Memory, belief and time

Abstract

I argue that what evidence an agent has does not supervene on how she currently is. Agents do not always have to infer what the past was like from how things currently seem; sometimes the facts about the past are retained pieces of evidence that can be the start of reasoning. The main argument is a variant on Frank Arntzenius’s Shangri La example, an example that is often used to motivate the thought that evidence does supervene on current features.

Keywords:

• Memory
• knowledge
• belief
• rationality
• time

Acknowledgements

Thanks to Maria Lasonen-Aarnio, Ishani Maitra, and the participants at the excellent Belief, Action, and Rationality over Time conference for helpful comments on this paper. Special thanks to Jonathan Weisberg, whose comments on the version of the paper presented at Madison saved me from a number of mistakes.

Notes

1. As with the transmissive view on testimony, I don’t take it to be essential to the transmissive view that all mnemonic knowledge is transmitted. Perhaps, as Lackey (2005) argues, memory can sometimes generate new knowledge. Even so, as long as it sometimes plays a purely preservative role, the transmissive theory is true.

2. Dummett (1994) also defends a transmissive account of memory, though the analogy between testimony and memory is important in his argument. Dokic (2001) endorses Dummett’s position on memory.

3. On children’s capacities to learn, see Saffran, Aslin, and Newport (1996a, 1996b) and Gopnik et al. (2001). For applications of this directly to the judgements of credibility, see among many others, Koenig, Clement, and Harris (2004) and Harris and Corriveau (2011). Jaswal, McKercher, and VanderBorght (2008) show that children don’t just track credibility of informants, they trade off credibility of informant against credibility of what is currently being said. In general, the lesson from the last 10–20 years of research is that children have more than enough capacity to perform the cognitive tasks that indirect theorists require of them.

4. I haven’t actually defended this in print yet, but it is correctly attributed to me by Horowitz (2013, 25).

5. Hamilton (2001, 78) says that this idea, that it is wrong to open an inquiry you know you can’t complete, plays a central role in the epistemology of the important Nyāya philosopher Gotama. The discussion of forecasting in Tetlock and Gardner (2015) might cast doubt on whether the kind of conservatism I’m endorsing here is empirically sound. There is a suggestion there that people who tinker with their credal states more frequently end up with more accurate credences. This is a topic that deserves revisiting as more data comes in.