Hoorcolleg 1
- Cognitieve processen: als je iets waarneemt en je daar vervolgens ‘iets mee doet’
- Goede wetenschappelijke studie:
- Wetenschappelijke methode: empirische cyclus gevolgd
- Repliceerbaar: helder en compleet gecommuniceerd
Empirische cyclus
1. Observatie (ik zie alleen witte zwanen)
2. Inductie (theorievorming) (alle zwanen zijn wit)
3. Deductie (hypothesevorming)) (alle zwanen zijn wit)
4. Toetsen (overal kijken of alle zwanen wit zijn)
5. Evalueren (niet alle zwanen zijn wit)
Anders geformuleerd (stappenplan):
1. Probleemanalyse
- Observatie (kan alledaags zijn, bijv. alle zwanen zijn wit)
- Identificeren ‘gat in de kennis’ (iets is nog niet onderzocht, bijv. kun je
meerdere objecten tegelijk waarnemen op de tast?)
- Redenen voor replicatie (bijv. tegen elkaar sprekende bevindingen of
rammelende methoden, ‘studie A concludeert dat bewustzijn en aandacht
hetzelfde is, studie B laat zien dat het verschillende processen zijn)
2. Onderzoeksplan
- Onderzoeksvraag (een goede onderzoeksvraag volgt uit de theoretische
achtergrond, bijv. zijn alle zwanen wit?)
- Hypothese (falsificatie: weerleggen; verificatie: bevestigen)
- Checklist:
- Is de hypothese gebaseerd op bevindingen uit eerdere studies
(theorie)?
- Kan er minstens één duidelijke voorspelling worden afgeleid uit
de hypothese?
- Zijn de voorspellingen die uit de hypothese kunnen worden
afgeleid ook daadwerkelijk testbaar in een wetenschappelijk
experiment?
- Heeft de voorspelling zowel een afhankelijke (wat je meet) als
onafhankelijke variabele (wat je manipuleert)?
- Hoe kunnen we de vraag beantwoorden?
- Check alle zwanen op de wereld, etc.
- Voorspellingen (als de hypothese waar is, dan komt er … uit het experiment.
Bijv.: geen enkele zwaan heeft een andere kleur dan wit).
3. Uitvoering
- Methoden
- Beschrijvend onderzoek (beschrijft alleen observaties, van een
populatie of een fenomeen)
- Correlationeel onderzoek (onderzoekt de samenhang tussen twee
soorten data) (let op: correlatie bewijst geen causaliteit; vaak sprake
van een spurieus verband (schijnverband) (meer inbraken bij meer
ijsverkoop -> komt door mooi weer))
, - Experimenteel onderzoek (directe manier om een hypothese te testen
over een oorzakelijk verband) (let op: er kan een contaminerende
variabele zijn die covarieert en die je effect/verband ook zou kunnen
verklaren (meer alcohol = grotere kater, maar leeftijd telt hierin ook
mee)
- Dataverzameling
4. Resultaten
- Verwerken
- Analyseren
- Visualiseren
5. Conclusies
- Interpretatie resultaten
- Hypothese aannemen of verwerpen (alle zwanen zijn wit -> verworpen als je
bijv. een zwarte zwaan hebt gezien. Aangenomen als je alleen witte zwanen
hebt gezien)
- Antwoord op onderzoeksvraag
- Vergelijking met bestaande kennis en literatuur
6. Rapporteren
- Delen van kennis (bijv. presentatie, poster etc.)
- Wetenschappelijk artikel - zandlopermodel
- Inleiding
- Wat is het onderwerp?
- Relevantie
- Wat is er al bekend?
- Inleiding
- Wat is er nog niet bekend?
- Onderzoeksvraag
- Hypothese en onderbouwing daarvan
- Opzet experiment (kort)
- Verwachte resultaten
- Methode
- Uitvoering
- Methoden
- Gedetailleerde beschrijving gebruikte methoden
- Repliceerbaarheid
- Dataverzameling
- Resultaten
- Resultaten (zonder interpretatie of conclusie)
- Grafieken/tabellen met complete ‘Figure Caption’
- Discussie
- Conclusies van experimenten
- Hypothesen: aannemen of verwerpen?
- Sluiten resultaten aan bij literatuur?
- Wat betekenen de resultaten en conclusies?
- Geven de conclusies nieuwe inzichten en ideeën?
- samen met bestaande literatuur?
- voor evt. vervolgonderzoek?
, - Vaak wordt in wetenschappelijke artikelen de empirische cyclus meerdere
keren doorlopen met meerdere experimenten, waarna in een algemene
discussie aan het eind beide experimenten nogmaals uitgebreid worden
besproken.
Methoden van cognitieonderzoek (zie ook kopje Uitvoering, bij 3.)
Mens/omgeving: genen -> cellen -> systeem -> gedrag
- Gedrag: cognitie; o.a. aandacht, leren, taal, beslissen.
1. Systeem meten - hersenactiviteit meten d.m.v. bijv. MRI-scans, EEG, etc. Hiermee
kunnen we bijv. vinden dat hersengebied X is betrokken bij aandachtstaken.
2. Modellen maken - cognitieve modellen bouwen. Computermodel: gedrag simuleren,
zodat we het kunnen begrijpen.
3. Experimenten - proefpersonen iets laten doen waarbij je iets kunt meten. Dan
verander je iets, en meet je het weer.
- Onderzoek met een gecontroleerde verandering/manipulatie (onafhankelijke
variabele)
- De manipulatie moet effect hebben op het proces dat je wilt onderzoeken
(afhankelijke variabelen)
3a. Meetmaten: experimenten binnen de cognitiewetenschap gebruiken vaak de
maten ‘tijd’ (reactietijd) en ‘aantal goed/fout’ (nauwkeurigheid). Het effect van de
manipulatie op de meetmaat is het effect van de onafhankelijke variabele op de
afhankelijke variabele.
- Donders, 1969: Reactietijd kan ons iets vertellen over het onderliggende
proces
- Proces:
- External world -> sensory encoding (iets zien met je oog) -> perceptie
-> function (detect/recognize/identify/choose) -> response (yes-no/cat-
dog etc.)
,Hoofdstuk 1
1.1 A brave new world
- What makes mind so difficult to study is that it is not something we can easily
observe, measure, or manipulate. In addition, the mind is the most complex entity in
the known universe.
- The human brain is estimated to contain ten billion to one hundred billion
individual nerve cells or neurons, which each can have as many as ten
thousand connections to other neurons.
1.2 What is cognitive science?
- The scientific interdisciplinary study of the mind. Primary methodology: scientific
method.
- Combination of philosophy, psychology, linguistics, AI etc.
- Theoretical perspective on the mind:
- Centers on the idea of computation (information processing)
- Information is ‘input’ into our minds through perception;
- Stored in our memories and processed in the form of thoughts;
- Which can then serve as the basis of ‘outputs’, such as language or
physical behaviour
1.3 Representation
- Four categories of representation:
- A concept: stands for a single entity or a group of entities (e.g. single words,
‘apple’)
- A proposition: a statement about the world that can be illustrated with
sentences (e.g. ‘Mary has black hair’; a proposition made up of concepts)
- A rule: can specify the relationships between propositions (e.g. ‘If it’s raining, I
will bring my umbrella’; makes the second proposition contingent on the first)
- An analogy: helps us to make comparisons between two similar situations.
- Four crucial aspects of any representation:
- A ‘representation bearer’ such as a human or a computer must realize a
representation
- A representation must have content - meaning it stands for one or more
objects
- The thing(s) in the external world that a representation stands for are
called ‘referents’
- A representation must be ‘grounded’; there must be a way in which the
representation and its referent come to be related
- A representation must be interpretable by some interpreter, either the
representation bearer himself or somebody else
- The fact that a representation stands for something means it is symbolic.
- Mental representations can stand for many different types of things and are by no
means limited to simple conceptual ideas.
- Human mental representations, especially linguistic ones, are said to be
semantic, which is to say they have meaning.
- Mental states and events are intentional: they are directed upon an object.
- Intentionality is considered to have at least two properties:
- Isomorphism (similarity of structure between a representation
and its referent
, - Appropriate causal relation: an intentional representation must
be triggered by its referent or things related to it, which causes
behaviors or actions that are somehow related to the referent.
- Digital representations:
- Also known as symbolic representations
- Information is coded in a discrete way with set values.
- The symbols can be operated on a more general set of processes than
analog structures
- Analog representations:
- Represent information in a continuous way
- Can theoretically take on any value not limited by resolution. Resolution =
the amount of detail contained in an analog representation.
- Visual images are the best example of mental analog representations.
- The Dual-Coding Hypothesis (Paivio, 1971)
- = the use of both digital/symbolic and image representations collectively
- Propositional Representations
- According to the propositional hypothesis, mental representations take the
form of abstract sentence-like structures.
- Propositions are good at capturing the relationships between concepts (‘Mary
looked at John’)
- Believed to lie in a deep format that is neither visual nor verbal; logical
relationship among constituent elements, denoted by a predicate calculus.
- A predicate calculus is a general system of logic that accurately
expresses a large variety of assertions and modes of reasoning. E.g.
predicate calculus of ‘Mary looked at John’: (relationship between
elements ((subject element), (object element)). Mary = subject, John =
object, looking = relationship.
1.4 Computation
- Categories of mental operations can be defined by the type of operation that is
performed or by the type of information acted upon. A list of these operations would
include sensation, perception, attention, memory, language etc.
- The Tri-Level Hypothesis
- According to the tri-level hypothesis, mental or artificial information-
processing events can be evaluated on at least three different levels (Marr,
1982).
- The highest or most abstract level is the computational level, at which
one is concerned with two tasks: a clear specification of what the
problem is and defining the purpose or reason for the process.
- Stepping down one level of abstraction, we can next inquire about the
actual way in which an information process is carried out. To do this,
we need an algorithm: a formal procedure or system that acts on
informational representations. Formal = well-defined: we know exactly
what occurs at each step of an algorithm and how a particular step
changes the information being acted on.
- The algorithmic level is like software, because software
contains instructions for the processing of data.
, - Needed, because the algorithmic level tells us how a particular
system performs a computation. Not all computational systems
solve a problem in the same way. Also gives us insights on
how these systems might compute solutions to other novel
problems that we might not understand.
- The most specific and concrete type of description is formulated at the
implementational level. Here we ask: What is the information
processor made of? What types of physical or material changes
underlie changes in the processing of the information?
- This level is sometimes referred to as the ‘hardware level’.
The classical and connectionist views of computation
- A computer is a formal symbol manipulator. Formal = syntactic or rule-governed;
symbol = form of representation, can assume a wide variety of forms. Manipulation
implies that computation is an active, embodied process that takes place over time.
- The network approach to computation differs from the classical formal systems
approach in cognitive science in several ways:
- Classical: Knowledge is presented locally in the form of symbols
- Connectionist: Knowledge is represented as a pattern of activation or weights
that is distributed throughout a network
- Processing style differs as well:
- Classical: has processing occurring in discrete stages
- Connectionist: processing occur s in parallel through the simultaneous
activation of nodes.
1.5 The interdisciplinary perspective
The philosophical approach
- The primary method of philosophical inquiry is reasoning
, - Deductive reasoning: involves the application of the rules of logic to
statements about the world (College students study three hours every night +
mary is a college student = mary studies three hours per night)
- Inductive reasoning: making observations about specific instances in the
world, noticing commonalities among them and drawing conclusions
(Whiskers the cat has four legs + Scruffy the cat has four legs = All cats have
four legs)
The psychological approach
- Psychologists attempt to understand not just internal mental phenomena, but also the
external behaviors that these internal phenomena can give rise to.
- They apply the scientific method to both.
- Scientific method: start with an idea or hypothesis about how the world works
and then design an experiment to see if the hypothesis has validity. The
resulting data support or fail to support the hypothesis.
- Psychoanalytic psychology conceives of mind as a collection of competing
entities, while behaviorism sees it as a device that maps stimuli onto
behaviors.
The cognitive approach
- Believes that the mind, like a computer, could be understood in terms of information
processing.
- Inherent in the cognitive approach is the idea of modularity. Modules are functionally
independent metal units that receive inputs from other modules, perform a specific
processing task and pass the results of their computation onto yet additional
modules.
- Computational modeling: involves carrying out a formal (typically software-
based) implementation of a proposed cognitive process
The neuroscience approach
- Neuroscience provides a description of mental events at the implementational level.
It attempts to describe the biological ‘hardware’ upon which mental ‘software’
supposedly runs. It studies the cell biology of individual neurons and of neuron-to-
neuron synaptic transmission, the patterns of activity in local cell populations and the
interrelations of larger brain areas.
- Neuroscientists employ a wide variety of machines to measure the performance of
the brain at work. These include PET-scanners, CAT-scanners and MRI-machines.
The network approach - see ch. 7
- Mind is seen as a collection of individual computing units that are connected to one
another and mutually influence one other’s activity via the connections.
The evolutionary approach
- Attempts to elucidate the selection forces that acted on our ancestors and how those
forces gave rise to the cognitive structures we now possess. Evolutionary
psychologists also adopt a modular approach to mind: the modules correspond to
‘favored’ cognitive capacities that were used by ancestors successful at solving
certain problems.
- A variant on this theme is evolutionary computing, in which the rules of evolution are
applied to create successful computer algorithms.
The linguistic approach
- Focuses exclusively on the domain of language.
, - Adopts a very eclectic methodological approach. Language researchers employ
experiments and computer models, study brain-damaged patients, track how
language ability changes during development and compare diverse languages.
The AI-approach
- Concerned with getting computers to perform tasks that have heretofore required
human intelligence.
- AI secondarily gives us insights into the function of human mental operations.
The robotics approach - see ch. 12
- While AI workers build devices that ‘think’, robotics researchers build machines that
must also ‘act’.
- Robotics helps us to think about the kinds of minds that underlie and produce
successful goal-oriented behaviors under certain conditions.
1.6: In Depth: Categories of Mental Representation
- A concept, an idea that represents things we have grouped together, is perhaps the
most basic form of mental representation.
- A proposition is a statement or assertion typically posed in the form of a simple
sentence, that can be proved true or false.
- A production rule is a conditional statement of the form: ‘If x, then y’, where X
and Y are propositions. The ‘if’ part is called the condition.
- Declarative knowledge is used to represent facts.
- Procedural knowledge represents skill.
- Analogy, a form of mental representation and/or reasoning, involves applying one’s
familiarity with an old situation to a new situation. They are a useful form of
representation because they allow us to generalize our learning.
