PSY270 Notes.pdf

Test info ● Need to know names and accomplishments Important stuff from Lec 1 ● Historical origins of cog psych ○ Hasan ibn al-haytham’s book of optics, he started the first experiments in vision, he is start of what we think of as experimental cog psych ■ Ran experiments on self, through deduction found how vision works, hundreds of years before fechner comes to similar conclusions ○ Fechner’s element of psychophysics ■ Relates physical stimuli to perceptions, he sort of consolidates hasan’s ideas ○ Donders’ reaction time ■ Detection vs decision ■ Difference in reaction time between detection and decision, and what this difference tells us about internal processes in the mind, depending on the time diff ○ Wundt’s introspective approach ■ Problems w introspection ● Differences in language and vocab + differences in subjective experiences, all of which has to be filtered through the system of language and expression which ends up being tricky ○ Titchner’s structuralism ■ Goal was to turn introspection into set of formalized laws for understanding inner workings of perceptual processes, based on introspective observation - though this didn’t go too well ○ Watson’s Behaviourism ■ What’s the overriding goal of behaviourism? ○ Pavlov’s classical conditioning & skinner and operant conditioning ■ Similarities and differences, what do they tell us about learning and cognition? Refer to the examples we talked about w the cats ○ Behaviourism and childrearing ■ Why might a focus on behaviourism be bad? ● It ignores internal mental states and any feelings of child or parent ● Cognitive revolution ○ Moore’s cognitive psych ■ It predated the cog revolution by decades ○ Tolman’s rate mazes ■ What do tolman’s rats tell us about contents of the black box? ■ Rat behaviour couldn’t be accounted for in strict behaviourist POV, thus we see him more as a cog psych because of this ○ Cherry’s dichotic listening paradigm ■ 2 inputs, one correct output, why? ○ Computational conception of cognition - cognitive processes as inputs, internal processes, and outputs like computers ○ What was chomsky’s main critique of skinner? ■ What does it say about limits of behaviourist theory? ■ That we can’t expect language acquisition to follow action-reward framework, since children could generate sentences that they had no reason or reward for generating it ● Ex. having child say to parent “i hate you”, why would they say that even though they wouldn’t be rewarded for it or anything ○ Process models & structural models ■ Differences, what can each tell us? ○ Information as core of cognitive psych ■ How is this diff from earlier schools of thought? ○ What happened to behaviourism? ● Cognitive neuroscience ○ Differences between cog neurosci (more focused on anatomical reasons as to why behaviours and such happen) vs cog psych (more focused on behaviour and how cog processes operate, less focused on anatomical causes for behaviours) ○ What can we learn from single neurons? ■ Neuronal selectivity, electrophysiology, etc. - how brain deals w visual information ○ What can’t we learn from single neurons and why? ■ 90 bil neurons makes it hard to learn and study each individual neuron and connecting it to entire brain ○ Neuronal selectivity ■ Hubel & wiesel, and gross ○ Coding complex information in the brain ■ Types - selective, population, sparse ○ neural/cortical representation ● Non-invasive methods in cog neurosci ○ fMRI Important stuff from Lec 2 ● Scene perception ○ Scene gist demo told us that we can get a general idea of a scene within the blink of an eye, even faster if its a static scene (the people on toilets in restaurant pic) ■ A lot of the time people wouldn’t notice the toilets that they were sitting on, which suggests that acquiring visual info isn’t just about where we’re looking at given point, its about entire visual field at some time ● How we acquire info ○ Focus of gaze (fovea) vs peripheral vision ■ Fovea - area of highest resolution on retina ■ Peripheral vision - the rest of the visual field ■ Just focusing on where you’re looking, where your gaze is, is not helpful since it would be ignoring 99% of retina, which is maladaptive ● Visual field - extent of worl you can see w your 2 eyes ○ Peripheral vision - what we see with our 2 eyes beyond the fovea ○ Can we see colour in periphery? Yes, and it is not blurry either ● How do we build up our perception of visual world? ○ we get a general gist, like scene gist, which then guides our eye movements to help fill in the blanks ● How do we recognize objects? ○ To be able to perceive something we need identification (what object is) and localization (where it is in world) ○ This comes from object recognition, and this has to do with image segmentation - what part of what belongs to which object? ■ Template theory - have templates of objects in world in our memory, and we compare what we see to that template ● Doesn’t work well since things may be in diff orientations, fonts, ect., so we’d need wayyyy too many templates ● Objects can be occluded, so they would also need a new template ● Templates are just too expensive in terms of how much space they’d take up ■ Feature analysis ● Break image down into constituent features in parallel across visual field ○ Line termination - where line ends ○ Orientation ○ Size ○ Curvature ○ Colour ● Then these features are used to build up and make an object ● Evidence for this is neuronal selectivity - simple cells respond best to oriented lines at specific location on retina, complex cells to moving stimuli regardless of orientation but prefer motion direction, and hypercomplex cells that respond to combination of specific features like orientation and motion direction ○ Though theory of neuronal selectivity as we know is not very accurate since this would be extremely inefficient ○ Evidence against neuronal selectivity is the macaque thing, since as we get further and further from a hand image, we got less and less activity rather than only getting activity when hand image shown and nothing else ○ Why might we want feature-based visual system? ■ Because its infinitely adaptable, whereas template is not, also hard to differentiate between existing template and potential new one ○ Why is object identification hard? ■ Variability in size, shapes, occlusions, etc. ○ Why does perceptual consistency matter? ■ It is the reason we are able to identify objects in various conditions, like diff lighting, orientation, etc. ○ What do aftereffects show us about representation? ■ Since they are cortically based, it proves neurons code for features since we get habituation, which is underlying cause of after effect without feature coding ● Low-level (features) to high-level (objects) ○ Features like colour, orientation, etc. ○ Brain starts coding at low level features and builds up into objects ● Disorders of object perception ○ Visual agnosia - can’t recognize objects, but can navigate and reach for them, use colours and stuff for cues ○ Prosopagnosia - inability to recognize faces, can use hair, voice, etc. to identify people, these people DO know about faces still, they can judge emotion and stuff based on face (ensemble perception) ○ Balints’ syndrome - can identify object but can’t interact with it (reach for it) ● Fusiform face area (FFA) and expertise or specialization ○ FFA is area of brain that selectively reponse to faces, first of object selection areas ○ Is FFA specialized? (only works for faces and nothing else? Or is it an expertise area (does faces and also anything else that we’re naturally really good at discriminating) greebles ● How computer object recognition doesn’t work ○ Computer algorithms not good at object identification for similar reasons to why object identification is hard for us, occlusions, segmentation, etc. ○ Algorithmic biases cause inaccurate recognitions, particularly for faces ■ Computer vision system trained 2 photos of world and embody developers’ biases ■ Don’t perform equally for everyone (like diff races) ■ Significant because algorithmic biases are similar problems to what humans face in some ways ● Connecting computer vision to human vision ○ Similarities between how we fail and how computers fail, and what we ought to do about them Important stuff from lec 3 ● Perceptual biases and assumption ○ Why does visual system make assumptions about the world? ■ World is largely predictable, if we know what parts of world are predictable then it makes processing visual world easier, which is optimal for the moving world ○ Likelihood principle ■ Perception of world biased by what we think and our experiences of world which help to explain what happens ■ Basically making inferences of world based on personal experiences or schema ○ Gestalt principles ■ Proximity (what’s close by), similarity (what’s similar), closure (perceive objects as complete, not sets of lines or assuming that occluded things still exist), good continuation (following overall form), common fate (moving as group, like flock of birds), pragnanz (perceive complex shapes as one object, like olympic rings) ■ Ways to understand some of the shortcuts visual system takes in terms of what, when and how objects group together ○ What do impossible figures tell us? ■ These principles don’t always work, they can be changed ● Perceptual assumptions ○ Where do they come from? What’s their impact on our perception of the world? ■ Light from above can cause oblique effect which messed w perception of world ● comes from our exposure to visual environment like sun from above ● Are faces special? Face inversion effect/thatcher effect ○ We expect faces to be upright ○ Thatcher effect is turning eyes and mouth upside down after turning picture upside down we still think it looks normal ○ Relates back to FFA, ● Assumptions abo