Models of the brain

Question 1

What is artificial intelligence (AI)?

Answer 1

the theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages
-developed out of symbolic logic

Question 2

What is reasoning with logic?

Answer 2

creating new knowledge from facts we already know/making inferences
-can replace words with symbols (aristotle) and machine makes same inference: symbolic logic

Question 3

Learning in neural networks

Answer 3

• learning might be changing the strength of neural connections
• conditioning
• which of the connections do we want to change and how?

Question 4

Why are AI models useful in cognitive psychology?

Answer 4

• ‘what i cannot create, i do not understand’ - building models can help us understand what we haven’t learnt yet
• Data-analysis model: data-driven, descriptive
• Box-and-arrow model: information processing model, conceptual, implicit assumptions e.g. memory model, not functional
• Computational model: information processing model implemented as a simulation, explicit assumptions, various levels of abstraction, model that can do something

Question 5

Why would we want to build a computational model?

Answer 5

• they make assumptions explicit so they can be tested
• can give specific predictions for outcome of an experiment so helps select which experiment to perform
• models can be explanatory even if not predictive e.g. c models of schiz can indicate causes of disorder without being able to predict individual cases - suggest treatment
• the brain is complicated, abstracted and idealized models can capture broad trends so is still useful

Question 6

David Marr + his 3 levels of understanding

Answer 6

• worked on visual processing - how can we understand info processing systems like the brain?
• came up with 3 levels of understanding: computation, algorithm, implementation
• we can build models at different levels but experimental techniques favour implementation level
• understanding the brain can cause bottom-up approach: implementation (neural circuits)–>rules–>problem (what we can solve)

Question 7

Computation

Answer 7

=why (the problem that needs to be solved)

Question 8

Algorithm

Answer 8

=what (rules to follow) - how can this computational theory be executed?

Question 9

Implementation

Answer 9

=how (physical action) how can the representation and algorithm be realised physically?
-experimental techniques prefer this level

Question 10

what does a top-down modelling approach aim to do?

Answer 10

aims to cover algorithmic function and might disregard biological implementation

Question 11

Epstein argues that computational models are:

Answer 11

often wrong, but can still be illuminating