Fabrizio Cariani

Contextology is the science of the dynamics of the conversational context. Contextology formulates laws governing how the shared information states of interlocutors evolve in response to assertion. More precisely, the contextologist attempts to construct a function which, when provided with just a conversation’s pre-update context and the content of an assertion, delivers that conversation’s post-update context. Most contextologists have assumed that the function governing the evolution of the context is simple: the post-update context is just the pre-update context intersected with the content of the assertion. We argue that this assumption is wrong: not only is it false, it is also incoherent given standard contextological assumptions. Moreover, it is impossible in principle to revise it to correctly describe the dynamics of context. We conclude that there can be no science of Contextology. The laws governing the evolution of the context in response to assertion must make essential reference to the private information states of interlocutors.

Many AI systems that make important decisions are black boxes: how they function is opaque even to their developers. This is due to their high complexity and to the fact that they are trained rather than programmed. Efforts to alleviate the opacity of black box systems are typically discussed in terms of transparency, interpretability, and explainability. However, there is little agreement about what these key concepts mean, which makes it difficult to adjudicate the success or promise of opacity alleviation methods. I argue for a unified account of these key concepts that treats the concept of understanding as fundamental. This allows resources from the philosophy of science and the epistemology of understanding to help guide opacity alleviation efforts. A first significant benefit of this understanding account is that it defuses one of the primary, in-principle objections to post hoc explainable AI (XAI) methods. This “rationalization objection” argues that XAI methods provide mere rationalizations rather than genuine explanations. This is because XAI methods involve using a separate “explanation” system to approximate the original black box system. These explanation systems function in a completely different way than the original system, yet XAI methods make inferences about the original system based on the behavior of the explanation system. I argue that, if we conceive of XAI methods as idealized scientific models, this rationalization worry is dissolved. Idealized scientific models misrepresent their target phenomena, yet are capable of providing significant and genuine understanding of their targets.

The way we represent the world in thought and language is shot through with indeterminacy: we speak of red apples and yellow apples without thereby committing to any sharp cutoff between the application of the predicate ‘red’ and of the predicate ‘yellow’. But can reality itself be indeterminate? In other words, can indeterminacy originate in the mind-independent world, and not only in our representations? If so, can the phenomenon also arise at the microscopic scale of fundamental physics? Section 1 of this Element provides a brief overview of the question of indeterminacy. Section 2 discusses the thesis that the world is comprised of indeterminate objects, whereas Section 3 focuses on the thesis that there are indeterminate states of affairs. Finally, Section 4 is devoted to the case study of indeterminacy in quantum physics.

This chapter was written in 2013 and was posted in the Semantics Archive in January 2014. The preprint of the published version has been in the Semantics Archive since 2016. The Semantics Archive is an electronic preprint archive hosted by the Linguistics Society of America. The chapter looks at indicative conditionals embedded under quantifiers, with a special emphasis on ‘one-case’, episodic, conditionals as in "No query was answered if it came from a doubtful address." It agrees with earlier assessments that a complete conditional (with antecedent and consequent) is embedded under a quantifier in those constructions, but then proceeds to create a dilemma by showing that we can’t always find the right interpretation for that conditional. Contrary to earlier assessments, Stalnaker’s conditional won’t always do. The paper concludes that the embedded conditional in the sentence above is a material implication, but the "if"-clause also plays a pragmatic role in restricting the domain of the embedding quantifier. That an appeal to pragmatics should be necessary at all goes with Edgington’s verdict that “we do not have a satisfactory general account of sentences with conditional constituents.” One of the conclusions drawn at the end of the chapter is that the English modal "will" should be given a Stalnaker selection function analysis. The idea is implemented by attaching to "will" an accessibility function variable that receives its value from the context of use and can be dynamically updated by preceding "if"-clauses, as proposed in von Fintel (1994). The argument for a Stalnaker selection function analysis of "will" provided by the paper is that, together with the restrictor analysis of "if"-clauses, it yields the correct truth-conditions for "will"-conditionals (as opposed to episodic, 'one case', conditionals) embedded under quantifiers. The proposal that the English modal "will" should be given a Stalnaker selection function analysis has since been taken up and elaborated considerably in Cariani & Santorio's (2018) paper on "will" in Mind, and is a central idea in Cariani's (2021) book The Modal Future.

This paper is about the principle that success entails ability, which I call Success. I argue the status of Success is highly puzzling: when we focus on past instances of actually successful action, Success is very compelling; but it is in tension with the idea that true ability claims require an action be in the agent's control. I make the above tension precise by considering the logic of ability. I argue Success is appealing because it is classically equivalent to two genuinely valid inferences, which I call Past Success and Can't-Entails-Won't; but also that Success itself has counterexamples. I show how to invalidate Success while validating Past Success and Can't-Entails-Won't by connecting the meaning of ‘can’ to facts about what is settled or open. I define an operator W with features attributed to ‘will’ in the literature on future contingents. I then give a conditional analysis of ability ascriptions stated with W-conditionals, where "S can A" says, roughly, there’s some action available to S such that if S does it, then W(S A). I show this semantics invalidates Success while still explaining its appeal.

This paper develops an information-sensitive theory of the semantics and probability of conditionals and statements involving epistemic modals. The theory validates a number of principles linking probability and modality, including the principle that the probability of a conditional If A, then C equals the probability of C, updated with A. The theory avoids so-called triviality results, which are standardly taken to show that principles of this sort cannot be validated. To achieve this, we deny that rational agents update their credences via conditionalization. We offer a new rule of update, Hyperconditionalization, which agrees with Conditionalization whenever nonmodal statements are at stake but differs for modal and conditional sentences.

This paper investigates the link between informational consequence and credence. I first suggest a natural constraint, namely that informational consequence should preserve certainty: on any rational credence distribution, when the premises of an informational inferences have credence 1, the conclusion also has credence 1. Then I show that the certainty‐preserving constraint leads to triviality. In particular, the following three claims are incompatible: (i) informational consequence is extensionally distinct from classical consequence; (ii) informational inferences preserve certainty; (iii) credences obey (a subset of) classical Bayesian constraints. The proof is straightforward, but the theoretical implications are substantial. Informational theorists need to either give up the idea that credence applies to epistemic discourse, or develop a nonclassical theory of credence and credal update. Moreover, the result also suggests that there is a connection between informational consequence and triviality results, including classical triviality results like Lewis's.

The Identity principle says that conditionals with the form 'If p, then p' are logical truths. Identity is overwhelmingly plausible, and has rarely been explicitly challenged. But a wide range of conditionals nonetheless invalidate it. I explain the problem, and argue that the culprit is the principle known as Import-Export, which we must thus reject. I then explore how we can reject Import-Export in a way that still makes sense of the intuitions that support it, arguing that the differences between indicative and subjunctive conditionals play a key role in solving this puzzle.

Since it was presented in 1963, Chisholm’s paradox has attracted constant attention in the deontic logic literature, but without the emergence of any definitive solution. We claim this is due to its having no single solution. The paradox actually presents many challenges to the formalization of deontic statements, including (1) context sensitivity of unconditional oughts, (2) formalizing conditional oughts, and (3) distinguishing generic from nongeneric oughts. Using the practical interpretation of ‘ought’ as a guideline, we propose a linguistically motivated logical solution to each of these problems, and explain the relation of the solution to the problem of contrary-to-duty obligations.

This paper concerns a puzzle about the interaction of epistemic modals and future tense. In cases of predictable forgetfulness, speakers cannot describe their future states of mind with epistemic modals under future tense, but promising theories of epistemic modals do not predict this. In §1, I outline the puzzle. In §2, I argue that it undermines a very general approach to epistemic modals that draws a tight connection between epistemic modality and evidence. In §3, I defend the assumption that tense can indeed scope over epistemic modals. In §4, I outline a new way of determining the domain of quantification of epistemic modals: epistemic modals quantify over the worlds compatible with the information accumulated within a certain interval. Information loss can change which interval is relevant for determining the domain. In §5, I defend the view from some objections. In §6, I explore the connections between my view of epistemic modality and circumstantial modality.

Assume that it is your evidence that determines what opinions you should have. I argue that since you should take peer disagreement seriously, evidence must have two features. (1) It must sometimes warrant being modest: uncertain what your evidence warrants, and (thus) uncertain whether you’re rational. (2) But it must always warrant being guided: disposed to treat your evidence as a guide. Surprisingly, it is very difficult to vindicate both (1) and (2). But diagnosing why this is so leads to a proposal—Trust—that is weak enough to allow modesty but strong enough to yield many guiding features. In fact, I claim that Trust is the Goldilocks principle—for it is necessary and sufficient to vindicate the claim that you should always prefer to use free evidence. Upshot: Trust lays the foundations for a theory of disagreement and, more generally, an epistemology that permits self-doubt—a modest epistemology.