Lecture 2- Levels of Analysis and Explanation Flashcards
Perception as an active process
start with an agent with some developmental history: how they develop a knowledge base
what if it wasn’t a human agent? what kind of agent: different types of sensors and uses that info differently
human sensory systems can be best thought of as…
…actively engaging sensory stimuli
it’s an interaction with the environment, focusing your eyes on something: they’re always moving, you’re always changing your perspective a bit
Environmental Stimulus
distal stimulus:
- ex: tree, object that you’re trying to detect
- it’s always at a distance
- try to orient towards it
some input will come from the d.s.
first step where things can go wrong: light going through the atmosphere, obscured by smoke
Receptor Processes
- stimulus (light energy) transduced into electrical energy
- transduction: changing in the type of energy
Neural Processing
happens at the local level and further on through the primary sensory cortex
Consciously Perceive (detect) that there is a thing
detected the distal stimulus
may not know the nature of the thing, can’t catagorize
if we have enough we go from perceptoin to recognition (OH! it’s a chair)
with or without recognition you’re still going to act
Perception vs. Recognition
you can have one without the other
example: indivual can describe an object in detail without being able to put it in a specific class: VISUAL FORM AGNOSIA where common objects cannot be recognize
Recognizing without Perceiving: Blind site: where people believe they are blind but under the right testing conditions they can recognize objects without consciously percieving it: different functions dissociated from one another: suggests that there are diff parts of the brain responsible for those functions (diff parts of the brain for perceptions and recognition)
Visual form agnosia
you recognize someone but you can’t place how you know them
perception without recognition
top-down processing
history biases you to interpret results in certain ways
experience literally makes you look at things differently
Ex: twins: one twin is an entomologist and one hates bugs: both detected and both attempted recognition but both had different sensory experiences: interact with the environmental stimulus differently
bottom up processing
receptors excited based on input
with always interact with the top down processing = never one or the other, it’s both
How do we measure effects at each state of the perceptual process? Where do we start?
define terms
framework
perception doesn’t happen at just one level, involves a lot, so our framework has to account for that =
Measurment
assigning numbers to observable entities or events
problem: how to observe what’s going on in someone’s head? you can’t see it?
Sensory qualia
the technical term for the experience in someone’s head, associated with some particular event
experience of a certain color, hearing a sound, detecting a taste
are all subjective
can’t be directly observed
Theoretical constructs
we believe that unobservable entities (color perception) exist in some sense, but can’t be directly observed
operational defintion
created for the internal theoretical construct
assigns observable behavior as a proxy
when we see it we can make an inference that come perception is being experienced
fundamental fact of science
when i measure that thing it tells me it’s there
precision is theoretically limiting
what if your operational defintion is wrong?
- what if mri isn’t a good indicator of brain activity? you’ve been thinking about things the wrong way
operational def.
observable act
pointing, speaking
theoretical constructs
recognition and perception
what is vision for? what is hearing for?
model of processing can’t help with that, assumes you know the problem
doesn’t tell you the problem itself
Marr’s Level’s of Analysis *****
computational level
algorithmic level
implementation level
complementary and interact with each other
david marr
first computational neuroscientist
computer scientist interested in computer vision
hippie
book: Vision
asked some basic questions trying to come up with computer vision systems
bigger question: what is the system trying to solve? what is vision doing for us?
computational level
basic question: what problems does the system solve and why?
what does vision do? danger: color, motion, edges
algorithmic level
if this system is trying to solve some problem (danger-threat level), how does it do that? what representations does it use?
goes through a serious of steps: looks through features that indicate threat
implementation level
what’s happing in the brain?
is it neural tissue? in a computer system? how is being physically instantiated? how does it act? does it go through some steps on the algorithmic level to solve some problem on the computational level?
