Cognitive Science Midterm Review PDF

Introduction to Cognitive Science Core Principles: ◦ Information Processing: Mental processes involve creating, storing, transforming, and retrieving information. ◦ Representation and Meaning: Symbols in the mind re ect external reality....

Introduction to Cognitive Science Core Principles: ◦ Information Processing: Mental processes involve creating, storing, transforming, and retrieving information. ◦ Representation and Meaning: Symbols in the mind re ect external reality. Distinct from Other Disciplines: ◦ Phenomenology: Cognitive science diverges by focusing on information processing rather than introspective experiences. Phenomenology is a philosophical approach to understanding the human experiences, founded by Edmund Husserl in the early 20th century—it focuses on rst-person and subjective experiences/ how things appear to us in our consciousness rather than examining the external world directly or explaining mental processes, phenomenology studies the structures of experience and consciousness from a descriptive perspective. ◦ Example: Sartre: “When I enter the café to look for Pierre, a synthetic organization of all the objects in the café is formed, against which Pierre is given as having to appear. And this organization of the café as a ground is a rst nihilation. Each element in the room—person, table, chair— tries to separate itself, to detach itself against the ground constituted by the totality of the other objects, and then collapses back into that undifferentiated ground, and is diluted within it.” ◦ Example: “[Pierre’s] absence freezes the cafe in its evanescence; the cafe remains as ground; it continues to present itself to my merely marginal attention as an undifferentiated totality; it slides away, in pursuit of its nihilation …Pierre, absent, haunts this cafe and is the condition of its nihilating organization as a ground.” ◦ Differences from phenomenology ◦ Focus on conscious experiences ◦ Description, not explanation ◦ Trust in introspection Focus and Goals: Cognitive Science: Seeks to understand the mechanisms and information-processing structures that underlie cognition, such as memory, perception, and decision-making. It is an interdisciplinary eld that uses computational models, neuroscience, and psychology to study mind and behavior scienti cally. Phenomenology: Aims to describe conscious experience itself, without reducing it to underlying mechanisms or explaining how it arises. Phenomenologists seek a pure, detailed account of subjective experiences as they appear in consciousness. Methodology: Cognitive Science: Uses experimental methods, neuroimaging, and computational modeling to gather quantitative data on behavior and brain activity. It aims to produce objective, replicable ndings that explain mental functions. Phenomenology: Relies on introspection and qualitative analysis to describe the essence of experiences. It does not prioritize objective measurement or data-driven models, focusing instead on the richness and nuance of subjective experience. fi fi fi fi fl fi  Explanation vs. Description: Cognitive Science: Tries to explain how cognitive functions are carried out by the brain or arti cial systems, often seeking cause-and-effect relationships. Phenomenology: Describes what experience is like without theorizing about underlying causes or neural mechanisms. Trust in Introspection: Cognitive Science: Generally does not rely on introspection as a primary source of data, since it considers subjective reports to be biased or incomplete. Phenomenology: Considers introspection a valid and valuable method, believing that only the individual experiencing a phenomenon can describe its qualities accurately. Cognitive Science isn’t Psychoanalysis Example: Little Hans was terri ed of horses with black muzzles or blinders because he associated these horses with his father’s mustache and glasses. His terror was an extension of his Oedipus complex. Example: Sergei Pankejeff was depressed because of a suppressed sexual fantasy in which his father was the predator and he was the prey Similarities: explanation, not mere description; focus on unconscious processes; distrust of introspection. Differences: focus on idiosyncratic behaviors, focus on complex behaviors, focus on sexual drives, methods include free association and transference; explanations are thermodynamic, but not mathematical. ◦ Behaviorism: Cognitive science considers internal processes, unlike behaviorism, which focuses on observable behaviors. ◦ Example: If the mouse is rewarded for pressing the lever, it will press the lever more often. ◦ Example: If the mouse is only rewarded for pressing the lever when the light is on, it will press the lever only when the light is on. ◦ Similarities: focus on simple. Measurable behavior; focus on universal behavior; distrust of introspection; use of mathematical models. ◦ Dissimilarities: explanations don’t reference internal processes. ◦ Pharmacology: example—Parkinson’s patients perform better on memory tasks after taking L-Dopa. ◦ Similarities: Interest in internal processes; focus on measurable effects; use of mathematical models. ◦ Differences: Explanations are not computational in the same sense; they are not about information processing?? ◦ Biology: Example: adding potassium to the exterior of a neuron causes it to re more often. ◦ Similarities: interest in internal processes; use of computational models; explanations are sometimes computational in the sense that they are about information processing (e.g. cell signaling). ◦ Differences: Often no direct connection to behavior; often not trying to explain phenomena with no obvious connection to perception, memory, decision-making, memory, motor control, etc. The Cognitive Approach Overview fi fi fi  Mind as Information Processor: The cognitive approach treats the mind as processing information, similar to a computer. Levels of Explanation: ◦ Goal-Constraint Level: De nes overall purpose. ◦ Algorithmic Level: Details speci c transformation rules. ◦ Implementation Level: Explores physical structures. Perception and Bias: ◦ Oddball Effect: Novel stimuli distort time perception, creating a sense of elongation. ◦ Flash Lag Effect: Moving objects appear further along in their path than they are, possibly to aid in real-time action coordination. Representation in Cognitive Science Representation Types: ◦ Symbols and Propositions: Logical statements that represent relationships (e.g., "Chris likes cheese"). ◦ Maps and Images: Maps as spatial representations (Tolman’s 1948 cognitive maps). Mental Imagery Debate: Kosslyn vs. Pylyshyn argued whether internal representations are picture-like (visual) or propositional (abstract). ◦ Behavioral Studies: Shepard & Metzler (1971) showed that reaction times in mental rotation tasks suggest visual representation. ◦ Population Codes: Multiple neurons represent values across combinations, enhancing cognitive exibility and resilience. Neural Network Computation Mind as a Computer: Traditional cognitive science viewed the mind as a classical computer executing algorithms. Arti cial Neural Networks (ANNs): ◦ Training: Networks learn by adjusting weights based on output errors (backpropagation). ◦ Applications: ANNs surpass classical models in tasks like image recognition and language generation. ▪ ImageNet (2012): Large-scale image classi cation, where ANNs began outperforming humans. ◦ Strengths and Weaknesses: ANNs excel in pattern recognition but are "black- box" systems, where internal processes are opaque. Nature vs. Nurture Nativism vs. Empiricism: ◦ Nativism: Posits that cognition is genetically preprogrammed, with brain regions developing autonomously. ◦ Empiricism: Argues cognitive processes emerge from environmental exposure. Shared Cognitive Features: ◦ Face Processing in Fusiform Face Area (FFA): Arcaro et al. (2017) raised macaques without exposure to faces and found they lacked typical face- processing abilities, supporting a nature-based explanation for FFA development. fi fl fi fi fi  ◦ Visual Illusions: Universally experienced illusions like the Ponzo illusion imply shared cognitive architecture. Heritability Studies: ◦ Twin Studies: Traits like intelligence and personality show varying heritability; monozygotic twins (identical) are more alike than dizygotic twins (fraternal). ◦ Turkheimer’s Laws of Behavioral Genetics: All traits are heritable, family effects are smaller than genetic ones, and a portion of trait variation is unexplained by genes or families. Modularity in Cognitive Science Modularity: Theory that the mind has specialized, independent "modules" responsible for speci c tasks (e.g., face recognition, language processing). ◦ De ning Features: Encapsulation (independence from other modules), specialization (speci c algorithms), and localization (speci c brain regions). Case Studies: ◦ Low-Level Perception: Studies on visual perception suggest it may be modular, though top-down effects can in uence low-level sensory processing. ▪ Mood Experiments: Banerjee et al. (2012) found thinking about unethical behavior made a room appear darker. ▪ Physical Load Experiment: Bhalla & Prof tt (1999) observed that carrying a heavy backpack made hills appear steeper. ▪ Face Categorization: Levin & Banaji (2006) found that Black faces appeared darker due to categorization biases. ◦ High-Level Perception (Face Perception): ▪ Prosopagnosia: Condition where people recognize non-faces but not faces, indicating face recognition is modular. ▪ Part-Whole Effect: Recognizing a face part (like an eye) is easier when seen as part of a whole face, supporting modularity. Linguistics and Syntax Linguistic Competence: Pinker (1994) argues humans uniquely combine words to convey sophisticated, hypothetical ideas. Animals like apes learn isolated words but lack complex syntax. Phonotactics: Phonology rules, such as in English (e.g., certain consonant clusters like “ng” are only allowed at the end of words), demonstrate the structured sounds permissible in each language. ◦ Examples: ▪ French lacks the “th” sound. ▪ Japanese con ates "l" and "r" sounds. ▪ Hawaiian prohibits codas (syllable endings). Syntax and Recursive Structure: ◦ Tree Diagrams: Linguists use these to illustrate how sentence components like subjects, verbs, and objects are organized. ◦ Recursive Rules: Allow in nitely complex sentences by embedding phrases within phrases, as shown in cumulative sentences like "This is the cat that chased the mouse...". fi fi fl fi fi fl fi fi  Semantic Ambiguities: Structural and lexical ambiguities, such as “John threw a ball” (ball as an object or a dance), highlight exibility and complexity in language interpretation. Language Learning and Universal Grammar Language Learning Basics: Language transmits complex information and involves phonetics (sounds), morphology (word construction), syntax (sentence construction), semantics (meaning), and pragmatics (language use in context). Theories of Language Acquisition: ◦ Nativism (Universal Grammar - UG): Noam Chomsky proposed that humans have an innate, language-speci c module. UG implies language principles are hardwired into the brain. ▪ Poverty of the Stimulus: Children quickly infer grammar rules despite limited input, suggesting an innate bias. ▪ Linguistic Universals: Universal grammar features appear across languages, like noun/verb distinctions and recursive sentence structures. ▪ Spontaneous Language Creation: Deaf children’s development of sign language grammar (home sign systems) and structured Creole languages from Pidgin support UG. Anti-UG Hypothesis: Empiricist view that language acquisition comes from general cognitive abilities rather than a language-speci c module. ◦ Against Poverty of the Stimulus: Children may rely on simple explanations (Occam’s Razor) and pattern recognition to learn syntax, with hierarchical structures arising through general learning processes. ◦ Counter-Examples to Universals: Languages like Straits Salish lack distinct word classes, while languages like Riau Indonesian don’t follow typical syntactic rules, challenging the concept of universal linguistic rules. Levels of Explanation Marr’s Three Levels: ◦ Computational: Focuses on the goals and properties of a system, such as a cash register’s goal to sum prices regardless of item order. ◦ Algorithmic: Describes the data manipulation processes, as with adding digits to calculate totals. ◦ Implementational: Physical realization of processes, like gears in a cash register adding numbers. Examples of Explanation: ◦ Donkey Kong: ▪ Computational: When Mario hits a barrel with a hammer, it disappears, illustrating goal-oriented design. ▪ Algorithmic: Position checks (position_Mario = position_barrel) to control interactions. ▪ Implementational: Storing position data in RAM highlights the hardware’s role. ◦ Interpersonal Distance: Huskey et al. (2020) studied reactions to interpersonal distance violations, noting that familiarity or liking in uences expected distance, which is tracked by different regions in the striatum. fi fl fi fl  ◦ Ultimatum Game: A game demonstrating computational theory in behavior. Henrich et al. (2006) used this to illustrate fairness concepts across cultures, highlighting computational and algorithmic levels of decision-making based on offers and rejections. Evolutionary Psychology Foundations: Evolutionary psychology investigates how the human brain developed to solve survival and reproductive challenges in ancestral environments. It uses the "computer analogy," proposing that the brain operates like pre-programmed systems designed by natural selection. Evolution and the Brain: Big brains are costly ("expensive"), so their evolution must have conferred signi cant survival advantages. Evolutionary psychology seeks to identify psychological traits that helped ancestors survive, like social cooperation or mate selection. Examples of Natural Selection: ◦ Peppered Moth: Classic example where dark and light-colored moths uctuated in frequency based on environmental changes due to pollution, showcasing survival-related phenotypic adaptation. ◦ Galápagos Finches: Studied by Darwin; nches developed different beak shapes based on available food, illustrating adaptive radiation. ◦ Mole Rats and Hypoxia: Mole rats can survive in low-oxygen environments for extended periods, demonstrating physiological adaptations for survival. ◦ Sea Slug and Photosynthesis: This species can photosynthesize, a unique adaptation to its environment. Phenotypic Plasticity: Genes can express differently depending on environmental conditions. For example, male salamanders might eat their offspring under stress, showing behavioral plasticity as an adaptive response. Methodologies: ◦ Backward Method: Identifying traits rst and then speculating on their evolutionary purpose (e.g., why do people laugh?). This method risks creating "just-so stories" without strong evidence. ◦ Forward Method: Analyzes ancestral needs from the Pleistocene (1.8 million to 10,000 years ago) and then seeks corresponding psychological traits today. Example needs include identifying edible plants, responding to alarm cries, assisting children, mate selection, and interpreting social cues. ◦ Criticisms of Forward Method: ▪ Hypotheses about the Pleistocene Era are challenging to test. ▪ Psychological traits could result from multiple genes or be in uenced by modern environments. ▪ Traits like chin development in humans may be byproducts ("spandrels") rather than direct evolutionary adaptations. Cognitive Development (Development) Nativism vs. Empiricism in Development: ◦ Nativism: Cognitive skills are innate and only re ned through experience. ◦ Empiricism: Cognitive abilities are learned through exposure to environmental patterns. fi fi fi fi fl fl  ◦ Core Cognition Hypothesis: Suggests infants possess specialized cognitive systems from birth, like numerical and object permanence concepts. Core Cognitive Abilities: ◦ Numerical Cognition: ▪ Wynn (1992): Used looking-time studies showing infants’ basic arithmetic understanding, e.g., infants look longer at unexpected numerical outcomes. ▪ Feigenson and Carey (2005): Demonstrated infants can accurately track up to four discrete objects, beyond which accuracy declines. ◦ Object Permanence: ▪ Baillargeon’s Studies: Infants as young as a few months understand object permanence, the concept that objects continue to exist when unseen. ▪ Spelke et al. (1995): Showed infants expect objects to move continuously unless obstructed, a foundational concept in physical reasoning. ◦ Social Cognition: ▪ Horner and Whiten (2004): Found that 4-year-olds and chimpanzees exhibit overimitation, copying even inef cient actions, possibly for social bonding or learning. ▪ Woodward (1999): Documented that 5-month-olds show longer looking times for goal-oriented behaviors (e.g., a hand reaching for a new toy), suggesting early agency recognition. Theory of Mind Development: ◦ False-Belief Task: Tested by showing children a scenario where one character holds a belief that is false (e.g., Sally and Anne test); children under four struggle, not understanding others’ minds. ◦ Perceptual Access Reasoning: Fabricius et al. (2021) suggested children rely on perception-based rules (e.g., if someone sees something, they know it) rather than understanding mental states. Language Acquisition: ◦ Stages of Overregularization: Children initially use correct irregular forms (e.g., “ate”), then overapply rules (e.g., “eated”) due to imperfect retrieval; this developmental stage aligns with Marcus et al. (1992)’s blocking and retrieval failure hypothesis. Classical Computation (ClassicalComputation) Historical Development of Computing Models: ◦ Antikythera Mechanism (~100 BC): An early analog device calculating astronomical positions, although not programmable. ◦ Charles Babbage and Ada Lovelace: Babbage’s Analytical Engine and Lovelace’s pioneering algorithms marked the beginning of programmable computing, illustrating early symbolic computation. Turing Machines: ◦ Concept: A theoretical machine with an in nite memory tape, a read/write head, and a lookup table for state-based actions, which embodies the concept of algorithmic computation. fi fi  ◦ Church-Turing Thesis: States that any process computable by humans can be computed by a Turing machine, establishing a foundation for computational theory. ◦ Binary and Logical Operations: McCulloch & Pitts (1943) showed neurons could model logical operations (e.g., NAND gates), suggesting brains can implement basic computational functions. Classical Computational Theory of Mind (CCTM): ◦ Core Principles: Asserts that the mind is a computational system, with processes akin to algorithms on symbolic representations. ◦ Example: ACT-R Model (Adaptive Control of Thought-Rational): This cognitive architecture models procedural and declarative memory in a rule-based, step-by- step format. General-Purpose Computers: ◦ Universal Turing Machine (UTM): Turing’s concept of a machine that can execute any algorithmic process given the correct input, a precursor to modern computers. ◦ Von Neumann Architecture: Describes digital computing architecture with stored programs, input/output processing, and sequential instructions—key to modern computing models. Brain Scanning and Measurement Techniques (BrainScanning) Techniques for Measuring Brain Activity: ◦ Implanted Electrodes: ▪ Advantages: Direct access to neuron ring rates, providing precise measurements. ▪ Disadvantages: Limited spatial coverage; requires invasive surgery (Surbeck et al., 2011). ◦ Electroencephalography (EEG): ▪ Advantages: Non-invasive, portable, and relatively affordable. ▪ Disadvantages: Signals blur at the scalp, limiting precise spatial resolution, especially in deep brain structures. ◦ Magnetoencephalography (MEG): ▪ Advantages: Non-invasive and capable of detecting magnetic signals, which pass through the skull more effectively. ▪ Disadvantages: Weak signal strength requires expensive equipment and doesn’t capture deep structures. ◦ Functional MRI (fMRI): ▪ Mechanism: Tracks blood oxygenation levels (BOLD signal) as a proxy for neuronal activity, since active neurons consume more oxygen. ▪ Advantages: High spatial resolution (~1mm) and access to deep brain regions. ▪ Disadvantages: Poor temporal resolution and indirect measurement of neuron activity. ◦ Diffusion MRI: Measures the integrity and orientation of axon bundles by tracking water diffusion along bers. Applications of Neuroimaging: fi fi  ◦ Timing in Cognitive Tasks: Borst & Anderson (2021) used EEG to track timing in cognitive stages, predicting retrieval speed differences (Fan-1 vs. Fan-2 conditions). ◦ Structured Learning in Hippocampus: Schapiro et al. (2018) showed hippocampal activation in response to learned sequences, observable even in infants. ◦ Behavior Prediction: Falk et al. (2011) used neural activity in the medial frontal cortex to predict behavior change (e.g., reduced smoking) more effectively than self-reported ef cacy. ◦ Individual Differences: Schwarzkopf et al. (2011) found that the size of visual area V1 correlates with the strength of visual illusions, suggesting low-level feature interactions. ◦ Tracking Representational Learning: Meshulam et al. (2021) studied fMRI representations in learning environments, nding shared spatial activity patterns in the brain for similar concepts. Brain Structure and Function (Brain) Historical Foundations: ◦ Ancient Egyptians linked brain injury to mental symptoms as early as the 17th century BC. ◦ Aristotle’s Heart Hypothesis (4th century BC): Posited the heart, not the brain, as the center of sensation, reasoning that the heart shows physical responses to emotion and pain. ◦ Santiago Ramón y Cajal (late 1800s): Discovered neuron structure, illustrating synaptic connections fundamental for information transmission. Basic Brain Processes: ◦ Neuronal Firing: ▪ Voltage Threshold: A neuron res when its voltage exceeds a threshold due to excitatory inputs, sending spikes down the axon to in uence other neurons. ▪ Synaptic In uence: Synapses modulate neuron ring by either increasing (excitatory) or decreasing (inhibitory) the likelihood of an action potential. ◦ Center-Surround Receptive Field: ▪ Neurons in retinal cells detect contrasts (e.g., edges) via spatial arrangements of excitatory and inhibitory inputs. Learning Mechanisms: ◦ Spike Timing Dependent Plasticity (STDP): ▪ Mechanism: Strengthens synapses if a presynaptic spike precedes a postsynaptic spike (increases weight) and weakens them if it follows (decreases weight) (Zhang et al., 1998). ▪ Signi cance: Supports unsupervised learning by enhancing sensitivity to input patterns without speci c goals or feedback, a fundamental mechanism for biological learning. Cortical Mapping: fi fl fi fi fi fi fi fl  ◦ Visual Cortex Maps: Regions in the visual cortex show retinotopic and orientation-speci c organization, helping structure visual inputs spatially and by stimulus orientation (e.g., Heeger, 2006). ◦ Functional Specialization: Different areas such as V1 ( rst visual processing region) handle unique aspects of sensory processing, supporting object recognition and spatial memory through distributed yet interconnected systems (e.g., Felleman & Van Essen, 1991). fi fi