The Psychology of AI
Summary of the articles and lectures
, Week 1: Introduction to the psychology of AI
Article: A (very) brief history of AI
Buchanan, 2005
Philosophers used the idea of machines to define ‘what it means to be human’.
● René Descartes: “mechanical man”
● Gottfried Wilhelm Leibniz: saw the possibility of mechanical reasoning devices using rules of logic
to settle arguments or disagreements.
Leibniz and Pascal designed machines that could calculate the so-called ‘calculators’. They never claimed
the calculators could think. Etienne Bonnot asked at what point machines would know enough to appear
to be intelligent.
In the 1940’s computers were referred to as ‘giant brains’, because of their abilities to calculate.
Robots → the public perception of an intelligent computer.
● Early efforts to make robots were not about intelligence, they were about mechanical
engineering.
● Now, robots are seen as powerful vehicles to test our ideas about intelligent behaviour.
AI is not just about robots → about understanding the nature of intelligent thought and action using
computers as experimental devices.
When AI was formed, it was influenced by many disciplines
● Engineering
● Biology
● Experimental psychology
● Communication theory
● Game theory
● Mathematics and statistics
● Logic and philosophy
● Linguistics
Now AI has grown and has influenced all these disciplines back.
Turing wrote a paper (1950) about ideas about the possibilities of programming an electronic computer
to behave intelligently, including a description of the landmark imitation game that is known as Turing’s
Test.
The first book about working AI programs was published in 1963 and was written by Edward Feigenbaum
and Julian Feldman.
Newell & Simon also wrote programs around the same time as Turing, and their visions were ahead of
their time, but they did not have the tools to actualize their programs.
1
,Satisficing: a fundamental principle of AI; in the absence of an effective method guaranteeing the
solution to a problem in a reasonable time, heuristics may guide a decision maker to a very satisfactory, if
not necessarily optimal, solution.
Minsky (1968) summarized the work in the first decade after 1950:
● The most central ideas were that of finding heuristic devices to control the range of trial and
error search and finding effective techniques for learning.
● After 1962, there came more focus on the problem of representation of knowledge.
○ How this knowledge was required was not important
Minsky’s work → network representations of knowledge in frames (society of minds)
● Since then, knowledge representation (formal and informal aspects), is a fundamental of every AI
program.
○ McCarthy: declarative knowledge representation that can be manipulated easily
Article: Computing machinery and intelligence
Turing, 1950
This is the article by Turing about ‘the imitation game’ and the turing machine.
Paragraph 1: The Imitation Game
“Can machines think?” → to answer this question, the terms ‘machine’ and ‘think’ need to be defined.
This is very difficult in this way (it could lead to thinking that the answer can be obtained in a statistical
way), so rephrasing this question is the best option. This new form is called the imitation game.
● Three people: the man (A), the woman (B) and the interrogator of either sex (C).
● C stays in a room apart from A and B.
○ C knows the two in the other room as X and Y and needs to determine who is A and who
is B.
● C can ask questions (like: what is the length of your hair) and A and B write their answer down.
● A’s objective is to try and cause C to make the wrong identifications.
● B’s objective is to help C make the correct identification → so give truthful answers
○ The best strategy is to give truthful answers and add ‘I am the woman, don’t listen to
him’ to her answers. But only this is not enough, since the man can do the same.
● The question here is: what will happen if the part of A is given to a machine?
○ That is the question that replaces ‘can machines think?’.
Paragraph 2: Critique of the New Problem
What is the advantage of ‘the new problem’?
● Draws a sharp line between the physical and the intellectual capacities of humans.
● There is little point in a ‘thinking machine’ that looks, sounds or feels like a human, only for the
purpose of testing intelligence.
An objection of the new problem is that machines may carry out ‘thinking’ in a very different way than
humans do → but if a machine can play the imitation game properly, this objection is rejected.
2
, Paragraph 3: The Machines concerned in the Game
What do we mean with the word machine? Three conditions need to be satisfied to get to the definition
of the machine used in the imitation game:
● Any kind of engineering technique can be used in making the machine.
● Allow the possibility that an engineer may construct a machine which works, but whose
operations can’t be described by its constructors, because the method that they applied was
experimental.
● The machine should be referred to as a ‘digital computer’ to exclude usual machines (humans,
biological creations, the specific kind of machine that was thought of in that time when talking
about machines).
Paragraph 4: Digital Computers
Digital computers are supposed to carry out any operation that could be carried out by a ‘human
computer’.
● The ‘human computer’ is supposed to be following fixed rules (can not deviate from them) using
an unlimited supply of paper on which calculations can be done.
The idea of a digital computer is old → Charles Babbage.
A digital computer can be defined as having three components:
1. Store → information for calculations and rules the computer has to follow.
● The information is broken into small packets (10 digits f.e.)
● The parts of information need to be assigned a number, so instructions can be more
specific (‘add number 709 to 698 and put the results in the latter storage position’)
○ Would be: 70969817 → 17 being the type of instruction as defined in the
rulebook.
2. Executive unit → carries out the operations.
3. Control → see that the instructions are obeyed to (according to the rules) and in the right order.
Programming = the creation of rules/instructions a computer must follow.
● Rules are important, because they make it possible to do a sequence over and over again (until a
condition is fulfilled).
● Define logic → mimicing the ability of a human computer.
Digital computer with a random element
● Instructions like throwing a die → sort of as having a free will.
Paragraph 5: Universality of Digital Computers
Discrete state machines - move by sudden jumps or clicks from one quite definite state to another.
● A machine can be on or off.
Digital computers fall within the class of discrete state machines, but the number of states of which such
a machine is capable is usually enormously large.
● Because it needs a huge store: all possible combinations that could be written by a human
computer, need to be known → store capacity.
3
Summary of the articles and lectures
, Week 1: Introduction to the psychology of AI
Article: A (very) brief history of AI
Buchanan, 2005
Philosophers used the idea of machines to define ‘what it means to be human’.
● René Descartes: “mechanical man”
● Gottfried Wilhelm Leibniz: saw the possibility of mechanical reasoning devices using rules of logic
to settle arguments or disagreements.
Leibniz and Pascal designed machines that could calculate the so-called ‘calculators’. They never claimed
the calculators could think. Etienne Bonnot asked at what point machines would know enough to appear
to be intelligent.
In the 1940’s computers were referred to as ‘giant brains’, because of their abilities to calculate.
Robots → the public perception of an intelligent computer.
● Early efforts to make robots were not about intelligence, they were about mechanical
engineering.
● Now, robots are seen as powerful vehicles to test our ideas about intelligent behaviour.
AI is not just about robots → about understanding the nature of intelligent thought and action using
computers as experimental devices.
When AI was formed, it was influenced by many disciplines
● Engineering
● Biology
● Experimental psychology
● Communication theory
● Game theory
● Mathematics and statistics
● Logic and philosophy
● Linguistics
Now AI has grown and has influenced all these disciplines back.
Turing wrote a paper (1950) about ideas about the possibilities of programming an electronic computer
to behave intelligently, including a description of the landmark imitation game that is known as Turing’s
Test.
The first book about working AI programs was published in 1963 and was written by Edward Feigenbaum
and Julian Feldman.
Newell & Simon also wrote programs around the same time as Turing, and their visions were ahead of
their time, but they did not have the tools to actualize their programs.
1
,Satisficing: a fundamental principle of AI; in the absence of an effective method guaranteeing the
solution to a problem in a reasonable time, heuristics may guide a decision maker to a very satisfactory, if
not necessarily optimal, solution.
Minsky (1968) summarized the work in the first decade after 1950:
● The most central ideas were that of finding heuristic devices to control the range of trial and
error search and finding effective techniques for learning.
● After 1962, there came more focus on the problem of representation of knowledge.
○ How this knowledge was required was not important
Minsky’s work → network representations of knowledge in frames (society of minds)
● Since then, knowledge representation (formal and informal aspects), is a fundamental of every AI
program.
○ McCarthy: declarative knowledge representation that can be manipulated easily
Article: Computing machinery and intelligence
Turing, 1950
This is the article by Turing about ‘the imitation game’ and the turing machine.
Paragraph 1: The Imitation Game
“Can machines think?” → to answer this question, the terms ‘machine’ and ‘think’ need to be defined.
This is very difficult in this way (it could lead to thinking that the answer can be obtained in a statistical
way), so rephrasing this question is the best option. This new form is called the imitation game.
● Three people: the man (A), the woman (B) and the interrogator of either sex (C).
● C stays in a room apart from A and B.
○ C knows the two in the other room as X and Y and needs to determine who is A and who
is B.
● C can ask questions (like: what is the length of your hair) and A and B write their answer down.
● A’s objective is to try and cause C to make the wrong identifications.
● B’s objective is to help C make the correct identification → so give truthful answers
○ The best strategy is to give truthful answers and add ‘I am the woman, don’t listen to
him’ to her answers. But only this is not enough, since the man can do the same.
● The question here is: what will happen if the part of A is given to a machine?
○ That is the question that replaces ‘can machines think?’.
Paragraph 2: Critique of the New Problem
What is the advantage of ‘the new problem’?
● Draws a sharp line between the physical and the intellectual capacities of humans.
● There is little point in a ‘thinking machine’ that looks, sounds or feels like a human, only for the
purpose of testing intelligence.
An objection of the new problem is that machines may carry out ‘thinking’ in a very different way than
humans do → but if a machine can play the imitation game properly, this objection is rejected.
2
, Paragraph 3: The Machines concerned in the Game
What do we mean with the word machine? Three conditions need to be satisfied to get to the definition
of the machine used in the imitation game:
● Any kind of engineering technique can be used in making the machine.
● Allow the possibility that an engineer may construct a machine which works, but whose
operations can’t be described by its constructors, because the method that they applied was
experimental.
● The machine should be referred to as a ‘digital computer’ to exclude usual machines (humans,
biological creations, the specific kind of machine that was thought of in that time when talking
about machines).
Paragraph 4: Digital Computers
Digital computers are supposed to carry out any operation that could be carried out by a ‘human
computer’.
● The ‘human computer’ is supposed to be following fixed rules (can not deviate from them) using
an unlimited supply of paper on which calculations can be done.
The idea of a digital computer is old → Charles Babbage.
A digital computer can be defined as having three components:
1. Store → information for calculations and rules the computer has to follow.
● The information is broken into small packets (10 digits f.e.)
● The parts of information need to be assigned a number, so instructions can be more
specific (‘add number 709 to 698 and put the results in the latter storage position’)
○ Would be: 70969817 → 17 being the type of instruction as defined in the
rulebook.
2. Executive unit → carries out the operations.
3. Control → see that the instructions are obeyed to (according to the rules) and in the right order.
Programming = the creation of rules/instructions a computer must follow.
● Rules are important, because they make it possible to do a sequence over and over again (until a
condition is fulfilled).
● Define logic → mimicing the ability of a human computer.
Digital computer with a random element
● Instructions like throwing a die → sort of as having a free will.
Paragraph 5: Universality of Digital Computers
Discrete state machines - move by sudden jumps or clicks from one quite definite state to another.
● A machine can be on or off.
Digital computers fall within the class of discrete state machines, but the number of states of which such
a machine is capable is usually enormously large.
● Because it needs a huge store: all possible combinations that could be written by a human
computer, need to be known → store capacity.
3
