 Hello, welcome, and thank you for joining us. I'm Kathleen Acock, the Director of Communications in the Graduate Division at the University of California, Berkeley. I encourage all viewers of this lecture to interact with us in the comments section of YouTube. If you could take a moment to share with us where you're joining us from and describe a little bit about your interest in philosophy, we would appreciate it. You may also submit questions throughout the lecture using our Google Form, which I've linked in the YouTube chat. This Google Form is also linked from our Berkeley Graduate Lecture's event webpage. I'm thrilled to introduce our moderator, Professor John Campbell, the Willis S. and Marion Susser Professor of Philosophy at UC Berkeley. Hello, good afternoon, and welcome. It's great that you're able to join us for this Howison lecture. Let me say something about what it means that it's a Howison lecture. George Holmes Howison was born in 1834. When he was 50, he came from the universe after an interesting sheckard career. He came from the University of Washington at St. Louis to take the first endowed chair in philosophy at Berkeley. And Howison built our philosophy department. He was clearly a gregarious, sharp, influential, generous, and much-loved individual. When he died, his friends and colleagues put together a fund, the fund that is funding today's lecture to continue his work. And the idea was to bring the most influential thinkers of the day out here to the rural wilderness of California. Now, for familiar public health reasons, today's session is remote. But of course, that brings opportunities as well as problems. And I very much hope that wherever you're writing from, you'll just let us know where you are and do that on the comment forum, as Kathleen suggested. And we're delighted to have Robert Stalnaker here with us today. He exactly fulfills the terms that Howison is thinking of. He is one of the most distinguished philosophers alive, having made many fundamental contributions to different areas of philosophy. Stalnaker received his PhD from Princeton in 1965. And over the next 50 years, he taught at Yale, Urbana-Champaign, Cornell, and MIT. His influence on generations of graduate students and faculty has been enormous. His work combines spare lucid rigor with great philosophical illumination and depth. He's worked on philosophy of language, particularly the foundations of pragmatics and metaphysics, issues concerning modality and conditionals, and epistemology, particularly the logic of knowledge and contextualism about knowledge. This talk today, I'm excited to tell you, is called Counterfactuals, Compatibilism, and Rational Choice. So please join me in sending vibrations of welcome to Professor Stalnaker across the ether, wherever you are. Professor Stalnaker. Thank you very much, John. And thanks to the House and Committee and the Department for inviting me. It's a great honor and a pleasure to have the opportunity to present these some development, present some ideas. So my talk, I have a handout which is available, normally, be passed out if we were all there, which is available on the website of the lecture. So if you haven't got it, you don't really need it, but it'll help a little bit to remind yourself some of the details. So there's been a lot of philosophical ink spilled over the years about the question whether determinism is compatible with free will. And I'll spill a little bit more on this issue later, but my main concern before getting to that will be with a more general and more abstract puzzle about counterfactual conditionals and determinism. A question about whether, in the context of determinism, we can make sense of the idea that history might have taken a different course from the one it is actually taking. Roughly stated, the puzzle is this. It seems that the assumption that we live in a deterministic world implies that a complete description of the intrinsic state of the world at any given time, together with the laws of nature that are true, entails all the truths about the state of the world at any other time. And this seems to imply that if the course of events had been different from the way it actually has been in any way, however trivial, then either some law of nature would have been false or else the intrinsic state of the world at every time would have been different, at least in small ways, at every time, no matter how far back. So, for example, if I had tied my left shoe before my right this morning, which is sort of the way I did, either some event that happened thousands of years ago would have happened differently or else the laws of nature would have been violated. Or as we'll see later is David Lewis's colorful way of putting this, a miracle would have occurred. Neither seems plausible. Now, the relevance of this puzzle to issues about the compatibility of determinism and free action is clear enough, but the puzzle itself is more general. Since even if the relevant counterfactual is about events that involve only inanimate objects where no issues about action were involved, the conclusion is still unsettling. It would be reasonable to believe that in some circumstances, what happened, my screen went, oh, there he's back, right there. It's reasonable to believe in some circumstances, if it hadn't rained last night, the streets would not be wet this morning. But this does not seem to imply, or it would be strange to say that it applies, that if it hadn't rained, then some intrinsic state of the world 2,000 years ago would have been different from the way it in fact was. Or else, some law of nature would have been violated. The general puzzle was made prominent by David Lewis' discussion of the semantics for conditionals, and by his application of that semantics to counterfactuals. On Lewis' general semantic account, a conditional is true in a given possible world, W, if and only if the consequence of the conditional is true in all of the possible worlds that are most similar in relevant respects to world W. The substantive task of a theory is to spell out the relevant respects of similarity. Now Lewis had a number of metaphysical axes to grind in carrying out this task, which we needn't necessarily share. First, his goal was a reductive analysis to specify the relevant respects of similarity in terms that are aimed to spell out the relevant respects of similarity in ways that are intelligible independently of any causal concepts. So that was one thing. Second, he was aiming in an account, a subsidiary aim, an account of respects of similarity that is temporally neutral. His thesis was that the temporal asymmetry of counterfactuals, obvious intuitively, the asymmetry, and so the temporal asymmetry of counterfactuals and derivatively of causation was the result of asymmetries in the global distribution of particular facts. So in particular, it was important for Lewis to make room for backward causation and time travel, even if our world does not exhibit those kinds of causation or events. So Lewis's formal semantics for conditionals helps to give the general puzzle a sharp formulation and his metaphysical aims and commitments help to explain his particular response to it, but the puzzle itself depends only on pre-theoretic intuitions about ordinary counterfactuals, intuitions that seem difficult to reconcile with determinism. Now the puzzle gets a thorough and careful discussion in a paper that was published about five years ago by Kian Dorr. After giving a rigorous definition of the metaphysical thesis of determinism, not exactly the way I'm gonna explain it, but close enough, Dorr states then three premises about a representative to using a representative counterfactuals as an example, three premises that seem intuitively plausible. And then he derives from them a conclusion that in effect reduces counterfactuals to triviality on the assumption that determinism in his sense he defines is true. That is trivial in the sense that all counterfactuals will turn out to be true. He then examines the premises in conclusion one by one, considering which bullet to bite. Should we reject premises one, two, or three, or should we accept the triviality conclusion in the counterfactuals really don't make sense? The choice is narrowed down to one, reject the premise that the relevant world in which counterfactual supposition is true must be worlds in which the intrinsic state of history, the past history is exactly as it was in the actual world. That's the first premise. Second premise, rejecting the premise that the laws of nature that are true in the counterfactual world must be true in the relevant counterfactual worlds. That no miracles occur in the counterfactual world. Dora's conclusion is that we should take option one and we should hold on to the second premise. There's reject option two. The argument proceeds by bolstering the second premise with examples that there are no miracles and arguing that we can get by with a weaker version of the first premise, one that assumes only that the past history of the counterfactual world must be approximately the same as they are in the actual world. Dora's argument was aimed at Lewis who chose option two and argued for it. That is Lewis wanted to defend the exact similarity of sequences of particular events of priority over small deviations from the laws of nature. Since Lewis had described law violating possible worlds as worlds in which a miracle occurs, Dora titled his paper against counterfactual miracles. Now, in the end, I will have to address, answer Dora's challenge on his terms. But in this talk, I'm gonna take a more indirect approach, aiming to set the problem up in a way that avoids choosing between the two options that he focuses on. And just to give you a spoiler alert, I'm gonna argue that we should reject both of the premises or at least it'll be a consequence of what I'm gonna argue, but once you reject both of the premises, the Dora asks us to choose between. So here's my plan for today. I'm going to first make some brief remarks about determinism, laws of nature, miracles. After this, I'm gonna put two examples of simple counterfactual situations on the table and use these examples to draw some intuitive conclusions about the relationship between determinism, counterfactuals, and other causal notions. I'll then digress to make a brief remark or two about the ways in which counterfactual conditionals are context dependent or context sensitive. And then, and about the relevance of context dependence to the puzzle. In the last part of the talk, I'm going to turn to the more specific tension between determinism and counterfactuals about voluntary action, arguing that getting clear about the general problem helps to clarify the question whether a version of compatibilism can be defended. In particular, I'm gonna argue that there is no incompatibility between determinism and the standard way in which decision situations are modeled in decision theory. I'm gonna conclude by looking at an example and an argument against this version, particular version of decision theory that I will put on the table. And it's an argument that involves the hypothesis that determinism is true. Okay, so now if you have a handout, we're on the section one on laws and determinism. Let's, and this is just really a sequence of quick definitions, not spelled out with any precision and qualification, but just a rough idea of what determinism says. So let's say that a dynamic theory is a theory that has the resources to define a notion of a state of some system, that is the system that's subject matter of the theory, at a time, so it defines the state at a time and it also has a set of laws that constrains the relationship between the states of the system at one time, any one time and the states at other times. Because that's what a dynamic, roughly what a dynamic theory is, a lot of the theories that scientists talk about are dynamic theories of this kind. Say that a theory of this kind is deterministic if and only if the propositions describing the complete state of the system at any one time together with a set of laws entails the state of the system at all other times. Or in some versions of determinism which are only forward looking, not backward looking, it entail the state of the system at all later times from that time to that state. Now we might say further that that's what a deterministic theory might roughly be defined, but we might say that a metaphysical thesis of determinism is true if and only if a deterministic theory of this kind is true and furthermore is comprehensive. Meaning by comprehensive that all the facts about the world are supervenient on determined by the state of the, by a theory of this kind. Okay, so of course, not all things we say about the world are stated in the language of a deterministic theory, but they got to be somehow a different way of talking about the same thing. We might say then further that a fact whether it's stated in the theory or not about a particular time is intrinsic to that time if it is supervenient on the state of the system at that time. Okay, this is a rough and ready characterization of determinism at least some questions open and it has some contentious metaphysical presuppositions but it's enough to raise the puzzle. I'm skeptical about some of the metaphysical presuppositions particularly concerning supervenience and comprehensiveness which I aren't as clear as people often take them to be but I set these worries aside here since I think the puzzle can be diffused without questioning the sort of metaphysical fundamentalism is involved in the definition of determinism. I do however want to point out that while my characterization of determinism does presuppose some kind of metaphysical fundamentalism it does not require the particular metaphysical foundation that David Lewis defended which he called humane supervenience. On Lewis's humane picture the fundamental properties that define the state of the world at a time are wholly independent of the laws of nature. On his account the basic purely natural properties on which the states of the world supervening have in themselves no causal or dispositional consequences. It's really the opposite of that. Some other people have defended the view that what it is to be a property at all fundamental property or not is to be defined in terms of causal powers. So Lewis takes the opposite line on this that fundamental properties have no causal consequences. The characterization of dynamic theories in terms of states at a time in laws and the general characterization of the thesis of metaphysical determinism does not rule out the possibility that there are metaphysically necessary connections even semantic connections between the laws and the states or that there are fundamental theoretical properties that are intrinsic to the things that have them at a time but are also dispositional meaning they have constitutive connections with the way things respond or the developers time goes on. If the fundamental natural properties were dispositional this would spoil Lewis's carefully developed reductive hierarchy essential it was kind of reductivist, humane subvenience. So the hierarchy goes first you explain laws of nature independently of any causal notions. Second, you define counterfactuals by giving criteria of comparative similarity of worlds that depend on the laws but don't depend on any causal notions. Then you define causal dependence in terms of rather counterfactual dependence in terms of counterfactuals and you define causation in terms of counterfactual dependence and each step involves a lot of back and forth. But anyway, that picture that reductive or hierarchical conceptually hierarch conceptual hierarchy is not something that we are going to assume and the puzzle still faces us even if we don't take that. Now I will assume following both Lewis and also most others including Dorr and his argument who talk at all about laws of nature that the laws are true. That is L is a law of nature in a possible world W if and only if L is true and not if and only if L is true in the world W. So it's a subset of the truths in that world. Laws are usually assumed to be contingent truths. So they are propositions that might be false but L would not be a law of a possible world W in which it was false. So if a miracle is defined as a violation of a law of nature then in a sense there can be no miracles. If L is a law of nature in world W then there are no violations of L in world W whether W is an actual world or a contractual world. And Lewis is perfectly clear on this point. He recognizes as a kind of playful misdirection in the very word miracle to describe a violation of the law of nature. But the upshot is that to characterize a possible situation or possible world as one in which a miracle occurs is to give it an essentially counterfactual characterization. To adapt Russell's famous example one might wish that one's yacht were longer than it is but this need not be a wish for the impossible. Similarly, one who wishes for a miracle in the literal and perhaps misleading sense in which it's taken in this puzzle is to wish that the laws of nature were different in a way that did not preclude a certain event that they in fact preclude. And I'll have more to say later about the idea of an essentially counterfactual characterization of a possible situation. Okay, now we're moving on to if you have the handout to the second section I'm just gonna give two examples, two simple examples. These aren't puzzle cases or wild and word counterfactual stories like Twitter or Swampman or something. They're simply straightforward stories of situations in which a counterfactual question naturally arises. But I'm gonna make two points about what I take the examples to show. Okay, first example and this is spelled out on the both spelled out on the handout. It's the last day of the tournament we're on the 18th green. Anika sets up for an 18 foot putt for a birdie which would give her a victory in the tournament. After careful consideration taking account of the wind conditions, the slope of the green and all that, as they always do, she sends the ball on its way and it looks like an excellent chance of reaching the cup and falling in. But when it is just a few feet away the droppings of a seagull who happened to be flying above land right on the ball, slowing and diverting it so that it stops a bit before its intended destination. I don't know what officials do in tournaments when things like this happen, but that's not what we're gonna talk about. Would the putt have sunk if it hadn't been for the bird? We can't know for sure since even putts that look like winners when they're just a couple of feet away sometimes stop just short of the edge of the cup or slide by, but it seems likely that it would have. And it seems certainly on the assumption of determinism that there's a factor of the matter in the actual world about whether it would have. But the spectators in the spectators later discussion of this question would putt have gone in or not, but the philosopher raises the following issue. In the counterfactual situation or situations we're considering what prevented the gull from being there and doing what it did? Did some small miracle intervene diverting the bird's course perhaps resulting in the dropping of its dropping, its load on a car in a nearby parking lot or should we assume that the history of the life of the gull was different at least in minor ways from its beginning? Or maybe because of a small difference in the distant past that gull never existed at all. The other spectators dismissed the philosopher's question as totally irrelevant. It's irrelevant they might have said because whatever explains why the bird happened to do what it did. That fact has no bearing on the event leading to the ball taking exactly the path that it took up to the point where the bird's behavior resulted in an intervention in that causal chain. Okay that's the first example. Second example, easier to follow if you're a poker player but that's not necessary. Sly Pete and Mr. Stone, these are two characters from a famous example and cocted by Alan Gibbard for a different purpose. These two guys are playing draw poker on a Mississippi riverboat. But this is a different story from Gibbard's with no cheating involved as his had. Pete has two pair, threes and tens and an ounce with an outside ace. Mr. Stone has the five, six, seven and eight of clubs along with a useless queen of diamonds. Pete deliberates about whether to trade in his ace in the hopes of turning his two pair into a full house but he decides to stand back. Mr. Stone trades in his queen, useless queen with a hope of turning his hand either into a straight if he draws any four or nine to complete the sequence of five cards in order of number or a flush if he draws any club turning it into a hand with all clubs. But since the top card on the deck is the 10 of spades his hand remains worthless after the draw. So he falls when Pete bets. But what would have happened if Pete had traded in his ace in order to try to get a full house? There's no question in this story no matter of debate, he would have drawn the 10 that in fact went to Mr. Stone, the top card on the deck and therefore he would have gotten his full house for three tens and two threes. Stone would have gotten the next card from the top of the deck which happened in fact to be the nine of clubs. So Mr. Stone would have gotten not just a straight or a flush but both, a straight flush. Both players would then have bet big since they both had excellent hands leading to a much more exciting finish with Mr. Stone's straight flush, beating Pete's full house. What the philosopher says, wait a minute we need to consider why Pete would have made a different decision in the relevant counterfactual situation. Was there a little miracle? Perhaps a tweak in his brain, alter his deliberation or was his past history different in perhaps minor ways from the beginning leading to a different thought process and a different decision? Once again, the others in the discussion dismiss his question as totally irrelevant. It's irrelevant they might have said because whatever factors might have led Pete to try for the full house they would not have changed the fact that the top two cards on the deck were as they were. Okay, now two observations about these stories. First, the judgment that the philosopher's questions are totally irrelevant seems intuitively to be completely correct. There is not a, this is not a response this observation is not a response to the argument given in stating the puzzle. We still need to answer it. And it's not a reason to dismiss the argument that Doran spells out in such detail. But it is a datum about counterfactuals on a par with the intuitions about the plausibility of the premises of the argument that are, it's a datum that our response to the puzzle should explain. That is we want to account of the respects and similarity to which there are counterfactual appeals to explain why the concerns of the philosopher, of the philosopher raises in these two examples play no role at all in our judgment about the counterfactuals, the truth or falsity of the counterfactuals that are at issue in these cases. Okay, that's the first observation. The second observation is really the main point is that the intuitive explanation of why the counterfactual questions have the answers they seem to have are entirely compatible with determinism. These explanations appeal to causal chains that are independent of each other, at least up to a certain point. That is to different sequences of events with causal stories that are independent up to a point at which they intersect. The notion of causation and causal independence may play no role at all in the characterization of a dynamic theory or in a definition of determinism, but causal distinctions can be made with the resources of such a theory. Determinism may imply that the state of the world as a whole, all the facts about a certain state of the world at a certain time, the state of the whole world at a certain time, contain the causes of the state of the world as a whole, any of the events at any subsequent time. So the cautious determinist, if asked for a determining cause of a certain event can safely say that it is to be found in a complete description of the state of the world at some time prior to the time in which the event occurred. In fact, at any time prior to the time in which that event occurred. One might, however, want a more informative answer that appeals to some more limited features of the prior state of the world that together with the laws suffice to determine the particular event in question. And while the thesis of determinism alone makes no claim beyond the global one, any specific dynamic theory deterministic or not can be expected to imply some more specific claims about ways that partial descriptions of the state of the world at one time constrain the state of the world at certain later times. And generalizations of this kind may contribute to characterization of notions of causal independence and causal change. And these may play a role in specifying the respects of similarity that are relevant to the interpretation of counterfactuals. In Lewis's reductive project with its hierarchy of definitions, causal and counterfactual dependence are defined in terms of counterfactuals. And any appeal to such notions in specifying the respects of similarity that are relevant to the truth conditions of counterfactuals would be circular. But we have left that project behind. And even if you go for that project, you've still got, then you've got analyses of these causal notions, but there still can be generalizations about the role of counterfactuals in relation to this relevant similarity notions. And so the compatibility with determinants is still one you have. Okay, we'll come back to the point, the second point, but I wanna digress now to say a little bit about context dependence because everybody agrees that counterfactuals are context dependent. And in particular, there are different kinds of counterfactual, famous examples of cases where an indicative conditional or the epistemic conditional has a totally different meaning from a counterfactual conditional. So the famous example of, if Shakespeare hadn't written Hamlet, nobody would have written it, but if Shakespeare didn't write Hamlet then somebody else did because Hamlet was written. So epistemic conditionals have different truth conditions from counterfactuals. And then there are so-called backtracking conditionals which sometimes have different truth values from forwards as where you say, if so and so it happened, that would have been because sudden such would have happened before we look back. But sometimes what is assumed is that the way in which counterfactuals are context-dependent is wholly determined by a general specification of respects of similarity used to interpret particular conditional sentences or thoughts. And it's sometimes also assumed that there's one standard compared to similarity relation that's relevant to the interpretation of the kind of counterfactuals, they give rise to the puzzle. One that contrasts with what are called backtracking conditionals and also with epistemic conditionals that are interpreted with a different comparative similarity relation. And while the comparative similarity relation is treated as a parameter determined by context, people making the point that counterfactual context-dependent say little about what a context is or what it is about the context that determines the value of this parameter relative to which counterfactuals are interpreted. Now, I think one has to look at the question, what is a context? I've defended the view in other places that conversational context should be modeled by the common ground, the set of possibilities that are compatible with what the participants in the conversation presume to be common knowledge. Common ground is a certain information state. It's definable in terms of propositional attitudes of a certain kind of the members of the relevant group, though it's not a propositional attitude of any one person, but rather something definable in terms of their attitudes. It's been suggested, and I'm in full agreement with this suggestion that an important element of certain propositional attitudes is this which understand them, both knowledge and belief, as an element of a propositional attitudes, a set of relevant alternatives determined by a set of questions at issue. So linguists have talked about this, Craig Roberts, for example, Jonathan Shaffer, contrastivism about knowledge in other places, talks about the partition of relevant alternatives. My favorite paper on this issue, defending this line is by Seth Yeltsin, which I put in a handout, references in the handout, talking about a kind of grain of representation. So you take a partition of a space of possibility as you're an element of what it is to believe. Believe is a matter of having a certain way of distinguishing between those relevant alternatives. A context set that models the common ground if you follow this suggestion is not just a set of possible total universes compatible with what is presupposed or taken to be common ground, but rather it's a partition of logical space that determines those distinctions, includes a partition of logical space that determines those distinctions between possible worlds that are relevant to the issues under discussion. So this assumption applies when we apply it to counterfactuals, that the parameter of interpretation of counterfactuals that context determines should be a comparative similarity in relation on the members of a partition of a set of possible worlds, the relevant alternatives. Now Saul Kripke among others has made the point that the formal apparatus, so if you can look at the formal models of these contexts and of attitudes and so on, the formal apparatus of possible world semantics applies naturally to cases in which what plays the role of the possible worlds are what Kripke called mini worlds, other people do, defined by perhaps a perhaps coarse-grained partition of a space of possibilities. Remember naming necessity in the introduction, Kripke talks about a little toy example of 36 possible states of the world. That are the possible ways that a pair of dice could land. And we do probability exercises with such a set of possible worlds. But as Kripke says, we just ignore the other differences, but the formal apparatus applies perfectly well. On this view, the context for interpretation of a conditional determines not just a particular comparative similarity relation, but also a set of elements on which the relation is defined. The context can be expected to determine some general constraints on the respects of similarity that should be held fixed in the most similar of the relevant alternatives. But the application of those constraints will depend on the particular partition of logical space that determines or is determined by the questions at issue. And the questions at issue may themselves help determine the general constraints that are relevant. Now, the hypothesis that the terms of the similarity relation are relevant alternatives modeled by a partitioned cells of logical space as consequences for both of the theses, the door asks us to choose between these discussion of the puzzle. First, that the past of the one of the assumptions that he wants to reject, the past of the counterfactual event must be exactly rather than merely approximately the same as it is in the actual situation. And two, that the actual laws of nature must be true in the counterfactual situation rather than just approximately true, permitting only small miracles. We might suppose without contradicting determinism, that the counterfactual past is exactly the same as the actual past with respect to the questions at issue. And that the causal relations between events in the counterfactual situation are fully compatible with all of the actual laws. But if we're talking not about the relevant alternative possibilities, but about the whole universes that realize those possibilities, then either of these similarities might be only approximate. The questions at issue and the assumptions that they are part of the context help to sharpen the kind of approximations that are permitted. I mean, one of the things when you move from exact similarity to approximate similarity, you immediately just open everything up because you can always sort of adjust what's approximately true to suit your purposes. But if you have questions at issue that helps to pin down, approximately means exactly with respect to those questions. Okay, in line of these, now we're moving on to section four and hand on. In light of these general considerations about counterfactuals, I'm gonna look at one particular kind of context in which counterfactuals as well as epistemic conditionals play a particularly a prominent role. Context of deliberation and context in which the decisions that are made are assessed. The main questions at issue in a context of deliberation are in the case of deliberation, which of the set of alternative actions to choose and what the consequences of deliberation are in the case of deliberation, which of the set of alternative actions to choose and what the consequences of those choices would be and in the case of assessment, whether the choice made was the right one and more generally, how that choice compares with alternatives to it. The way that decision theory, at least causal decision theory, models such contexts, the questions at issue and the notion of causal independence play an explicit role. Now, in David Lewis's way of spelling out the model and David Lewis is here. This is not connected with his humane supervenience and dependence project, but Lewis's way of spelling out a model for decision situation. There are two cross-cutting partitions of a space of alternatives. One determined by a set of actions that are presumed to be available to the agent and the other specified in terms of the action partition. The second partition, which Lewis labels the dependency hypotheses is defined as follows for a particular agent and time. A dependency hypothesis is in this quoting Lewis a maximally specific proposition about how things the agent cares about do and do not depend causally on his present actions. The idea is that the dependency hypotheses are propositions that are causally or counterfactually independent of the propositions in the action partition, but that when conjoined with one of the action propositions entails answers to all of the questions about things the agent values positively or negatively. So the grain or partition in this case is going to be the intersections of these two cross-cutting partitions. And that's a partition itself. So the main lesson I wanna take from a general discussion is that this kind of model of a decision situation and the notion of causal independence that it presupposes is not in conflict with determinism. If our general discussion succeeds in diffusing the general puzzle, I think it goes a long way toward diffusing the consequence of it. Though it must be acknowledged that I fully satisfactory compatible response to that argument will require that we say something constructive about the basis in a model of the kind that Lewis outlines, decision theory model for the basis for describing the first partition as the set of alternative actions or action propositions. That is for describing it as a set of alternatives that are in some sense available actions for the agent at the time. But getting clear about counterfactuals and their relation to determinism helps make room for a representation of a decision situation that clarifies and sharpens the challenge that a compatibilist needs to meet. Now most of this is further work that I'm here just trying to set up but in carrying out that further work the specter of determinism will continue to raise its ugly head. I'm going to conclude today by taking a quick look at an argument against the adequacy of this kind of model of a decision problem. At least in the context where the thesis of determinism is presupposed. Okay, so now we're on to the last part of the handout betting on the past. So Ari Amit has developed the most resourceful and imaginative critique of causal decision theory. And one of his arguments is that if we assume determinism this decision theory gives the wrong answer to what a rational agent should do in the following decision problem. Okay, so this is a quotation from him stating the problem. In my pocket says, Bob, I have a slip of paper on which you've written a proposition P. You must choose between two bets. You got a bet on P but you can do it at two different sets of odds. Bet one is a bet of P on 10 to one for a stake of $1. Bet two is a bet on P at one to 10 for a stake of $10, things of opposite odds. Before you choose whether to take bet one or bet two I should tell you what P is. It is the proposition that the past state of the world was such that such as to cause you now to take bet two. Amit suggests that we might make P this proposition more precise by identifying a particular time say one that is many years ago, many years before you and or Bob were born. You know, 1900s at just 1000 BC, whatever. He then argues that it's obvious you ought to take bet two. This is the normative premise, the intuitive premise about what rationality requires. And that causal decision, the second conclusion he draws is causal decision theory implies you should take that one by a principle of dominance that the causal decision theorist accepts. So just to get this clear, this is the matrix that he draws for this problem. So P and not P are distant past statements, statements about the distant past. So we take them to be the dependency hypotheses. Bet one, if P is true, you win $10 if you take bet one and you win $1 if you take bet two. So bet one is better if P is true. If P is false, then you lose a dollar if you take bet one and you lose $10 if you take bet two. So it's still better to take bet one. But bet two is obviously what you should take. And I agree about the normative premise, but the mistake in Ahmed's argument is that is that one should take bet one or the other one should take bet two. The mistake in Ahmed's argument is the assumption because the truth or falsity of P was settled by facts about the distant past as all facts are, it is therefore entailed by the true dependency hypothesis. But the point of the causal decision theorist framework was to distinguish features of the world that are counterfactually independent of the action from those that are not. The compatibilist acknowledges, Ahmed is a compatibilist. He just accepts the premises of the argument against compatibilism. The compatibilist acknowledges that the agent's choice has a causal history and that if determinism is true, then events in the distant past determine both the agent's choices and the events that would have happened, whatever choice had been made. So I think the person facing the choice, let's call her Alice, should reason like this. Assuming the thesis of determinism, which both Bob and I presuppose, to say that the past is such as to determine that I choose bet two, is just a roundabout way of saying that I choose bet two or similarly for bet one. So there is really just one dependency hypothesis that says the following. If I choose bet one, I lose a dollar. And if I choose bet two, I gain a dollar. It's a no-brainer. I go for bet two. In fact, unlike Newcombe's problem, you can easily play this game at home. Assuming that it's, for Newcombe's problem, you need a super predictor. But here you don't need any special powers. All you have to do is to believe determinism or to be prepared to presuppose it. So I might offer you a dollar to play the role of Bob. And you can ask yourself, would you accept if I did? If you accept, you probably won't make any money, but you can't lose any. In deciding whether to take my offer, Alice's choice is your dependency hypothesis. As you don't have any control over that, that's something that she'll do whatever whether whoever it is for the bet. If she takes bet one, you gain an extra dollar. And if she takes bet two, as she surely will, you break even. So, if you wanna test out your views about free will, we'll go home and play this game. Now, the reason you can play this game at home without having any special insight into the distant past or into Alice's state of mind is that Alice's choice is causally relevant to whether Bob pays or collects, even if it's not causally relevant to whether P is true or false. Now, this is a loophole which Ahmed recognizes in his story and he suggests that we might block this feature of the story by supposing that the payoffs in the matrix above represent not money that Bob would give her or collect from her, but rather just utility values that she attaches directly, that's quoting him, values that she attaches directly through the possible outcomes. Suppose for some reason, it doesn't matter what it is, she ranks her preferences over possible states of affairs this way. Best outcome, take bet one while P is true. Second best outcome, take bet two while P is true. Third best, take bet one while P is false. Worst outcome, take bet two while P is false. And then you can read that off the matrix by saying what the four alternative possible outcomes are as represented in that matrix. But notice that the best and the worst of these four descriptions of possible situations are essentially counterfactual specifications like the situation you wish for when you wish that a yacht were longer than it is. A situation in which bet one is taken while the past is such as to determine that bet one is not taken, I'm sorry, a situation in which bet one is taken while the past is such as to determine that bet one is not taken must be a situation in which a miracle happens and things that were determined to happen don't happen where some actual law of nature is false. But no miracle occurs in this counterfactual situation. The laws that hold in that situation are not vital. Situations defined by an essentially counterfactual specification are not options available in deliberation. Although in a retrospective assessment of a decision, some options one did have may be specified in that way. Let's define, for example, a road not taken as an action when considers in deliberation but in the end rejects. O'Leary is always second guessing his decisions, often regretting the choices he has made. Noting this trend, he resolves in the future to avoid regrettable choices by sometimes taking roads not taken. And he's frustrated by the fact that he never succeeds. He concludes that he is not as free as he had thought. The fatalism argument, not the consequence argument, but the fatalism argument turns on the conflation of the sense in which one cannot choose what one does not choose and the sense in which sometimes one can or at least could have. The consequence argument, I think, turns on a more subtle version of his fallacy, although there's obviously a lot more else going on. Now, in my homie version of Ahmet's story, the one you can play at home, Bob's knowledge of whether P is true or false is causally downstream from Alice's action. But as Ahmet observes, it might be that Bob knew this prior to and independently of his choice. This is certainly possible. And in fact, it's not at all unrealistic. Many free decisions are predictable and in many cases one could say that someone before the fact knew what I was gonna do and that wouldn't prevent me from freely doing it. Many free decisions are predictable, particularly when there are no brainers like Alice's choice. She knew what she was gonna do before the time came to commit herself and Bob, knowing that she was clearheaded, might have known that she was the kind of person who would make the rational choice. That does not imply that Bob's knowledge and so Alice's choice is a consequence of the true dependency hypothesis. The counterfactual, if Alice had chosen differently, Bob's belief would not have been knowledge is a true counterfactual. The contexts of deliberation and retrospective assessment of decisions taken involve a complex interaction of temporal, causal, normative and epistemic considerations to give rise to a range of puzzles about choice and about an agent's capacity for choice. I think the causal decision theorist model helps to sharpen and clarify some of those puzzles, but we need to disarm the tacit assumption that determinism obliterates all causal distinctions in order to make room for a project of this kind. Thank you. Okay. Okay. Well, thank you for that beautiful talk. Thank you. Professor Zellmacher. Okay. We'd usually, at this point, stop for buns and cookies and wine, but not being able to do that. Maybe while we were waiting for people to type in their questions, I could just ask you a little bit about what you just said. Yes, go ahead. In particular, how you're handling backtracking conditionals, the idea is that to determine which possible worlds are relevant to the interpretation of a counterfactual, we should be looking at the common ground and what is happening to the common ground and the conversation, and in doing that, we can help ourselves to the notion of causation. Is it causation? Yes. Yeah. So we don't need the Lewis, the very austere thing where we drop causation here and try and define similarity without causation. Right, yeah. That's an important move. But that's not really essential to the story because Lewis would not disagree that causation is going to play a role in practical assess evaluation, but causation itself should be analyzed in other terms. But still, I do think that this whole notion of a hierarchy of definitions plays a role in a number of places in these arguments. And I think the idea that we're looking here for conceptual interconnections and we can have them, constitutive connections, say between causal notions and temporal notions and so on, even if we don't have any analysis that's sort of reducing things step by step in some way. I don't want to say about backtracking conditional. I mean, one of the things that, the way normal way of thinking about backtracking additional is that there's just different criteria of similarity. But I think the main thing is different questions that issue. So the question, if the question is, how did it get that way? How did a certain fact get that way? One looks back to the causes of it. And when considers what would have had to been different for the thing to be different. So they're different questions at issue. And in some cases, that's gonna force you to have different priorities with respect to what's more possibilities and more similar to than others. But it's different possibilities that you're considering is the main thing. And as for the most part, a backtracking conditional doesn't assume different criteria of similarity. It just asks a different question. But in sort of memorable cases as different, like Frank Jackson has this example of, you know, if he jumped off the bridge, he would have died, you know, no, he wouldn't have because he wouldn't have jumped off the bridge unless there were a safety net below. Yeah. And so that's going to a totally different counterfactual situation. But you can also ask a backtracking conditional, which really orders things in the same ways. And in my slide, Pete exam, he might ask, how come he didn't take a card? He should have. It would have been much better strategy given what he knew to take another card, but he didn't. Why didn't he? Well, that's a different question, but it's a question about the same situation. Okay. Just one last thing in this, while we're waiting for questions to come in, which I hope people will type in. When you were talking about practical deliberation, when you moved to talking about practical deliberation, I may have been naively hearing this, but I was thinking that sometimes backtracking conditionals are very relevant to deliberation as to what to do. I mean, if, for example, you and I are burgling a house together, you know, something we've done many times and your task is to get inside the house and loosen a window for me to get in. When I get to the house, if I find the window loose, I think, ah, if it's loose, then my colleague must be inside the house. That can only be because my colleague is inside the house. So now I know how to proceed. So in practical deliberation, then we're often going to have to consider backtracking conditionals. Yeah. So it's certainly not that we want some way of thinking about the interpretation of the conditionals in practical reasoning that will simply let us sideline the backtracking and say, I only wanna look at the future. In fact, the context of deliberation, these are epistemic conditionals too. I mean, that is you're trying to, you haven't decided yet what you're gonna do. The main question that issue involves, if I do this, then so and so. But in trying to figure out what the dependency hypotheses are, that is what the causally independent facts are, you may use backtracking reasoning. That's an epistemic form of reasoning. As you say, you wanna, you don't know whether your partner is in the house yet. And so that's an open question. And the backtracking is a way to get at what the answer to that question is. So quick counterfactuals of any kind may enter into, but the specific kind of causal context and the notion of causal independence is playing a role in all of these, in all of these things. So in fact, you say, well, whether he's in the house or not, it's not causally independent of whether the window is loose, but if it's, the cause goes the other way, in that case, from him being in there to the window. Okay, we only have a couple of minutes, but we have a question from Roger and Berkeley, who says, you refer to the laws of nature as if they are deterministic. Since we discovered that the world appears to operate under the laws of quantum mechanics where all events proceed only probabilistic, i measurements are always approximate. There are many possible outcomes and sequences of events, et cetera. Does this modern physics affect your view of counterfactuals and determinism? Okay, so first, I wanna separate the question what a law of nature is or how it's modeled and defining a dynamic theory from the question whether the laws are deterministic. So you can have laws of, I mean, what's essential to the laws of nature is that they constrain in some way, or they have made claims about the relationship between the state of the world at one time and the state of the world at other times, but there are non-deterministic laws of nature which do constrain, but they constrain it in a non-deterministic way by saying that the outcome have to be one of these and it may assign, the laws may assign a probability value to those things. So even if chance is a feature of the most basic laws of nature, as Louis thought and other people think, those questions about exactly how quantum mechanics should be spelled out, I'm sure, I know, I don't know much about it, but the idea of a deterministic law is just a special kind of law, but not part of the definition of it. I hope that gets at what he's asking. I think so, but I'll watch out for anything else coming in on this. Just, we'll have to stop, I think, in just a second. On your, you said you had some reservations about the thesis of metaphysical determinism, very strong thesis. I guess I wondered about pluralism. I mean, if you think about Nancy Cartwright's picture, for example, of there being no one physical theory that governs all the world, but just many overlapping theories, would you be sympathetic to those kinds of doubts about? Oh yeah, no, and more than that, I mean, I think one can separate the question, what's fundamental science like from the question? I mean, there is a question, does that determine all the other? I mean, the questions that mix normative and factual in a way that's hard to sort out, there's fagness and so on. And so the very claim that all the facts are supervenient on the fact stated in one fundamental scientific theory is it can be called into question both on the grounds that maybe there isn't just one, maybe the left side scientific theories that overlap or that govern different domains and so on and there's no way of unifying them all together, but to add to that, the rest of our discourse, the rest of the things we talk about and we talk about the world and how much of that is not reducible in a piece by piece way, but it is in some way wholly determined by that and that's what I'm skeptical about. Okay, that's wonderful. Well, thank you very much. It's frustrating, but I think we do have to stop here and we'll look forward to carrying on discussion with you tomorrow in the department colloquium. Very good, thank you. So meantime, on behalf of what I'm sure is a vast scattered audience, let me thank you very much for that. Thanks, thank you. Thank you all. I stand in for everyone. Okay, thanks very much. Thank you.