Tuesday, 16 April 2019

Confabulation as Unreliable Imagining

This is the second in a series of posts featuring new research on confabulation. Today's contribution is by Kirk Michaelian (Centre for Philosophy of Memory) who summarises his paper, "Confabulation as Unreliable Imagining", for the special issue of Topoi on confabulation guest edited by Sophie Stammers and Lisa Bortolotti.




The context for my contribution to the special issue is a debate over the nature of confabulation that has been unfolding for several years now within the philosophy of memory community.

In my 2016 book, Mental Time Travel: Episodic Memory and Our Knowledge of the Personal Past, I developed and defended a simulation theory of memory. In opposition to the causal theory, the simulation theory denies that remembering an event presupposes the existence of an "appropriate" causal connection between the subject's present representation of the event and his past experience of it, maintaining, instead, that the difference between genuine and merely apparent remembering is a matter of reliability: genuine remembering is carried out by a properly functioning -- and hence reliable -- episodic construction system.

The simulation theory naturally suggests an account of confabulation according to which confabulation is distinguished from remembering by its unreliability: in remembering, the process that produces the apparent memory is reliable; in confabulation, it is not. While the book briefly sketches this account of confabulation, however, it doesn't develop it in any detail, and I hadn't thought carefully about confabulation until I was prompted to do so by a pair of stimulating articles by Sarah Robins on confabulation and misremembering.

Opposing the simulationist account sketched in the book, Robins develops a causalist account according to which confabulation is distinguished from successful remembering by its falsity and by the absence of appropriate causal connection, whereas misremembering -- seen, for example, in the DRM effect, in which subjects who study a list of thematically-related words tend to "remember" non-presented but thematically-consistent lure words -- is distinguished from successful remembering by its falsity and by the presence of appropriate causal connection.

Unconvinced by Robins' argument, I wrote an article devoted to critiquing the causalist account, in part on the ground that it failed to acknowledge the possibility of veridical confabulation, and to refining the simulationist account. According to the refined version of the simulationist account developed in this article, falsidical confabulation is characterized by falsity and unreliability, veridical confabulation is characterized by truth and unreliability, and misremembering is characterized by falsity and reliability; successful remembering, in line with the simulation theory, is characterized by truth and reliability.

This refined simulationist account was then attacked in an article by Bernecker, who argued that it could accommodate neither the possibility of unjustified memory nor that of justified confabulation.

My contribution to the special issue argues that, unlike the epistemic accounts with which Bernecker groups it, the simulationist account can indeed accommodate unjustified memory and that, since justified confabulation is not a genuine possibility, it need not accommodate justified confabulation.

It also further refines the simulationist account by taking into account the role of failures of metacognitive monitoring in unsuccessful remembering. Earlier versions of the simulationist account emphasized that reliability and accuracy can come apart at the level of apparent remembering; this happens in misremembering (in which a reliable process produces an inaccurate representation) and in veridical confabulation (in which an unreliable process produces an accurate representation). The current version of the simulationist account points out that this can happen at the metacognitive level as well. A form of luck thus plays a role at two distinct levels in the production of memory errors.

The result is a more complex but (I believe) more precise taxonomy of memory errors. In order to tame the complexity of the taxonomy, the article distinguishes among four groups of subjects -- those with no malfunction, those with malfunction at the level of remembering only, those with malfunction at the level of metacognition only, and those with malfunction at both levels -- and sorts the errors characteristic of each group according to whether they involve no luck, luck at one or the other level, or luck at both levels.

Table: A new simulationist classification, second attempt; alternative presentation


The obvious next step is to consider which of the errors acknowledged by the account have been or should be subjected to empirical investigation. And there will -- the conciliatory tone of Robins' contribution to the special issue notwithstanding -- no doubt be further attacks from the causalist camp. So, while the current version of the simulationist account is an improvement over previous versions, more work remains to be done.

No comments:

Post a Comment

Comments are moderated.