Abductive reasoning is the process of finding a best explanation for a
set of observations. In many abductive problems, like medical
diagnosis, scientific discovery, debugging or troubleshooting, an
amount of information far beyond the capacity limits of working memory
(WM) must be processed. Although WM plays a central role in theories
of human cognition, theories of abductive reasoning do not specify WM
processes during the generation of explanations. On the basis of a
computational model of abductive reasoning and of theories of text
comprehension a mechanism is proposed that reduces WM load during
abductive reasoning. The computational model views abductive reasoning
as the sequential comprehension and integration of observations into a
situation model that represents the current best explanation for the
observations. The proposed WM mechanism assumes that the situation
model is only partly kept in WM, whereas other pieces are stored in
long-term memory. These long-term representation part can be reliably
accessed through retrieval structures to reinstatiate information in
WM during abductive reasoning. It is assumed that unexplained
observations are actively maintained in WM until an explanation for
them could be generated. Thereafter their representation is lost from
WM. But these explained observations can be recalled from long-term
memory via their integration into the situation model.
This mechanism makes predictions about the availability of the mental
representation of explained and unexplained observations. These
predictions were tested in four experiments, using different memory
tests for observations. In Experiments 1 and 2 a recognition test was
used, in Experiment 3 an implicit menory test was used and in
Experiment 4 the participants had to perform an unexpected recall
after task interruption.
The results show that unexplained observations are accessed faster
than explained ones during abductive reasoning. This confirms the
mechanism's assumption that unexplained observations are kept in WM and
explained ones not. But explained observations seem not to be
represented in long-term memory. Rather, it seems that observations
are rapidly forgotten afer they are explained. Different possible
reasons for this pattern of result are discussed.
| Identifer | oai:union.ndltd.org:DRESDEN/oai:qucosa:de:qucosa:17738 |
| Date | 08 February 2001 |
| Creators | Baumann, Martin |
| Contributors | Technische Universität Chemnitz |
| Source Sets | Hochschulschriftenserver (HSSS) der SLUB Dresden |
| Language | German |
| Detected Language | English |
| Type | doc-type:doctoralThesis, info:eu-repo/semantics/doctoralThesis, doc-type:Text |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.1906 seconds