FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2768395108> ?p ?o ?g. }

• W2768395108 abstract "A Bayesian Account of Reconstructive Memory Pernille Hemmer (phemmer@uci.edu) Department of Cognitive Sciences, University of California, Irvine Irvine, CA, 92697-5100 Mark Steyvers (msteyver@uci.edu) Department of Cognitive Sciences, University of California, Irvine Irvine, CA, 92697-5100 human agents. They showed that Bayesian and human learners revert to their prior when inferring the underlying function of a set of coordinates. While serial reproduction is about the evolution from iteration to iteration, the approach presented here will focus the retrieval from memory based on a single specific event. Previous work by Huttenlocher and colleagues (Crawford, Huttenlocher, & Engebretson, 2000; Huttenlocher, Hedges, & Duncan, 1991; Huttenlocher, Hedges, & Vevea, 2000) has shown that that prior knowledge exerts strong influences on reconstruction from memory. Huttenlocher, Hedges, & Duncan (1991) presented a Bayesian model of category effects positing that reconstruction from memory is a weighted average of specific memory traces and category information. This weighted average ‘cleans up’ noisy memory traces and prevents large errors in reconstruction. In this paper, we first present the basic approach of the model presented by Huttenlocher and colleagues and then introduce a series of extensions to this model. We assume that the observer is presented with an object during study and is instructed to retrieve from memory a feature of that object at a later time. In the experiment reported in this paper, we test memory for one-dimensional stimulus values, such as the size of an object. In this context, the goal for the observer is to reconstruct the original size μ of an object using noisy samples y that are retrieved from memory. Bayes’ rule gives us a principled way of combining prior knowledge and evidence from memory: Abstract It is well established that prior knowledge influences reconstruction from memory, but the specific interactions of memory and knowledge are unclear. Extending work by Huttenlocher et al. (1991, 2000) we propose a hierarchical Bayesian model of reconstructive memory in which prior knowledge interacts with episodic memory at multiple levels of abstraction. The combination of prior knowledge and noisy memory representations is dependent on familiarity. We present empirical evidence of the hierarchical influences of prior knowledge, showing that the reconstruction of familiar objects is influenced toward the specific prior for that object, while unfamiliar objects are influenced toward the overall category. Keywords: Long term memory; Prior knowledge; Bayesian models; Reconstructive memory Introduction Knowledge is essential for our interactions with the environment. We learn more easily by using what we know to relate to new information and associations for objects are learned over a lifetime. The challenge, however, is to understand how this knowledge interacts with memory. Bartlett (1932) showed that memories are guided by schemas that help to fill in the details of memories. For example, providing labels can activate schemas that guide the interpretation of the stimulus and serves as an aid to memory. Carmichael, Hogan, & Walter (1932) showed that providing labels can facilitate and influence later reconstruction. They had subjects study a simple line drawing, e.g., two circles and a line (o-o), along with a label. Subjects who were given the label ‘eyeglasses’ later tended to reconstruct the drawing with a curve rather than a line connecting the circles (o^o), representing the nosepiece on a pair of glasses. Subjects who were given the label ‘dumbbell’ tended to reconstruct a thicker line (o=o) similar to the handle on a dumbbell. Biases need not be from labels provided by the experimenter, but may arise from internal sources as well. Bartlett showed that the participants themselves bring certain biases to the task. In both temporal and serial reproduction he demonstrated how a person’s cultural and social experiences influence their reconstruction to conform to their idiosyncratic biases. Kalish, Griffiths, and Lewandowsky (2007) formalized Bartlett’s serial reproduction task using iterated learning with Bayesian and p ( μ | y ) ∝ p ( y | μ ) p ( μ ) The posterior probability p( μ | y) gives the likely stimulus values μ given the noisy memory contents y. This posterior probability is based on a combination of p(μ), the prior knowledge of the likely sizes of the object and p(y|μ), the likelihood of obtaining evidence y from memory. This Bayesian approach gives a principled account of how prior knowledge of the world is combined with memory contents to recall information about events. For example, suppose the feature values of objects are Gaussian distributed, μ ~ N( μ o , σ o2 ), where μ o and σ o2 are the prior mean and variance of the feature values. Furthermore, when a specific object value μ s is studied, suppose this leads to samples y drawn from episodic memory with the samples having a Gaussian noise distribution centered on the original studied value, y ~ N( μ s , σ m2 ). The variance of the noise process, σ m2 , controls the degree to which the" @default.
• W2768395108 created "2017-12-04" @default.
• W2768395108 creator A5051768325 @default.
• W2768395108 creator A5082806282 @default.
• W2768395108 date "2008-01-01" @default.
• W2768395108 modified "2023-09-26" @default.
• W2768395108 title "Higher-Level Cognition Modeling Prize: A Bayesian Account of Reconstructive Memory" @default.
• W2768395108 hasPublicationYear "2008" @default.
• W2768395108 type Work @default.
• W2768395108 sameAs 2768395108 @default.
• W2768395108 citedByCount "0" @default.
• W2768395108 crossrefType "journal-article" @default.
• W2768395108 hasAuthorship W2768395108A5051768325 @default.
• W2768395108 hasAuthorship W2768395108A5082806282 @default.
• W2768395108 hasConcept C107673813 @default.
• W2768395108 hasConcept C151730666 @default.
• W2768395108 hasConcept C154945302 @default.
• W2768395108 hasConcept C15744967 @default.
• W2768395108 hasConcept C161407221 @default.
• W2768395108 hasConcept C169760540 @default.
• W2768395108 hasConcept C169900460 @default.
• W2768395108 hasConcept C180747234 @default.
• W2768395108 hasConcept C188147891 @default.
• W2768395108 hasConcept C21963081 @default.
• W2768395108 hasConcept C2779343474 @default.
• W2768395108 hasConcept C2781238097 @default.
• W2768395108 hasConcept C41008148 @default.
• W2768395108 hasConcept C86803240 @default.
• W2768395108 hasConceptScore W2768395108C107673813 @default.
• W2768395108 hasConceptScore W2768395108C151730666 @default.
• W2768395108 hasConceptScore W2768395108C154945302 @default.
• W2768395108 hasConceptScore W2768395108C15744967 @default.
• W2768395108 hasConceptScore W2768395108C161407221 @default.
• W2768395108 hasConceptScore W2768395108C169760540 @default.
• W2768395108 hasConceptScore W2768395108C169900460 @default.
• W2768395108 hasConceptScore W2768395108C180747234 @default.
• W2768395108 hasConceptScore W2768395108C188147891 @default.
• W2768395108 hasConceptScore W2768395108C21963081 @default.
• W2768395108 hasConceptScore W2768395108C2779343474 @default.
• W2768395108 hasConceptScore W2768395108C2781238097 @default.
• W2768395108 hasConceptScore W2768395108C41008148 @default.
• W2768395108 hasConceptScore W2768395108C86803240 @default.
• W2768395108 hasIssue "30" @default.
• W2768395108 hasLocation W27683951081 @default.
• W2768395108 hasOpenAccess W2768395108 @default.
• W2768395108 hasPrimaryLocation W27683951081 @default.
• W2768395108 hasRelatedWork W171582815 @default.
• W2768395108 hasRelatedWork W2134305700 @default.
• W2768395108 hasRelatedWork W2206217378 @default.
• W2768395108 hasRelatedWork W2397166815 @default.
• W2768395108 hasRelatedWork W2402670475 @default.
• W2768395108 hasRelatedWork W2407106071 @default.
• W2768395108 hasRelatedWork W2408603481 @default.
• W2768395108 hasRelatedWork W2523967999 @default.
• W2768395108 hasRelatedWork W2589068497 @default.
• W2768395108 hasRelatedWork W2589855323 @default.
• W2768395108 hasRelatedWork W2617908825 @default.
• W2768395108 hasRelatedWork W2622740767 @default.
• W2768395108 hasRelatedWork W2624372095 @default.
• W2768395108 hasRelatedWork W2765185783 @default.
• W2768395108 hasRelatedWork W2765279935 @default.
• W2768395108 hasRelatedWork W2766090218 @default.
• W2768395108 hasRelatedWork W282514718 @default.
• W2768395108 hasRelatedWork W3119170277 @default.
• W2768395108 hasRelatedWork W3119850932 @default.
• W2768395108 hasRelatedWork W3183574632 @default.
• W2768395108 hasVolume "30" @default.
• W2768395108 isParatext "false" @default.
• W2768395108 isRetracted "false" @default.
• W2768395108 magId "2768395108" @default.
• W2768395108 workType "article" @default.