François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2023-2024

Colloque - Indexical Dynamics : Now and Then: The Dynamics of Self-Locating Beliefs

Colloque organisé par François Recanati, Professeur du Collège de France, chaire Philosophie du langage et de l'esprit

Intervenant(s)

Matheus Valente, University of Barcelona (LOGOS) & University of Valencia

It's often said within epistemology circles that self-locating beliefs about now and then change in peculiar ways incompatible with traditional Bayesian update rules, and so, that these beliefs are epistemically exceptional. The point is clear enough when we consider subjects who lose track of time - e.g. Rip van Winkle (Kaplan 1989) - and even clearer when some funny business with their memories are added to the equation, as in the famous story of Sleeping Beauty (Elga 2000). But that's like killing a fly with a bulldozer, after all, the dynamics of self-locating beliefs seems exceptional even when attention is restricted to idealised agents that are assumed to never forget any information nor to lose track of time. Such is the case of Chronos, an omniscient god in a deterministic world who not only knows the complete history of her universe but is never uncertain about what time it is. Since Chronos is de dicto omniscient, her beliefs are always entirely concentrated on one possible world. Since she's self-locating omniscient, her self-locating beliefs are, at any given time, concentrated on a single temporal location. Though Chronos is never uncertain about anything, it appears that she must be constantly shifting her self-locating beliefs just to keep up with time: when the present time is n, she believes 'now is n', a moment later when it's n+1, she instead believes 'now is n+1' etc. To use Evans' (1982) metaphor, self-locating beliefs seem to require us to run to keep still. Call the peculiar type of dynamics that even omniscient gods must subject their self-locating beliefs to Shifting (Arntzenius 2003; Bradley (2011) calls it 'Belief Mutation'; Recanati (2016) calls it 'conversion').

To account for Shifting we need to explain why Chronos knows at n that she'll believe 'now is n+1' at n+1 but still refrains from presently believing it. In other words, we need to explain why self- locating beliefs violate van Frassen's (1984) Reflection Principle which holds that we ought to defer to our future selves as experts (under the assumption that their epistemic standing is at least as good as our current one). One way to do so favoured by the majority of epistemologist is to countenance tensed propositions whose truth-values change with time (Titelbaum 2013, p. 171-278). Another way is to hold that the passage of time changes what times/events subjects are acquainted with, which then changes which de re beliefs they can hold at each time. The first camp holds that self-locating beliefs are epistemically special because their truth is tensed. The second camp holds that they so are because their accessibility is tensed. Given how easily this argument seems to roll off the tongue, one wonders whether the last decade of debates with sceptics like Cappelen & Dever (2013) and Magidor (2015) would have taken a different shape if more focus had been given to self-locating beliefs involving instead of to de se beliefs involving 'I'.

But there's an issue. Some epistemologists hold that Shifting is a particularly type of sterile belief update: when the only change in a subject's belief across times is due to Shifting, it's never rational for that subject to revise her de dicto beliefs. As Titelbaum (2013, 233) remarks, it's intuitive that "finding oneself passing through the world in exactly the way one was certain one was going to shouldn't change one's opinions about what that world is like". This suggests a different view where Shifting is not taken to be a type of belief change but instead of belief retention, a view which Prosser (2005) calls 'the

Frege-Evans dynamic theory'. On that approach, Chronos never changes any of her beliefs, and the permutations of indexicals which arise in virtue of the passage of time - now 'now', then 'then' - are really just ways of expressing a single persisting dynamic belief. If the Frege-Evans dynamic theory is tenable, we'd lose one important reason to think that self-locating beliefs are epistemically exceptional. This is so because, as I'll argue, the only puzzling cases that they would still give rise to would be ones where subjects' epistemic states become deteriorated across time due to cognitive mishaps like failures of memory or of one's tracking/discriminatory capacities. Since nobody doubts that weird things happen when our cognitive powers deteriorate, the problem of cognitive dynamics turns out to pertain less about self-locating beliefs per se than about the unsurprising fact that cognitively deteriorated subjects are epistemically exceptional.

If the Frege-Evans dynamic view can be upheld would thus have significant implications. Whether this can be done ultimately depends on what we should say about the interplay between self- location and rational action. Consider a temporal variant of Perry's famous bear attack case: Chronos intends to whistle once at all odd times and twice at all even times. Since Chronos acts differently as time passes, it seems that we must conclude that doxastic states coordinated by nothing but Shifting really change. Can the Frege-Evans dynamic view accommodate the fact that a single persisting dynamic belief might lead to distinct actions at distinct times?

My main objective in this talk is to motivate an affirmative answer. My hypothesis is that rational "changes" due only to Shifting are just as questionably a real type of change as corresponding "changes" in self-locating beliefs, and so, that the Frege-Evans dynamic view can be applied to the realm of reasons, intentions, and decision theory, just as well as it can be applied to the realm of belief and confirmation. The outcome is that when a subject's epistemic state changes only by Shifting, both their beliefs and their reasons/intentions can be said to remain stable regardless of their distinct (respective) linguistic and practical manifestations. To cash this out, I first defuse a set of cases which have taken some to think that Shifting can by itself require a revision of de dicto beliefs (Arntzenius' (2003) Prisoner, Elga's (2000) Sleeping Beauty, Shaw's (2019) forgetful god Lethe, and Topey's (forthcoming) hypoxia-affected Aisha), and argue that Shifting isn't the culprit of these cases' peculiar features. Then, I'll draw a parallel between the Frege-Evans dynamic view and recent work in the philosophy of action which holds that the distinction between intentions for the future and intentions for the present is ill-motivated (McDowell 2011, Brozzo 2021). These two philosophical debates bear promising structural analogies that have only recently started to be acknowledged. I'll deploy this analogy to argue that (1) the Frege-Evans dynamic view allows us to demystify the cognitive dynamics of self-locating beliefs, and so, that the epistemic exceptionality of self-locating beliefs must reside elsewhere (if it resides anywhere at all) and (2) that this view's credentials can be defended by showing how it fits with an independently plausible picture of how rational agents act on the basis of intentions formed at times prior to the action's execution.