非表示:
キーワード:
-
要旨:
There have been many battles about how best to formalise the affective states of humans andother animals in ways that can be self-evidently tied to quantifiable behaviours. One recent suggestion isthat positive and negative moods can be treated as prior expectations over the future delivery of rewardsand punishments, and that these priors affect behaviour through the conventional workings of Bayesiandecision theory (Mendl et al., 2010). Amongst other characteristics, this suggestion provides an inferentialfoundation for a task that has become a widely-used method for assessing mood states in animals (Hardinget al., 2004). This so-called ‘cognitive bias’ task extracts information about affect from the optimistic orpessimistic manner in which subjects resolve ambiguities in sensory input. Here, we describe experimentsin humans and rodents aimed at elucidating further aspects of this notion. The human studies assessed theextent to which subjects can incorporate information about explicitly-imposed external loss functions intotheir inference about ambiguous inputs, and the way this incorporation interacts with mood. Subjects foundit hard to integrate these sources of information well, which was unexpected given their apparently admirablecapacities in related circumstances (Whiteley & Sahani, 2008), so we are exploring modifications. Therodent studies sought to examine the interaction between the experimenter-imposed instrumental demandsof the task and inherent Pavlovian effects, such as ineluctable approach and avoidance in the face of theprospect respectively of rewards and punishment (Guitart-Masip et al. 2014). The latter might provide anaccount of the differences between rats and mice that we were surprised to observe.
