We argue that atomistic learning?learning that requires training only on a novel item to be learned?is problematic for networks in which every weight is available for change in every learning situation. This is potentially significant because atomistic learning appears to be commonplace in humans and most non-human animals. We briefly review various proposed fixes, concluding that the most promising strategy to date involves training on pseudo-patterns along with novel items, a form of learning that is not strictly atomistic, but which (...) looks very much like it ?from the outside? (shrink)

What entitles you to rely on information received from others? What entitles you to rely on information retrieved from your own memory? Intuitively, you are entitled simply to trust yourself, while you should monitor others for signs of untrustworthiness. This article makes a case for inverting the intuitive view, arguing that metacognitive monitoring of oneself is fundamental to the reliability of memory, while monitoring of others does not play a significant role in ensuring the reliability of testimony.

Real agents rely, when forming their beliefs, on imperfect informational sources (sources which deliver, even under normal conditions of operation, both accurate and inaccurate information). They therefore face the ‘endorsement problem’: how can beliefs produced by endorsing information received from imperfect sources be formed in an epistemically acceptable manner? Focussing on the case of episodic memory and drawing on empirical work on metamemory, this article argues that metacognition likely plays a crucial role in explaining how agents solve the endorsement problem.

The central supposition of the sceptical controversy regarding rationality in the theory of argumentation is that either there are universal standards against which the reasonableness of arguments can be evaluate or, conversely, that there are no determinate standards against which arguments can be evaluated, and hence no methods by which disputes can be rationally resolved. The paper argues that the basic terms of this debate are erroneously defined and that there is a middle path in this sceptical controversy. The paper (...) adopts the later Wittgensteinian approach to language and applies it to the concept of rationality. Furthermore, it maintains that the search for universal standards of reasonableness in theories of argumentation is likely to come to naught, and that, nonetheless, there are discoverable methods by which arguments are evaluated, facts constituted, and disputes resolved. (shrink)