Keep ‘hoping’ for rationality: a solution to the backward induction paradox

We formalise a notion of dynamic rationality in terms of a logic of conditional beliefs on (doxastic) plausibility models. Similarly to other epistemic statements (e.g. negations of Moore sentences and of Muddy Children announcements), dynamic rationality changes its meaning after every act of learning, and it may become true after players learn it is false. Applying this to extensive games, we “simulate” the play of a game as a succession of dynamic updates of the original plausibility model: the epistemic situation when a given node is reached can be thought of as the result of a joint act of learning (via public announcements) that the node is reached. We then use the notion of “stable belief”, i.e. belief that is preserved during the play of the game, in order to give an epistemic condition for backward induction: rationality and common knowledge of stable belief in rationality. This condition is weaker than Aumann’s and compatible with the implicit assumptions (the “epistemic openness of the future”) underlying Stalnaker’s criticism of Aumann’s proof. The “dynamic” nature of our concept of rationality explains why our condition avoids the apparent circularity of the “backward induction paradox”: it is consistent to (continue to) believe in a player’s rationality after updating with his irrationality.     

Bridging learning theory and dynamic epistemic logic

This paper discusses the possibility of modelling inductive inference (Gold 1967) in dynamic epistemic logic (see e.g. van Ditmarsch et al. 2007). The general purpose is to propose a semantic basis for designing a modal logic for learning in the limit. First, we analyze a variety of epistemological notions involved in identification in the limit and match it with traditional epistemic and doxastic logic approaches. Then, we provide a comparison of learning by erasing (Lange et al. 1996) and iterated epistemic update (Baltag and Moss 2004) as analyzed in dynamic epistemic logic. We show that finite identification can be modelled in dynamic epistemic logic, and that the elimination process of learning by erasing can be seen as iterated belief-revision modelled in dynamic doxastic logic. Finally, we propose viewing hypothesis spaces as temporal frames and discuss possible advantages of that perspective.     

Logical information and epistemic space

Gaining information can be modelled as a narrowing of epistemic space. Intuitively, becoming informed that such-and-such is the case rules out certain scenarios or would-be possibilities. Chalmers’s account of epistemic space treats it as a space of a priori possibility and so has trouble in dealing with the information which we intuitively feel can be gained from logical inference. I propose a more inclusive notion of epistemic space, based on Priest’s notion of open worlds yet which contains only those epistemic scenarios which are not obviously impossible. Whether something is obvious is not always a determinate matter and so the resulting picture is of an epistemic space with fuzzy boundaries.     

Contrastivism and Lucky Questions

There’s something deeply right in the idea that knowledge requires an ability to discriminate truth from falsity. Failing to incorporate some version of the discrimination requirement into one’s epistemology generates cases of putative knowledge that are at best problematic. On the other hand, many theories that include a discrimination requirement thereby appear to entail violations of closure. This prima facie tension is resolved nicely in Jonathan Schaffer’s contrastivism, which I describe herein. The contrastivist take on relevant alternatives is implausible, however, and this then threatens to undermine contrastivism’s anti-skeptical results.

Consciousness as a concrete physical phenomenon

The typical empirical approach to studying consciousness holds that we can only observe the neural correlates of experiences, not the experiences themselves. In this paper we argue, in contrast, that experiences are concrete physical phenomena that can causally interact with other phenomena, including observers. Hence, experiences can be observed and scientifically modelled. We propose that the epistemic gap between an experience and a scientific model of its neural mechanisms stems from the fact that the model is merely a theoretical construct based on observations, and distinct from the concrete phenomenon it models, namely the experience itself. In this sense, there is a gap between any natural phenomenon and its scientific model. On this approach, a neuroscientific theory of the constitutive mechanisms of an experience is literally a model of the subjective experience itself. We argue that this metatheoretical framework provides a solid basis for the empirical study of consciousness.     

Internal reasons and practical limits on rational deliberation

I argue that agent-causal libertarianism has a strong initial rejoinder to Mele's luck argument against it, but that his claim that it has yet to be explained how agent-causation yields responsibility-conferring control has significant force. I suggest an avenue of response. Subsequently, I raise objections to Mele's criticisms of my four-case manipulation argument against compatibilism.     

Preschoolers trust novel members of accurate speakers' groups and judge them favourably

By age 3, children track a speaker's record of past accuracy and use it as a cue to current reliability. Two experiments (N?=?95 children) explored whether preschoolers' judgements about, and trust in, the accuracy of a previously reliable informant extend to other members of the informant's group. In Experiment 1, both 3- and 4-year-olds consistently judged an animated character who was associated with a previously accurate speaker more likely to be correct than a character associated with a previously inaccurate speaker, despite possessing no information about these characters' individual records of reliability. They continued to show this preference one week later. Experiment 2 presented 4- and 5-year-olds with a related task using videos of human actors. Both showed preferences for members of previously accurate speakers' groups on a common measure of epistemic trust. This result suggests that by at least age 4, children's trust in speaker testimony spreads to members of a previously accurate speaker's group.     

Implicit bias,confabulation, and epistemic innocence

In this paper I explore the nature of confabulatory explanations of action guided by implicit bias. I claim that such explanations can have significant epistemic benefits in spite of their obvious epistemic costs, and that such benefits are not otherwise obtainable by the subject at the time at which the explanation is offered. I start by outlining the kinds of cases I have in mind, before characterising the phenomenon of confabulation by focusing on a few common features. Then I introduce the notion of epistemic innocence to capture the epistemic status of those cognitions which have both obvious epistemic faults and some significant epistemic benefit. A cognition is epistemically innocent if it delivers some epistemic benefit to the subject which would not be attainable otherwise because alternative (less epistemically faulty) cognitions that could deliver the same benefit are unavailable to the subject at that time. I ask whether confabulatory explanations of actions guided by implicit bias have epistemic benefits and whether there are genuine alternatives to forming a confabulatory explanation in the circumstances in which subjects confabulate. On the basis of my analysis of confabulatory explanations of actions guided by implicit bias, I argue that such explanations have the potential for epistemic innocence. I conclude that epistemic evaluation of confabulatory explanations of action guided by implicit bias ought to tell a richer story, one which takes into account the context in which the explanation occurs.