C03 Basic Concepts in CogSys

• predecessor to cognitive science

• assumptions:

  • brain is the organ of the mind

  • human mental poser is composed of differnet independent faculties located at specific regions in the brain

  • size of a brains region indicates how much the corresponding faculty contributes to a persons character

  • there is a close correspondence between the sizes of brain reginos and the shape of the skull

NOT A SCIENCE!!!

• brain as the center of thought

• modular organization of the brain into anatomical regions

• functional specialization of individual brain regions

• different (52) regions in the cerebral cortex

• based on cytoarchitectinic properties

  • -> cellular composition under microscope…

• cut brain into slices of 20 micrometer (7400 slices)

• stain every section to identify the cell bodies

• digitize every slice/section with resolution of 13k by 11k pixels

• process each section to remove atifacts from teh scanned images

• align slices for a 3D reconstruction

• quantify brain structure

  
  

• magnetic resonance imaging (MRI) (kernspin)

• diffusion tensor imaging (DTI)

-> both based n nuclear magnetic resonance

• non-invasive method used for imaging anatomic structue of tissue in vivo

• active sensing method based on magnetic properties of hydrogen atoms that dissipate energy differently depending on surrounding tissue

• extension of MRI

• able to measure diffusion movement of water molecules

• -> can thereby detect nerve fibers

• study the function of the brain

• instead of its structure

• electroenchepalography (EEG) and magnetoencephalography (MEG)

• positron emmision tomography (PET) and single-photon emmision computer tomography (SPECT)

• funcitonal MRI (fMRI)

• enable non-invasive monitoring of brain activity at high temporal resolution

• eeg: measures electrical potentials resulting from neural activity via eectrodes placed on skill

• MEG: detects corresponding magnetic field

• nuclear imaging techniques

• -> require medication with radiotraces

• -> brain activity monitored implicitely by observing regional cerebral blood flow

• extension of MRI

• -> visualizes changes of blood oxygen level

• caused by brain activity

• elicited through external stimuli or cognitive processing at high spatial resolution

• augments functional brain imaging

• -> haptic fMRI augments fMRI with haptic input device

• during fMRI scan, person looks at virtual objects projected on a screen

• haptic input device enables interactoin with the virtual objects

• barin activity is monitored via fMRI

• engineering

  • AI

  • ML

  • Computer Science

  • robotics and control

  • chip design

  • subtopics of mathematics

• Natural science

  • neurosciece

  • neuobiology

  • developmental sciences

  • biomechancis

  • neural engineering

  • neuroimaging

• humanities

  • cognitive psychology

  • linguistics

  • epistemology

  • philospohgy

  • ethics

• computaitonal models

• bio-inspired models

• hybrid models

• purely computational

• implement cognitive funcitons solely based on functinal view of the system

• without any reference to biology

• implement cognitive functions by

• replicating known or hypothesized mechanisms of cignitive processing

• from biological organisms

• distinction between two approaches that try to explain the evolution of cognitive behavior

• from different perspectives but complementary

• cognition specific to an agent and its environment

• -> in the course of evolution

• -> living creatures have developed cognitive skills enabling them to act and survive in their native environments

• ultimate explanation:

  • why a behavior exists

• proximate explanation:

  • how it works

• the successfull development of a cognitive system requires

• that the implementaion of cognitive functions wiht ultimate explanations

• match the systems specific working environment

• -> why does a function exist? => must fit the environemt (makes only sense if the reason it exists aligns with what we can percieve…)

• computational theory

• representation and algorithms

• hardware implementation

=> depending on model paradigm, levels are loosly or thighly coupled

• high-level description of the system

• that states the goal of teh computation

• and the logic of the strategy by which it is carried out

• realization of the computational theory w.r.t.

• input, output, representatnios and algorithms

• that perform the reqired transformaitons

(datenstrukturen und tatsächliche algorithmen…)

• physical implementation of the representations and algorithms

• cognitivist cognitive modelling based on hypothesis

• that congitinon is a form of computation

• -> cognitive functions modelled as working computer programs

• here, cognition is a continuous self-organizing process

• that is driven by the interaction between agent and its environment

• cenrtral hypothesis:

  • cognition is achieved by oprations on symbols htat have a one-to-one correspondence to real world aciotns and objects

  • -> external infromation from environemnt gathered and filtered by perceptual process

  • -> generates set of symbols

  • -> processed in “brain”

  • -> agian output set of symbols

• all cognitive processes can be explained mechanistically

• through some form of computation

• addresses certain questions that challenge the simplistic view of cognitivism

• -> what is the basic nature of mental phenomena (such as thought, feeling, perception, consciousness, and sensory experience)?

• -> can these phenomena be described solely by chemical / physical processes or can they exist independently of them (i.e. inside of a computer)?

• -> how to explain subjective conscious experiences (qualia)?

• how is the mental world related to the physical world?

• substance dualism (also cartesian dualism)

• monism (also physicalism)

• from rene descartes

• -> postulates that mond and body are two kinds of different substances

• => separation makes soul immortal (contrary to mortal body) and enables free will

• there is only one substance

  • -> meaning mental states are physical states

• => would imply that individuals that share a mental property also share a corresponding physical propertay

• -> therefore nonreductive physicalism postulates that mental properties are not identical physical ones even though there is only one substance

• theroetical basis of cognitivism

• -> holds that the mind is a digital computer

• i.e

  • -> a discrete-state device

  • that stores symbolic representations and manipulates them accordig to syntactic rules

  • that thoughs are mental represenattions (more spefcifically symbolic representations in a language of thought)

  • and that mental processes are causal sequences driven by the syntactic but not semantic properties of symbols

• syntax: structure of langauge

• semantics: meaning of words

• for the emergence of general intelligence

• physical pattenrs called symbols

• -> which can form expressions that are

  • created

  • modified

  • copied

  • destroyed

• by processes

• iff (<=>) the system can affect it or if it is affected by it

• <=> it designates a process

• that can be carried out by the system

• resucrively

• -> symbols can designate symbols

• processes can produce other processes

• by searching

• chess

  • -> symbols are pieces

  • -> processes are legal chess moves (indicate by i.e. piece, source, dest,…)

  • expressions are the positions of all the pieces on the board…

• PSS has the necessary and sufficient means

• for general intelligent action

• => every cognitive system is a PSS

• the solutions to problems are represented as symbol structures

• -> PSS exercises its intelligence in problem solving by search

• => progressively generating and modifygin symbol structures until it produces a solution structure…

=> cognition is reflected in an intelligent search strategy

-> that finds a solution without exhaustive brute-force search

• an agent acts in a sensible and purposeful way to achieve its goals

• -> rational agent will always use as much of its knowledge as possible to guide its behavior

• optimal rationality

• bounded rationality

• rationality exhibited by decision makers of limited abilities

• in reality: usually no complete information or all actions possible -> thus no real optimality feasible thus in real world -> bounded optimal thus bounded ratinoality

• success criterion

• agents prior knowledge of the environment

• actions the agent can perform

• past percepts

• rise of a system

• that cannot be predicted or explained

• from antecedent (previous) conditions

=> show some behavior that was not programmed into them…

• on the continuous self-organization of the system

• build their own model of reality through

• -> interaction with the environment

• and filtering of relevant information

• knowledge and represenation

• -> depend on its history of interaction (experience)

• in the form of additional anticipatory skills

• embodiment (-> to interact wiht env.)

• and

• development (-> to change…)

• model cognition based on networks of

• simple interconnected computational units

• -> cognitive processing in these networks is

  • distributed

  • parallel

  • based on statsitical properties instead of formal rules

    
    

• neural networks

• simple elements called units

• equatinos that determine an activation value for each unit at each point in time

• weighted connections between units which permit the activity of one unit to infuence the ativitay of other units

• optionally:

  • learning rules which change the networks behavior by changing the weights of its connectoins

• in the connection weights of the system

• develops its own understanding of the world

• through its interaction with its environment

• phylogeny

• ontogeny

• interactoin between the system and its environment is

• structurally determined

• by the innate embodied physical and cocnitive capabilities

• => i.e. human: has two legs, eyes,brain,… -> body of humans determine its interaciotn with the environment…

• system is structurally coupled with the environment

• and makes sense of it based on

  • its actions

  • and the perception of resulting actions of the envoironment

• => thorugh semse-making, system generates its own specific epistemology

  • -> that incerases its cognitive capacity by capturing the regularities of interactoin (exploration)

• => gradual incerase in cognitive capacity results in development

• phyolgeny: comparabe to capabilities of species that determine their interacton with the environment

• ontogeny: indivdual history -> i.e. individual bird learns and buiilds its own knwoledge

-> knowledge and capablities acquired by birth

vs acquired through interaciton and experiecne…

• based on time-dependent differential equations

• that allos for a compact representation of complex system behavior

• and the applicaiton of analysis methods from dynamical systems theory (attractors, bufircators, stability,.etc.)

• dissipation

  • number of reachable states reduces over time

• non-qeuilibrium system

  • stable funciton requires an external “energy” supply

• non-linearity

  • complex behavior can emerge from a small set of state parameters

• collective variables

  • system is represented by a small set of state variables

• can (but not neccesarily do) represent entities by

• means of system states or trajectories

• -> these representations are transient and context-dependent

• cognitivist:

  • rational goal-achievement

• emergent:

  • self-maintenance and self-development

cognitivist:

• syntactic manipulation of symbols

emergent:

• concurrent self-organizaiton of a network

cognitivist:

• patterns of symbols and tokens

emergent:

• global system states

cognitivist:

• atemporal

emergent:

• synchronous real-time entrainment

cognitivist:

• no role implied: funcitonalits

emergent:

• direct constitutive role: non-funcitonalits

cognitivist:

• abstract symbol representations

emergent:

• pertubation by the environment

cognitivist:

• causal result of a symbol manupulation

emergnet:

• pertubation by the system

cognitivist:

• learn new knowledge

emergent:

• develop new dynamics

cognitivist:

• criteria for goal selection

emergent:

• increase space of interaction