Knowledge, Memory, and Forgetting Lecture Notes PDF

KNOWLEDGE, MEMORY, AND FORGETTING BA - COG PSY 1. LECTURE NAGY MÁRTON [email protected]  KNOWLEDGE, MEMORY, AND FORGETTING Important theories of semantic memory: Schemas: Bartlett Scripts: Schank Semantic networks (Conc...

KNOWLEDGE, MEMORY, AND FORGETTING BA - COG PSY 1. LECTURE NAGY MÁRTON [email protected]  KNOWLEDGE, MEMORY, AND FORGETTING Important theories of semantic memory: Schemas: Bartlett Scripts: Schank Semantic networks (Concepts, categorization lecture) Prototype and exemplar theory (Concepts, categorization lecture) The role of knowledge in memory: Deese-Roediger-McDermott (DRM) paradigm Encoding context affects later memory recall (labeling effect) - Carmichael, 1932 Expert knowledge - specific area involved Rule learning as the creation of a general rule from individual experiences and its application to entirely new stimuli (complex learning, Szabó Eszter) Healthy forms of forgetting: Trace decay hypothesis Interference theory: proactive & retroactive interference Consolidation disorders: classic theory of consolidation (Frankland & Bontempi, 2005) & multiple trace theory (Nadel Moscovitch,1997) Memory inhibition phenomena: intentional forgetting (IF): think vs. don't think; and repetition-induced inhibition (RIF) Draw a giraffe! What are these? : What is a car? Where did you go to high school? 3 https://quickdraw.withgoogle.com/#  KNOWLEDGE, REPRESENTATION Mental representations of knowledge and concepts Main question: How can something be "described"? We can do experiments to try to infer the characteristics. Describing/modeling complex situational knowledge is a challenging task. Mental Representation: Stored & retrievable Manipulable Sketchy/General - most of the time MODELS OF SEMANTIC NETWORKS: KNOWLEDGE AS A MEMORY SYSTEM SEMANTIC NET WORK MODELS Collins & Quillian (1969) Collins & Loftus (1975) HIERARCHICAL ORGANIZATION MODEL OF SPREADING ACTIVATION PRINCIPLE OF COGNITIVE ECONOMY SEMANTIC DISTANCE IS IMPORTANT QUESTIONS TO RESOLVE IN RT MEASUREMENT: AMONG THE ELEMENTS. IS THE CANARY YELLOW? < CAN THE CANARY FLY? / DOES IT HAVE SKIN? MORE FLEXIBLE. PROBLEMS: FAMILIARITY AND TYPICALITY EFFECTS  SCHEMA SIR FREDERIC BARTLETT Connected knowledge set about the world, specific events, people, and actions (Baddeley, Eysenck & Anderson textbook: Frameworks that organize and facilitate our knowledge but also limit the organization of our memories. (Pléh Csaba, Lukács Ágnes (eds.): Psycholinguistics) War of the Ghosts story: An Indian tale was read to experimental subjects – English university students Culture-specific effects in recall: Omissions Reformulations Memory is constructive: Memory: Representation constructed based on schemas  WHAT IS IT ABOUT? "The procedure is not complicated at all. First, organize the items into groups. Of course, a single pile may also be suf cient, depending on how much work awaits us. If something is still important, go ahead and retrieve it; otherwise, this much preparation should be enough to get started. It’s very important not to overdo it. It's better to do less at a time than too much. It might not seem so signi cant, but complications can easily arise. Mistakes can be very costly. At rst, the whole procedure may seem complicated. However, soon it will not cause more trouble than anything else in life. It’s very hard to foresee how soon you will need this, but you can never know. After completing the procedure, reorganize the items. Then, you can put them back in their place. If they are used again, then the whole cycle must be repeated. This too is part of life.” (- translated from Hungarian not exact match with original text) (Rumelhart, 1975) Without a title (activated schema), it's dif cult to interpret this text. It's hard to highlight the essence and remember the text/steps. fi fi fi fi  THE BIG WASHING OF CLOTHES "The procedure is not complicated at all. First, organize the items into groups. Of course, a single pile may also be suf cient, depending on how much work awaits us. If something is still important, go ahead and retrieve it; otherwise, this much preparation should be enough to get started. It’s very important not to overdo it. It's better to do less at a time than too much. It might not seem so signi cant, but complications can easily arise. Mistakes can be very costly. At rst, the whole procedure may seem complicated. However, soon it will not cause more trouble than anything else in life. It’s very hard to foresee how soon you will need this, but you can never know. After completing the procedure, reorganize the items. Then, you can put them back in their place. If they are used again, then the whole cycle must be repeated. This too is part of life.” (- translated from Hungarian not exact match with original text) (Rumelhart, 1975) Without a title (activated schema), it's dif cult to interpret this text. It's hard to highlight the essence and remember the text/steps. The abstraction (schema): structuring of prior expectations, highlighting the essence, and aiding memory. fi fi fi fi  SCRIPTS AND FRAMES ROGER C. SC HANK Scripts organize our memories similarly to schemas. A script is a type of schema that describes the typical sequence of events and their consequences in everyday situations (e.g., a restaurant dinner). Frames are representations that describe certain aspects of the world or objects and their characteristics (e.g., buildings). Structured information and "empty values" are inherent in these models, helping to guide and limit our interpretation and memory based on the structured knowledge.  THE EFFECT OF KNOWLEDGE REPRESENTATIONS ON OUR MEMORY DEMONSTRATION DEESE ROEDIGER MCDERMOT (DRM) PARADIGM (ORIGINAL WAS WITH WORDS) THE EFFECT OF KNOWLEDGE ON MEMORY INTERPRETATION/L ABELING EFFECT TATIONAL BIOLOGY THE EFFECT OF KNOWLEDGE ON MEMORY PLOS COMPUTATIONAL BIOLOGY Optimal forgetting: Semantic compression of episodic memories RESEARCH ARTICLE PLOS COMPUTATIONAL BIOLOGY Optimal forgetting: Semantic compression of episodic memories Optimal forgetting: Semantic compression of episodic memories David G. Nagy ID1,2*, Balázs Török1,3, Gergő Orbán ID1 1 Computational Systems Neuroscience Lab, Wigner Research Centre for Physics, Budapest, Hungary, 2 Institute of Physics, Eötvös Loránd University, Budapest, Hungary, 3 Department of Cognitive Science, Budapest University of Technology and Economics, Budapest, Hungary * [email protected] Abstract Context (label) at encoding has It has extensively been documented that human memory exhibits a wide range of system- atic distortions, which have been associated with resource constraints. Resource con- an effect on later recall from straints on memory can be formalised in the normative framework of lossy compression, however traditional lossy compression algorithms result in qualitatively different distortions memory. to those found in experiments with humans. We argue that the form of distortions is charac- teristic of relying on a generative model adapted to the environment for compression. We n G (2020) show that this semantic compression framework can provide a unifying explanation of a pression of Biol 16(10): wide variety of memory phenomena. We harness recent advances in learning deep genera- /journal. tive models, that yield powerful tools to approximate generative models of complex data. We use three datasets, chess games, natural text, and hand-drawn sketches, to demon- ard University, strate the effects of semantic compression on memory performance. Our model accounts for memory distortions related to domain expertise, gist-based distortions, contextual effects, and delayed recall. is an open Quickdraw Author summary drawings Human memory performs surprisingly poorly in many everyday tasks, which have been e terms of the richly documented in laboratory experiments. While constraints on memory resources ense, which necessarily imply a loss of information, it is possible to do well or badly in relation to avail- on, and ided the original able memory resources. In this paper we recruit information theory, which establishes how to optimally lose information based on prior and complete knowledge of environ- mental statistics. For this, we address two challenges. 1, The environmental statistics is not FICS games w.ficsgames.org known for the brain, rather these have to be learned over time from limited observations. available at 2, Information theory does not specify how different distortions of original experiences ab/quickdraw- should be penalised. In this paper we tackle these challenges by assuming that a latent var- available at iable generative model of the environment is maintained in semantic memory. We show Fig 3. Context effects on reconstruction of line drawing from memory. A: Middle column, Ambiguous line drawings from the QuickDraw data set of eyeglasses ki/. that compression of experiences through a generative model gives rise to systematic dis- and dumbbells. Left and right columns, Reconstructions of the image from memory in the dumbbell and eyeglasses contexts, respectively. Context is modelled by tortions that qualitatively correspond to a diverse range of observations in the experimen- using a sketch-VAE trained on sketches from a single category with beta = 2. B: Examples of ambiguous drawings (middle column) and their reconstructions (side ported by the columns) when cues are provided to participants (shown as text labels). Data is reproduced from. C: Effect of contextual information on the visual features in nd Innovation tal literature. recalled stimuli in the model. Quantitative changes (top), qualitative changes (middle), and subtle changes in characteristics (bottom) occur as a result of contextual K125343, https:// recall. D: Quantitative changes in visual features with changing context (proportion of the length of the line connecting circular features in the eyeglasses and Fig 3. Context effects on reconstruction of line drawing from memory. A: Middle column, Ambiguous line drawings from the QuickDraw data set of eyeglasses  THE EFFECT OF KNOWLEDGE REPRESENTATIONS ON OUR MEMORY DEESE-ROEDIGER -MCDERMOT (DRM) PARADIGM bucket garden table spaceship painter photographer bride phone tourist bus/coach beach ball  THE EFFECT OF KNOWLEDGE REPRESENTATIONS ON OUR MEMORY DEESE-ROEDIGER -MCDERMOT (DRM) PARADIGM bucket (Learned) garden table (Learned) spaceship (Critical element) painter (Novel) photographer (Learned) bride (Novel) phone (Critical element) tourist bus/coach (Critical element) beach ball (Critical element)  THE EFFECT OF KNOWLEDGE REPRESENTATIONS ON OUR MEMORY PLOS COMPUTATIONAL BIOLOGY EXPERT KNOWLEDGE Optimal forgetting: Semantic compression of episodic memories Fig 1. Effect of domain expertise on memorising positions in chess. A: Top, Chess board configurations from real game settings (game configurations). Bottom, reconstructions of the configurations from memory. These configurations are individual samples generated by the expert model based on the encoding of the presented configurations. Green frames indicate correctly reconstructed pieces, red frames indicate positions where a piece is missing or erroneously appears in the reconstructed game; purple frames indicate pieces whose identity is switched in the reconstruction. B: Same as A but instead of game configurations randomly shuffled pieces are presented (random configurations) and reconstructed. C: Reconstruction accuracy of the model for game and random configurations as a function of the training size. D: Reconstruction accuracy of human participants as a function of chess skill. Data reproduced from. Nagy, Török, Orbán (2020) https://doi.org/10.1371/journal.pcbi.1008367.g001 the learned generative model to encode either game or random boards into a latent representa- tion. Then, conditioning on the stored latent state we used the generative model to decode the memory trace into a reconstruction of the chess board configuration (Fig 1A and 1B). Recon- structions by the model show the monotonic increase in accuracy for ‘game’ boards as a func-  THE EFFECT OF KNOWLEDGE REPRESENTATIONS ON OUR MEMORY RULE LEARNING ABA or ABB rule in infants (~8 months olds) “ga ti ga” or “ga ti ti” tri-syllable pseudowords (Marcus et al., 1999) - rule learning and its generalization to new contexts“wo fe wo” vs. “wo fe fe” With visual stimuli also (Saffran et al., 2007)  FORGETTING Philosophical question: does it even exist? Accessibility (retrievable at a given moment) and availability (stored, but not currently retrievable) (Tulving: recall vs. recognition) Wagenaar (1986) diary studies - successful retrieval alongside the presentation of appropriate cues (what, where, with whom, what?) Is it bad? - Luria: the case of S. FORGETTING CURVE  WHAT IS BEHIND FORGETTING? 1. Trace decay hypothesis 2. Interference theory 3. Consolidation disorders 4. Memory inhibition phenomena  DECAY OF MEMORY TRACE Example from previous research: STM Brown-Peterson task But it is not possible to measure the effect of only time on the memory trace. There are also other processes, information coming in to the system.  INTERFERENCE Similar memories disrupt each other. Proactive and retroactive interference Proactive: The tendency for earlier memories to compete (disrupt retrieval) at retrieval with new memories. Retroactive: New memories disrupt the retrieval of earlier memories. FORGETTING CONSOLIDATION DEFICIT Standard consolidation model (Frankland & Bontempi (2005)  FORGETTING CONSOLIDATION DEFICIT Multiple trace theory (Nadel, Moscovitch , 1997) (Frankland & Bontempi (2005)  FORGETTING AS AN ACTIVE, EFFORTFUL PROCESS BJORK & RETRIEVAL INHIBITION Previously relevant but currently irrelevant memory elements are inhibited with respect to the given goals. This means their access will be limited without affecting their availability. Inhibition signifies active suppression, meaning that the inhibition is not an automatic consequence of the strengthening of other connections.  INHIBITION BASED EXPERIMENTAL PARADIGMS Intentional forgetting paradigms: Directed Forgetting –DF Think/No-Think task Retrieval Induced forgetting – RIF  DIRECTED FORGETTING (BJORK ET AL., 1968, 1970, 1996) First list “Remember” instruction “Forget” instruction Second list Recall DIRECTED FORGETTING (Hanslmayr et al (2012)  THINKS VS. NO-THINK TASK Sullivan, Danielle & Marx, Brian & Chen, May & Depue, Brendan & Hayes, Scott & Hayes, Jasmeet. (2019). Sullivan, Danielle & Marx, Brian & Chen, May & Depue, Brendan & Hayes, Scott & Hayes, Jasmeet. (2019). RETRIEVAL INDUCED FORGETTING Buchli, 2023  AMNESIA INJURY Retrograde amnesia Anterograde amnesia deficit in previously acquired memories deficit in forming new memories Usually with temporal gradient Less deficit Significant deficit remote memories in recent memories Past Future Based on Markowitch (2008) https://doi.org/10.1016/S0072-9752(07)88007-9  AMNESIAS READ WITH C AUTION , NOT UP TO DATE TAXONOMY Parkin & Leng (1993)  DONEC QUIS NUNC H.M: amnesic syndrome Anterograde amnesia with temporal gradient retrograde K.C.: Anterograde amnesia with extensive retrograde amnesia for episodic memories, while semantic memory was retained.  ALTERNATIVE MEMORY TAXONOMY 3 basic processing modes: Rapid vs. Slow encoding Associative vs. Single item encoding Flexible (compositional) vs. Rigid (unitized) Henke (2010)