Approaches to Cognitive Modeling Approaches to Cognitive Modeling Symbolic Models Symbolic Models Connectionist Models Connectionist Models Hybrid Models Hybrid Models Cognitive Architectures Cognitive Architectures Approaches to Cognitive Modeling 1 What makes a good model? Information Processing Connectionist School Cognitive Modeling Symbolic School Lecture 2: Approaches Symbolic Models 2 Symbolic Representations Production Systems Frank Keller Connectionist Models 3 Parallel Distributed Processing School of Informatics University of Edinburgh Feature Based Representations keller@inf.ed.ac.uk Learning, Generalization, Degradation Hybrid Models January 31, 2005 4 Cognitive Architectures 5 Reading: Cooper (2002: Ch. 1) Frank Keller Cognitive Modeling 1 Frank Keller Cognitive Modeling 2 Approaches to Cognitive Modeling Approaches to Cognitive Modeling What makes a good model? What makes a good model? Symbolic Models Symbolic Models Information Processing Information Processing Connectionist Models Connectionist Models Connectionist School Connectionist School Hybrid Models Hybrid Models Symbolic School Symbolic School Cognitive Architectures Cognitive Architectures A Cognitive Model of a Task What makes a good model? Example: a teacher trying to diagnose the problems a student has with learning subtraction. A good model has two critical properties: Model may consist of a computer program that: 1 it is complete – it does not abstract properties that are takes some representation of the stimulus (the arithmetic test important; items) as input; 2 it is faithful – it does not introduce confounding details during produces a prediction of student s answer as output; abstraction. perhaps also describes the difference between this model and that expected if the student were able to perform the task correctly. Frank Keller Cognitive Modeling 3 Frank Keller Cognitive Modeling 4
Approaches to Cognitive Modeling Approaches to Cognitive Modeling What makes a good model? What makes a good model? Symbolic Models Symbolic Models Information Processing Information Processing Connectionist Models Connectionist Models Connectionist School Connectionist School Hybrid Models Hybrid Models Symbolic School Symbolic School Cognitive Architectures Cognitive Architectures Rise of Cognitive Modeling Information Processing Sensory processes act as input devices: information from the Ebbinghaus: empirical/cognitive psychologist, learned lists of environment converted into internal representation. nonsense words to study memory (1885). Mental processes manipulate and transform these representations, Mid 20th century, it was demonstrated that for higher mental triggering responses via output processes. processes (e.g., language, Chomsky vs. Skinner debate): Major changes in second half of the 20th century: stimulus-response links cannot explain range of behavior; 1 computer simulation techniques used to explore theories of stimulus-response links mediated by internal mental states; cognitive processing and evaluate competing theories of internal mental states essential for causal explanations of empirical phenomena; cognitive processes. 2 brain imaging techniques developed to localize cognitive The mind as an information processor, cognition as information processing and relate functioning of the mind to that of the processing. brain – now primary focus of Cognitive Neuroscience. Frank Keller Cognitive Modeling 5 Frank Keller Cognitive Modeling 6 Approaches to Cognitive Modeling Approaches to Cognitive Modeling What makes a good model? What makes a good model? Symbolic Models Symbolic Models Information Processing Information Processing Connectionist Models Connectionist Models Connectionist School Connectionist School Hybrid Models Hybrid Models Symbolic School Symbolic School Cognitive Architectures Cognitive Architectures Behavior, Process, and Theory Modeling as an Added Dimension Model Classical relation between behavior, cognitive process underlying Simulates Generates the behavior, and theory of the process: Implements Cognitive Behavior Generates Process Cognitive Behavior Generates Process Describes Explains Describes Explains Theory Theory Frank Keller Cognitive Modeling 7 Frank Keller Cognitive Modeling 8
Approaches to Cognitive Modeling Approaches to Cognitive Modeling What makes a good model? What makes a good model? Symbolic Models Symbolic Models Information Processing Information Processing Connectionist Models Connectionist Models Connectionist School Connectionist School Hybrid Models Hybrid Models Symbolic School Symbolic School Cognitive Architectures Cognitive Architectures Modeling and Cognitive Neuropsychology Connectionist School Cognitive Neuropsychology is concerned with different patterns of Two main schools of thought in cognitive modeling: connectionist behavior following neurological damage, and using such patterns to and symbolic. inform normal cognitive functioning. The connectionist school: A Cognitive Model of normal functioning can be damaged or assumes that properties of the neural tissue that implement lesioned in a principled way. information processing in the mind are critical; We compare behavior of the damaged model with that of builds models of many simple interacting units functioning in neurological patients, to account for both normal and impaired parallel; performance. typically regards these models as analogues of neurons or Cognitive neuropsychology can also provide data against which neural cell assemblies. models can be tested. Frank Keller Cognitive Modeling 9 Frank Keller Cognitive Modeling 10 Approaches to Cognitive Modeling Approaches to Cognitive Modeling What makes a good model? What makes a good model? Symbolic Models Symbolic Models Information Processing Information Processing Connectionist Models Connectionist Models Connectionist School Connectionist School Hybrid Models Hybrid Models Symbolic School Symbolic School Cognitive Architectures Cognitive Architectures Symbolic School Other Approaches The symbolic school: describes information processing in terms of the manipulation Hybrid approaches: try to combine the strengths (and avoid the of symbol representations; weaknesses) of both connectionist and symbolic approaches. views the neural substrate as an implementation medium of secondary importance. Architectural approach: Share: idea that the functioning of the mind is computational and hypothesized organization of complete set of information can be simulated by machine. processing structures that comprise the mind/brain; use this as a guide to developing models, e.g., ACT-R and Differ in: Cogent. their approaches; assumptions about mental representations; views on relation between a cognitive model and the brain. Frank Keller Cognitive Modeling 11 Frank Keller Cognitive Modeling 12
Approaches to Cognitive Modeling Approaches to Cognitive Modeling Symbolic Models Symbolic Models Symbolic Representations Symbolic Representations Connectionist Models Connectionist Models Production Systems Production Systems Hybrid Models Hybrid Models Cognitive Architectures Cognitive Architectures Symbolic Propositional Representations Properties of Symbolic Representations A representation is systematic if: it consists of a number of parts and Conjunctions of propositions concerning objects, their properties, replacing some part with other parts of the same kind is also a and relations between them. meaningful representation. Example: the red pyramid is on the blue cube. A representation is compositional if: pyramid ( p ) & red ( p ) & cube ( c ) & blue ( c ) & on ( p , c ) it consists of a number of parts and the meaning of the whole is a function of the meaning of the Propositions may be true or false, depending on the state of the parts (Fodor and Pylyshyn 1988). objects to which they apply. Representations that are systematic and compositional may be manipulated by rules dependent only on the form of the representation, not its meaning. Frank Keller Cognitive Modeling 13 Frank Keller Cognitive Modeling 14 Approaches to Cognitive Modeling Approaches to Cognitive Modeling Symbolic Models Symbolic Models Symbolic Representations Symbolic Representations Connectionist Models Connectionist Models Production Systems Production Systems Hybrid Models Hybrid Models Cognitive Architectures Cognitive Architectures Symbolic Representations Production Systems Provide a general means of representing information. Provide a general, yet constrained, framework for symbolic models: Are supplemented by symbol manipulation rules that operate on Propositional store: propositions referred to as Working memory the representations to transform them, or build new ones. elements (WME) (analogue of working or short term memory). Symbolic programming languages commonly used for developing Rule database: inference rules (productions) correspond to long cognitive models: term (general and task specific) knowledge (analogue of long term Lisp (early 60s, John McCarthy & co. at MIT): based in part memory). on work by Newell, Shaw and Simon at CMU; Recognize phase: inference rule selected; Prolog (early 70s, Alain Colmeraur & co., France). act phase: selected rule applied; Problem: general purpose programming languages too conflict resolution procedure for selection between rules. unconstrained for cognitive modeling. Alternative: production systems. Frank Keller Cognitive Modeling 15 Frank Keller Cognitive Modeling 16
Recommend
More recommend
