Confirmed speakers so far:
A Kratzer-style semantics for indicative conditionals invokes a modal base and a selection function. It is argued that in any given context of use these parameters are fixed by speaker- and audience-independent features of the world: they depend only on what is *objectively* possible and on the causal dependencies of these objective possibilities. This implies that when dealing with determinate matters (e.g. decidable claims concerning the present and past) the proposition expressed by an indicative conditional is true in a context of use c iff the corresponding material conditional is true in c. For epistemic might-claims we get, in a similar fashion, the consequence that Might A is true in c iff A is true in c. It is argued that the added structure provided by a Kratzer-style semantics is primarily of interest in characterising the propositional content of epistemic states and the entailment relations between propositions. The content of an epistemic state can be represented as a set of *centered* worlds -- worlds combined with an epistemic perspective within that world (a modal base, a selection function) along the lines of the centered worlds invoked in order to explain self-locating (de se) knowledge. It is argued that the above model provides the best way to understand disagreement data. Two objections are considered: the objection that it violates the knowledge norm of assertion, and the objection that a semantically self-conscious speaker will not be able to coherently believe the proposed theory. It is argued that neither objection is valid.
This paper connects triviality, in the sense of Lewis (1976), to model-theoretic constraints on accessibility relations for modals in natural language, like Euclideanness. The aim is to expand the theoretical pressures of triviality beyond their usual boundaries, while also undermining certain standard ways of ameliorating such pressures.
I begin by discussing the best-known approach to using causal models to interpret subjunctive conditionals---interventionism---and I discuss Rachel Briggs' observation that this approach respects an intuitive counterexample to Modus Ponens. I then offer a class of counterexamples to these interventionist theories and offer some consideration of how my counterexamples should inform our thinking about Modus Ponens and backtracking interpretations of conditionals.
I have long argued for a kind of ‘counterfactual skepticism’: most counterfactuals are false. Very briefly, I maintain that the indeterminism and indeterminacy associated with most counterfactuals entail their falsehood. For example, I claim that these counterfactuals are both false:
(Indeterminism) If the chancy coin were tossed, it would land heads.
(Indeterminacy) If I had a son, he would have an even number of hairs on his head at his birth.
And I argue that most counterfactuals are relevantly similar to one or both of these, as far as their truth-values go.
However, ordinary speakers judge many counterfactuals that they utter to be true. And a number of philosophers have defended such judgments against counterfactual scepticism. David Lewis appeals to ‘quasi-miracles’; Robbie Williams to ‘typicality’; Sarah Moss to pragmatics; Moritz Schulz to an epsilon-operator semantics; Karen Lewis to contextually-sensitive ‘relevance’; and Orri Stefánsson to ‘counterfacts’. I argue against each of these proposals. Most counterfactuals are still false.
Kratzer's now orthodox semantics for modals famously embeds poorly in consequents. It does not assign intuitive truth conditions to sentences like "If the kids are going to drive drunk, then they should drive drunk" and "If you want sugar in your soup, you should see a waiter". In work presented at the 2nd Belgrade Conference on Conditionals, I argued that we should regard certain variables in Kratzer's semantics as determined by the context not the index. In this paper, I argue that so construing Kratzer's semantics also solves her problems with conditionals and refine my general proposal by appealing to recent work in default logic, especially that of John Horty.
The truth-conditions of counterfactuals vary across contexts of use. However, I argue that there is a set of (broadly speaking) causal principles that are central to their truth-conditions across a wide range of contexts. Moreover, I outline a functional explanation of the fact that we use these principles in counterfactual reasoning.
Test semantics for indicative conditionals (e.g., Gillies 2004; Yalcin 2012) are known to solve a range of problems for the interaction of conditionals with other operators, such as probability operators, epistemic modals, and attitude verbs. They also face well-known problems. In this talk we lay out the strengths and weaknesses of test semantics, and evaluate several strategies for remedying the shortcomings. We argue that ranking theory provides the best framework for addressing an important range of shortcomings.
The problem of reverse Sobel sequences is taken by some to be an important objection to the classic Lewis-Stalnaker semantics for counterfactuals. Responses to the problem have been wide-ranging. Some (von Fintel, Gillies) have argued that the Lewis-Stalnaker semantics should be rejected, and a version of a strict conditional semantics, which better handles the troublesome sequences, should be endorsed in its place. Others (Karen Lewis, Ichikawa) have argued that the problem motivates a contextualist rendering of counterfactuals similar to contextualist accounts of knowledge or taste. And Moss has argued that there is a plausible, entirely pragmatic way to account for the (apparently) problematic sequences.
It is my contention that none of these responses to the problem is right. After showing why I think each extant solution is inadequate, I defend a novel way to make sense of the troublesome sequences. The solution I endorse avoids the problems faced by the alternative analyses. In addition, there is good independent reason to think that it is right. There is, however, a difficulty for my view: its truth suggests that many ordinarily accepted counterfactuals are not true. I argue that this (apparent) cost is an acceptable one.
Suppose we're at a party, and everyone is under-dressed. I say:
MIJON: If Mijon were here he'd be wearing a bowtie.
Then suppose we found out that Mijon is here, in the very next room. How would we evaluate the truth value of MIJON? We'd probably think that it's true if Mijon is wearing a bowtie, and false if he isn't.
The attractiveness of this inference pattern has led many if not most philosophers writing on counterfactual conditionals to assume the truth of something like the following, which I call the True Antecedent Principle (TAP):
TAP: If A is true, then A>C is true iff C is true.
I argue that TAP is invalid. There are a few putative counterexamples from the relevant literature; I offer a few more of my own that I take to be more forceful. If you share my intuitions about these cases, you'll be inclined to reject TAP.
Unfortunately, TAP is difficult to resist on the standard Lewisian account (SLA) of counterfactuals:
SLA: A>C is true iff (roughly) C is true at the closest A-worlds.
Part of the Lewisian package is also a principle called Strong Centering (SC):
SC: Every world is closer to itself than any other world.
Together SLA and SC straightforwardly entail TAP. If you want to reject TAP, this leaves two options: reject SLA, or reject SC. I argue we should reject SLA. But then of course we need a new theory of counterfactuals to replace it.
I happen to have an alternative, independently motivated account of counterfactuals that may be well suited to solve cases like these. On my view it is not always just the very closest A-worlds that are relevant, but often some contextually determined wider variety of A-worlds instead. Cases like these allow for violations of TAP, but the account can also accommodate the instances of its apparent validity. This constitutes an additional, separate motivation for my theory over the standard one.
One of the motivations given by Anderson and Belnap for their logics E and R was the Use Criterion, which said that in the derivation of a conditional, the antecedent had to be used to obtain the consequent. The Use Criterion largely dropped out of subsequent work on relevant logics. I will present an interpretation of proofs in terms of informational flow that, I think, captures a core idea of the Use Criterion. I briefly explain how this can be used to respond to some objections to the Use Criterion.