- 1) Assumes digital computational theory of mind
- 2) Uses physical symbol systems
- -Thought it result of symbol manipulation
- 1) Physical Kinds: Things are defined by what they are made of (mind=brain)
- 2) Functional Kinds: Things are defined by their function (doughnut=doorstop)
Digital Computational Theory
- 1) Cognitive states are mental representations
- 2) Cognitive processes are computational operations performed on mental representations
- 3) Computations are digital
- 4) Use algorithms and programs
A sequence of instructions and effective procedures. Independent of programming language
- Implement algorithms.
- Written in specific programming languages
- Algorithms are ideal, programs are specific
Thought experiment asking if a machine could fool a human. Do we count this as thought? Does this mean a machine is intelligent? No.
Classical AI vs. Connectionism
- Classical AI relied on physical symbol systems hypothesis. Results were slow and only worked in very small worlds
- Connectionism uses a neural net (ANN) to show a pattern of activation
Artificial Neural Network
- Pattern of activation, like real neural network
- Mathematical model
Components of ANNs
- Nodes: Have activation values, thresholds, can fire if input exceeds threshold. Like Cell Bodies
- Links: Connections between nodes. Have Weights, or numbers associated with the link.
- Weights: Can be anywhere from +1 to -1. Closer to |1|=heavier weight.
- Sum of the activation values ties the connection weights.
I don't know yet
Lesioning and ANN
Cutting out certain links in an ANN to see if network performs like person with a brain injury. Can hypothesize relationship between organization of network and the organization of neural system.
- 1) Mechanical creature that can function with no operator
- 2) Can adapt to changing environment
- 3) Can function even if some parts break
- 4) Can move and change things
Primitives of robotic system
- Sense: Take information from environment, convert to information system, either from sensor of knowledge base
- Plan: Use information and generate set of tasks to execute. Analogue to human cognition.
- Act: Commands for action
- Remove the Plan stage
- Subsumption Architecture
- Interaction between two layers (avoid objects, wander around) looks like intelligent behavior
- No central planning
Gap in graph of similarity to familiarity in robots
Cognition is not confined to a single brain. It is spread across an individual's mind, body, the environment, and other people.
Pragmatic vs. Epistemic actions
- Pragmatic: Change the world in order to have things in a certain position
- Epistemic: Change the world to make mental tasks easier. To aid with cognition
Similarities and Differences in Animal and Human Cognition
- How we think is intimately tied to our human bodies and interacting with the world with these bodies.
- Support: Affordances, Simulation
How a person thinks an object can be used or interacted with
- 1) Evolutionary: Natural Selection
- 2) Epigenetic: from egg to embryo
- 3) Developmental: As we grow
- 4) Online: Self-organization of multiple organisms
- Recognition of Intention: Understanding that others have mental states/goals
- Joint Attention: Mutual focus on an object AND knowing the focus is mutual
- Theory of Mind: Understanding that others may have different mental states than your own
- Communicative Intention: You can change someone else's mental state via communicative action
Show a stimulus to infant until bored, then show another stimulus and gauge reaction
- May cause empathy.
- We understand others because we simulate their actions or emotions.
How humans acquire knowledge over the course of a lifetime
Change in the gray matter density in the brain, number of connections. Connections that are not used are removed.
- Assimilation: New experiences are linked to existing schema
- Accommodation: Schema is modified to include new experiences
Piagetian stages of cognitive development
- Sensorimotor: beginning of intentional action and object permanence
- Preoperational: Lean to use language and egocentric thinking
- Concrete Operational: Have conservation, classify objects by more than one property
- Formal Operations: Abstract reasoning, hypothesis formation
You can move things around while still having the same amount of stuff
Things continue to exist even though they are not seen
Nativist Approach to Cognitive Science
- Language: Critical Period. Pro-Genie, Con-Don't know existing damage
Deception in Children
Before 4, children can deceive to change someone's action, after 4 can deceive to change mental state
- General Approach
Mental representation of a problem, including initial, final, and intermediate states
Genreal Problem Solver
- Used mean-end analysis
- Worked with logic problems
- Can't solve real world problems
Pattern of Similarity
First Order vs. Higher Relations
- First Order: direct similarities
- Higher order: relations among relations. (x is like y in the same way a is like b)
- Accessing: find items in memory
- Mapping: transfer information from source to target
- Inferencing: Generate new inferences
- Adapting/evaluating: choosing which information to map and which inferences not to draw
When tested in experimental situation, people only transfer when superficial properties are preserved.
Human Factors Engineering
Field focusing on designing objects so that they are compatible with human body and human cognition.
Principles of Good Design
The study of human information processing in design
Design vs. Problem solving
Design is creating a plan. Problem solving is navigating through problem space to find a solution.
Cognitive Science of Art