PSYB57 Final: Knowledge
We use knowledge to make ___?
Inferences that go beyond the data. Knowledge is from semantic memory, which are facts we know about the world and ourselves.
The fact that we make similar inferences as others
Shows us that we have a common way to organize knowledge
Commonsense knowledge
Implicit knowledge, knowledge we automatically think is there, knowledge we don't ever have to express/think about.
Hard for computers, not for humans.
Knowledge is interconnected
We use different pieces of knowledge to build our understanding of the world.
Categories
Groups of things that are similiar biologically, functionally or perceptually.
Made out of examplars, which are equal and interchangeable.
No common consesus for how catgorization occurs
Examplars
Individual items in a category
Concepts
Mental representations of objects, ideas and events.
Why items in a category go together.
Idea that includes all that is characteristically assoicated with it.
Might or might not be real.
Commonsense knowledge problem
Ai don't possess the same commonsense knowledge as humans. It all needs to be explicitly encoded in which will take forever.
However, new computers might be able to take in so much information that they might be able to make generalizations across that information and make inferrences.
Classical view of categorization
Categories are defined by set of defining features that are neccesary and sufficient for category membership.
Neccesary: needed for membership.
Sufficient: enough to grand membership
2 reasond why classical view of categorization suck
1. impossivle to have defining feature for most categories.
2. When people categorixe, they don't use the all-or-bnothing strategy. Some exemplars are just better than others and get high typicality ratings.
Typicality ratings
Participants rate how good an exemplar is of a category. Most people agree on these ratings.
Typicality effect
We treat typical items differently than atypical ones.
They are catgorized faster and they are listed first.
Participants were shown a group of letters and they were asked if those were words and words were identified faster if they will primed by a word from the same category as before.
Prototype theory of categorization
We categorize based on the characteristic features of the category.
probablistic theory
other name for prototype theory of categorization
Characteristic features
Features that category members are likely to have but are not required. They store together is a prototype
Prototype
The average of all category members. The one with the most characteristic features. Examplars are compared to them, the the more similar means the more typical they are.
family resemblance
Category members cal all belong to the same category just by sharing a feature with one other family member. The most typical member is the one that shares the most feature with the others.
Problems with prototype theory
Context effects: what is typical depends on the context.
there is no consideration for what we do with atypical members.
Exemplar theory
We store actual examples of items we have encountered in the past. Categorixation occurs by comparing more items to the ones you have in memory and looking for similarities. They are placed into a category with whom they have the most similar features to.
Explains typicality and context effects.
Failuares for both prototypical and exemplar theory
ratings of similarity are often given to things that can be categorixed that are rule based.
No explanation for how we deicde which features are important to compare even though there seems to be a consensus.
Knowledge-based approaches to categorization
We rely on existing category knowledge to explain category membership implicityly. Everything in the past that we know are determining these things.
Physcological essentialism
Category members possess a sundemental essence unique to that category that determins membershio.
Risks:
Sterotyping. The people that support this most are the ones who have higher essentialist beliefs.
When we learn about categories
We try to make meaningful combinations of features to teach about the world.
Categories and reduced cognitive loads.
They are just shortcuts.
Heiarchal model of semantic knowledge
Base-level: people find most natural, informative and distinctive.
Subordinate levels: more specific than base-levels, more informativve less disitnctive.
Superordinate levels: less specific than base-levels, less informative more distinctive.
Knowledge stored as concepts in nodes that are individual units highlighting properties of the concept and other nodes.
Human knowledge and computer
Knowledge must be represented in an information processing system.
Propety inheritance
Subordinate categories have all the properties of superordinare categories.
Cognitive economy
tendency to conserve cognitive effort and resources. Information does not have to be stored at every level in the heiarchy, it is just done so at the highest possible level and it can be inferred for everything else.
Spreading activation of nodes
Enviornment can trigger activation of nodes, and it spreads from one node to another. The further away the node is, the longer it will take to trigger it.
Study asked to verify a question that will tap into different levels. the ones that will tap into levels farthur from eachother were verified much slower.
Spreading activation model of semantic memory
concepts are organized based on semantic similarity to eachother.
Based on semantic priming.
Nodes are connected to eaother based on how semantically similar they are. Those that are more similar are closer and will have more connections.
Activation of a node spreads and activates the closer more connected nodes as fades as it gets into the further concepts.
Explains typicaly effects: typical exemplars are so similar so activation will spread quickly there.
semantic priming
participants were faster to respond to a word with another words that was related rather than unrelated.
Schemata
Plural of schema.
Cogntive stucture representing their knowledge about an item/situation.
Includes everything you know about a topic informed by past experiences.
People with shared experiences will have similar schemas.
Method of repeated reproduction
Simulus shown to participant, they have to recreate it, the more they recreate it, the more it looks the the schemata not the orgional photo because schemata is helpin them fill in the blanks of what they forgot.
Symbol grounding problem
any symbol system can only replace one symbol with another.
Does not effect humans.
Connectivist model
relies on interconnections between nodes but knowledge is stored in the weights of the nodes instead that determine pattern from input produce specific outputs. Network is trained by tuning weights.
Graceful degradation
Harming one part of the network won't cause too much of a deficiency because knowledge is spread otu everywhere.
Black bos problem of neural network models
We can observe resposes from NNM but not the reason behind them.
How do humans ground symbols
With interactions to the enviornment.
Embodied cognition and its 2 theories
The body and our bheaviour influences cognition.
AKA grounded cognition.
Body influences cognition. Body position can influence how well you remember infromation.
Body causes cognition. Sensory experiences are used to understand abstract cognitive thought. Ex: our knowledge of an apple is dependent on modality specific neurons that tell us its taste, looks and feel, etc.
Evidence for body causing cognition
Similar brain activity for physical movement and movement-related thoughts.
Ex: moving tongue would light up the same brain areas as reading "lick".
Unsure if its just related for an actual cause.
Classic theory knowledge bs embodied theory knowledge
Used to be that knowledge is abstract and unrelated to context.
Now it is goal-oriented, flexible, and context-dependent.
Knowlegde pulled is often ones that are relevant to the situation. And knowledge differ based on setting.
Ex: people read about eggs being in a bridge. They are presented with a picture and ased to press a button if the item is in the story. Button is pressed faster when the egg is uncracked because that would fit better with the context.
What are they concerned with in the brain
Processes used, not how it was done individually.
And how brain controls thought
Semantic dementia
Inability to name objects due to deficits with naming ability, language comprehension and ocjiect used. They are losing the knowledge itself, not any processing from the senses.
Asociated with the anterior temporal cortex.
Brain region and knowledge
anterior temporal cortex
It deos not contain all knowledge because damage to it does not always cause loss of semantic memory and it is not active during tasks involving semantic memory
Hub and store model
Abstrack concept information is stored in the central hub in the ATL while context-dependant and modelit0specific information is stored in spokes all around the brain. The knowledge of an apply is in the central hub while everything else like its taste is in its sensory locations.
When ATL ws stimulated, people took slower time to name many objects in pictures. When only specific modalities were stimulated, only corresponding object naming ability was hindered.
Where is semantic knowledge in the brain
Everywhere
What does typicality influence
language, memory, attractiveness.