between them was actually made empirically, which really only supports the empirically-based decision, not the parsimony-based one. We only know Copernicus was "right" because of observations: different predictions.

Second, Copernicus wasn't in fact "right." He actually abandoned every correct scientific advance Ptolemy had made, and produced a system that was worse at making predictions. The truth wasn't as simple as Copernicus thought, in fact he abandoned non-circular orbits and inconstant velocities because he found the very notion of them messy and repugnant, when in fact those were actually correct, which is why Kepler only succeeded in beating Ptolemy when he reintroduced them (and improved on them).

In other words, parsimony actually failed in this case: the simpler system (Copernicus) was false, and the truth was a more complex hybrid of Copernicus and Ptolemy (Kepler), which was simpler than Ptolemy's model but more complex than the Copernican model...

Richard Carrier said...

...and now the correct solar model is arguably far more complex than even Ptolemy could have imagined, combining planetary mechanics, thermodynamics, relativity, and complexity theory, with the mutual gravitational interaction of several million orbiting bodies, the rotational effects of the sun on its inertial field, the pressure effects of its solar flares, the ripple effects of its contractions, and now possibly the involvement of dark matter and dark energy as well.

Scientists today can only wish it had all just been epicycles!

Similarly, chemistry didn't turn out simple either: instead of a parsimonious four elements, there turned out to be over ninety of them, which combine to produce billions of different compounds and are in turn the product of dozens of different subatomic particles.

Moral of the story: parsimony is not so reliable a principle.

It has its utility, when carefully defined and applied, but in the real world its application is limited and problematic. We need to go back to observations. Those are the only things that really matter: what can we show is actually the case.

I want that to be the point we keep in mind from now on as we move to the question of moral theory...

Richard Carrier said...

Okay, moral theory...

Fyfe, your main point here isn't clear to me. You are either (a) saying our theories are observationally identical (i.e. they make no different predictions of the real world whatsoever), in which case they are semantically identical as well, and thus in fact the exact same theory, just stated differently. Or (b) you are saying they only entail different predictions that we just haven't yet the technological means to observe, in which case we should be discussing what those observational differences are and how we might gain technological access to them.

It doesn't sound like you are arguing (b), but something closer to (a), but without the recognition that if (a), then one theory cannot be simpler than the other, they can only be identical (i.e. semantically equivalent). If so, then you seem to think that it is possible for our two theories of morality (let's call them F and C) to entail real differences in the world that can never even in principle be observed, and thus concluding the only way left to decide between them is with some criterion of parsimony.

If that is the case, I deny the "can never even in principle be observed" element of this line of reasoning. If our theories differ at all (other than semantically), they must differ in respect to real world facts, and those facts must be in principle observable (even if we don't yet know how to observe them). My book Sense and Goodness without God: A Defense of Metaphysical Naturalism explains epistemologically why I would say this, so I won't defend the point here, I just want to get clear what our disagreement even is, before going any further.

Exactly how do our theories differ, in respect to physical facts? (e.g. human biology, psychology, social dynamics, whatever)

I would argue they differ in the very ways I suggested we could experimentally test: in the way human brains actually work, to produce moral decisions (and thus how they would answer on certain tests, and how the facts would bear out on relevant brain scans and life studies).

To that next...

Richard Carrier said...

Here is the predictive difference I was alluding to in my original email...

In actual fact, humans operate on a system of dispositions (compassion, integrity, self-image, lust, etc.), which have causal consequences on the whole gamut of their decision making, which in turn has an aggregate effect on the conditions of their life (what sorts of friends they have, whether the cops are hunting them down, how they feel about themselves, whether all their actual needs are hierarchically satisfied, their mental and physical health, etc.), which in turn affects their baseline of happiness. I elaborate considerably and more systematically in my book.

All that is scientific fact, not philosophical speculation. Though there is a 'sect' in psychology that tries to deny the existence of character, their experimental apparatus is flawed--it arbitrarily excludes subjects of strongly defined character and instead only examines the behavior of people of conformist character and thus always finds what its looking for by discarding everything else. That's a debate for another day. For now, I'll simply say there are ample facts attesting to the fact that dispositions physically exist and have causal effects on decision making and that those effects are differential in their effects on happiness and well-being (see my book for the relevant bibliographies).

The question then becomes: do we want to know (a) what people will do in this or that situation (imaginary or real), or (b) what they would do if they were fully informed and thought everything through? Question (a) is useful to answer, but all it will get us is descriptive ethics, not prescriptive ethics. To get the latter, we need to answer question (b). Again, my book gives a complete demonstration of why this is the case.

Thus it is not useful for prescriptive ethics to ponder how people will behave when not adequately informed. We must ponder (which means: predict) how people will behave when they are adequately informed.

Thus, I said our theories differ (if they differ at all) in respect to their predictions of how people would behave when fully informed. Though you focus on the single data point of a single decision they would make in a binary condition, that isn't where our theories make different predictions. I said the differential prediction would not occur there, but in questioning them as to why they chose as they did, and the prediction of what they would answer (assuming we can confirm their answer is honest) would only be relevant if the conditions were correct: i.e. as I stated, they have to be fully informed and fully thinking through their decision, so that the decision is fully rational and informed.

In those conditions, I predict my theory would be observationally confirmed, and yours would not.

But I must emphasize: the actual differences in our theories are tremendously small and not IMO of very great consequence.

Richard Carrier said...

Now to some particulars...

"My first question is: Why would he suspect that? There is no argument here as to why satisfaction is necessary – merely an assertion that it is necessary."

Actually, as you even quote me as saying, I predict it can be physically confirmed as a biological fact. I am not assuming anything--I am inferring from what we know of the neurophysics of human decision-making, and what we have learned from the psychological study of why people behave as they do, and how they would change that behavior if they could, and what is actually physically necessary, in terms of changes in the brain, for their behavior to change.

But more importantly, I predict that when people are placed in such situations (in a large sample where answers can be compared), and are confirmed to be fully informed and deciding rationally (and answering queries honestly), they will answer that they made their choice because it made them happier to do so, and that for them the choice was between what sort of person the two choices would make them, and that it satisfied them more to select the choice that made them the sort of person they wanted more to be.

I predict that if the conditions of the experiment are ever correctly fulfilled as described, that will be the observed result.

Indeed, I go one further: coordinated brain scans will directly confirm that these people experience more pleasure at the choice they made than the one they wouldn't have made--if, for example, they are asked to make the other choice contrary to their desire.

That would probably be impossible in reality, but in a test scenario they will be able to do this, yet it will likely still have an observable effect as predicted, since we have seen the same effect in other related tests, e.g. people swallowing aspirin from bottles they were just asked to label as cyanide: despite fully knowing the label is bogus they still have a harder time taking the pill and show more anxiety at doing so, than if they weren't instructed to mislabel the bottle. I suspect forcing themselves to choose Option 1 when they want to choose Option 2, even though the scenario is entirely fictional, will generate a similar level of unease that will be directly measurable with existing instruments.

Not only will we thus be able to confirm greater pleasure at choosing Option 2 for the same subjects, I'll bet when we compare subjects who willingly chose Option 1 the intensity of the pleasure response will be greater in the Option 2 people, except when Option 1 choosers are confirmed sociopaths.

But then, the differential prediction will be that the Option 1 choosers will have shittier lives overall than the Option 2 choosers (i.e. Option 2 people will score higher on life satisfaction, both on subjective and objective markers of personal well-being), which is in fact the main prediction of my theory.

Richard Carrier said...

Regarding the computer analogy, computers only make the decisions we tell them to.

If we wanted to program a computer that could make decisions on its own (as in, informed rational decisions, not merely unexpected ones), we would have to give it desires (otherwise it would have no basis for ever choosing one thing over another), and the satisfaction (fulfillment) of those desires would have for them the functional equivalence of pleasure responses in us. It might feel different to them, in the way a wolf sees things differently than we do, but just as a wolf nevertheless still sees things, so a computer will nevertheless be defining its decisions according to some ultimate goal of satisfaction (of a hierarchy of desires) which will be that computer's functional equivalent of happiness.

Richard Carrier said...

"There is a state of affairs called "satisfaction" (S) and we are motivated to pursue S. However, for all other states of affairs T that are not equal to S, not only is it the case that we have no motivation to pursue T, but we cannot have any motivation to pursue T, according to Carrier."

I don't recognize anything I have said in this description. I don't know what T is supposed to refer to. For example, if T = {hacking off our arms and legs to win a bet at cards}, then wouldn't the above statement be obviously true?

For any T in which the above statement is not true, T contributes to S, and therefore S provides ample motivation to pursue T. I highly recommend you read Drescher's Good and Real for a decisive proof of this.

Richard Carrier said...

"It is possible for the mother in the above case to be motivated to bring about some state of personal satisfaction, but it is not possible for the mother to be motivated by the state of her child being healthy and successful except insofar as the child's wellbeing is a means (a tool, an instrument) for bringing about a state of satisfaction.

That's too narrow-sighted.

The idea of the child's wellbeing is also a means for bringing about that satisfaction, as is the knowledge of self. The mother has a choice really not between two fates for the child, but between two fates for herself: will she become (in her moment of deciding) the sort of woman who would choose Option 1 or the sort of woman who would choose Option 2?

Her satisfaction comes from the effect of that decision, which is wholly independent of any causal physical connection to what actually subsequently happens to the child. It is rather caused by the knowledge of what will (or was promised) to happen to the child, and the causal effect that that knowledge has on the mother's brain when conjoined with the additional causal effect of the knowledge of what sort of person her choice makes her, is an overriding satisfaction (i.e. a greater satisfaction than would occur had she chosen Option 1). And in fact, on a complete analysis (if she is fully informed and has made a fully rational choice), I predict that any woman for whom this is not the case will choose Option 1.

Now, we could design a mommy robot that has as its supreme goal the wellbeing of her child (i.e. fulfilling no other desire brings as great a satisfaction), though that would not accurately model human reality or the human mind or human behavior or decision-making. But even then the general principle will be confirmed: the satisfaction of that supreme desire will be the functional equivalent of happiness for the robot.

But perhaps there is a confusion here: what, after all, is supposed to be the difference between fulfilling a desire and satisfying a desire? In neurophysiology, is it even physically possible for a human to have a desire the fulfillment of which produces no satisfaction? I predict the answer will be no, if that isn't already a scientifically confirmed fact.

But even if such desires exist, desires that do produce satisfaction will always override them, unless there is some physical causal mechanism by which they can override satisfaction-causing desires, and your theory would have to make a prediction in that regard, and would remain an unconfirmed theory until that causal mechanism was observationally confirmed.

I suspect the only reason desires cause satisfaction is that we evolved that mechanism as a motivator (the basic pleasure-pain response is the substrate on which all more complex higher thinking is built), which is why it is very unlikely we have any other mechanism that is more effective. If we did, we would call it's effect "satisfaction," or something even more satisfying than satisfaction, unless those satisfaction-overriding desires caused us to act in a way that dissatisfied us (like some sort of monstrous puppet routine that causes us to act against our will), in which case you would have to explain why we would do things that never satisfied us (i.e. what physical mechanism does that, and how did it evolve in us, and why should we agree with what it tells us to do, rather than instead obeying the desires that satisfy us).

"I would suggest that the desire fulfillment theory leaves fewer mysteries then the satisfaction theories. If it is true that both theories yield the same predictions, the fact that the desire fulfillment theory does not leave these types of questions unanswered gives it an advantage.I don't see any unanswered questions. All I see is a considerable amount of ignoring all the scientific facts we have learned about actual human decision making. You sound a lot like Copernicus, when I'm arguing from the position of Kepler. Perhaps I'm missing your point?