17 terms

Embodied Cognition


Terms in this set (...)

The Outfielder Problem - Option 1
Option 1: Predict the final location of the ball by computing its trajectory

What do we need?
-initial velocity of the ball
-angle of departure from bat
-height from ground
-wind, etc

-evidence for such a model lacking
-even skilled baseball players can't predict trajectories over that distance
-ball is so small it's nearly impossible to view necessary variables
The Outfielder Problem - Option 2
Option 2: He fixed the game

What do we need?
-money and a motive
-an unrealistically skilled batter
-entirely predictable conditions
The Outfielder Problem - Option 3
Option 3: He beings to run and, while keeping his eye on the ball, continuously updates his movements in a FEEDBACK LOOP b/t action and perception

What do we need?
-can solve this problem to only one variable
-trajectory of the ball w/ respect to the viewer (linear optical trajectory)
The Outfielder Problem: Linear Optical Trajectory
Perception-Action Loop
-outfielder MOVES, varying both a (vertical angle to the ball) and b (angle along the ground)
-so as to make the ball appear to follow a straight line path

-fielder uses visual info to continuously guide his movements
-avoids dependence on an internal model or computation of the trajectory

One variable: does the ball trajectory look straight?
-a simple rule over time
Marr's Level of Analysis
-what problem are we solving?

-how do we solve it?
-general algorithm: should accommodate all possible physical realizations

-how is this physically realized?
-running speed, stride, height, visual acuity different for everyone

-the nature of out implementation is not independent of our algorithm
-Level's of Analysis do not allow for this "loopiness"
Embodied Cognition
-a radical new approach to cognition
-a growing research program is cognitive science that emphasizes the formative role the BODY and ENVIRONMENT play in the development of cognitive processes

Embodied accounts of cognition proposed in many cognitive science subfields:
-developmental psychology, robotics, neuroscience, AI, philosophy of mind, & linguistics

Central tenets:
1) The primacy of goal-directed action occurring in real-time
2) The belief that the form of embodiment (type of body you have) determines the type of cognition
3) The view that cognition is constructive

-all fields share common goal of developing cognitive explanations that capture the manner in which MIND, BODY, & WORLD mutually INTERACT and INFLUENCE one another to PROMOTE an organism's ADAPTIVE SUCCESS
Classic Cognitive Science
-sensory input into the brain
-processing to identify it, remember it, manipulate it, decide what to do
-output in form of a RESPONSE; ACTION; BEHAVIOR
-BRAIN is where all "cognition" occurs

Job of Cognitive Science:
-to figure out how the brain converts sensory system input into motor system output

Sense-Think-Act Cycle
Telephone Poles (Example)***
Off-loading computation load to PERCEPTION-ACTION-CYCLE
-simplifying the task
-are telephone poles the same height?
-which telephone pole is in front?
-see image***
-we use heuristics everyday
-depth perception
Kitten Carousel Experiment**
-kittens kept on dark for 8 weeks (except for one hour per day when kept in kitten carousel)
-2 groups: Passive Kittens & Active Kittens

-both groups received equal amount and type of visual experience, but visual experience (PERCEPTION) of action kitten tied to their behavior (ACTION)

-Active Kittens developed normally
-Passive Kittens had no depth perception

Normal perceptual-motor development depends on active exploration
-or feedback b/t perception, action, and environment
Theoretical Orientation Matters: Environmental Differences
Traditional Cog Sci
-constrain everything but variables of interest to reduce "noise"
-subject passive receiver of stimulus

Embodied Cog Sci
-interaction b/t mind, body, and environment are essential if you want to understand cognition
-"noise" isn't noise at all (it is a constituent of the cognitive process of interest)
-ecological validity (??)
-subject is active receiver of stimulus
Novel Object Recognition in Rats
Common test of recognition memory
-rats spend more time investing new subjects

-the orientation of an animal's snout toward the object (sniffing or touching with snout) while running around the object
-NOT sitting or climbing on it

-findings were greatly influenced by the properties of the object (with rats preferring objects they could climb on to those they could not)
-rats showed preference for objects that have affordances for common rat activities
Theoretical Orientation Matters: Robot Design
Compare traditional AI to traditional Cog Sci
-Input, Computation performed by a set of algorithms, Output

Shakey the Robot (1966)
-first "electronic person"
-navigated through a room with an off-board computer
-planned full movements
-sent those commands to the behaving portion of the robot

-controlled by a SDS-940 computer w/ 64K 24-bit words of memory
-could not learn, could barely navigate, and required an off-board w/ "representation" of the room it was navigating though
Robot Design Today
-GOFAI (Good Old Fashioned AI) Success Stories
-Deep Blue (beat chess grand-master)
-Watson (defeated several former champions of Jeopardy)
-Asimo (brought to you by Honda)
Behavior-Based Robotics
-situated AI
-instead of guiding behavior by symbolic mental representations of the world, SUBSUMPTION ARCHITECTURE couples sensory info to action selection

-decomposes complete behavior into sub-behaviors, organizes them into hierarchy of layers, and each layer implements a specified "perception-action"

-attacks problems of intelligence from significantly different perspective than traditional AI
Turbo-Charged Engine
-engine that re-uses the exhaust it produces
-rather than exhaust exciting engine, it spins a turbine that causes air to be compressed more compactly in cylinders
-explosions in cylinders more powerful

Key: Engine does one thing, but it is not the whole picture
-the process wouldn't work w/o output (exhaust) feeding back into the system

-engine = brain
-pistons, etc = body
-exhaust feeding back = relevant environment
-all of these processes combined should be considered the mind

-the brain is an essential component of the mind, but the mind also includes the BODY and directly coupled ENVIRONMENTAL FEEDBACK

-the mind extends beyond the brain AND body
The Extended Mind Thesis
-the mind extends beyond the boundaries of the skull and into our environment

-tools, instruments, and other aspects of our environment that we organize can in some cases be considered a part of our mind

-the MIND, BODY, & ENVIRONMENT act as a coupled dynamical system and it is impossible to separate them from one another in a meaningful way
Criticisms - The Extended Mind Thesis
-when focusing on cognition, the thesis confuses claims about what is constitutive about the concept of cognition w/ claim about casual influences on cognition (casual-constitutional fallacy)

-Ex: why did the pencil think that 2 + 2 = 4? (b/c it was coupled to the mathematician)

Andy Clark's Response
-coupling is important for cognition (but not necessarily sufficient --> coupling must play functional role in cognition)