LIN 463

if in a single-layer perceptron, the net sum is equal to 0.5 and the threshold is equal to 1 (this is different from the regular threshold). what should be the output? a. 0 b. 0.05 c. 1 d. 0.5

0

Why is a single scale, utility, possible in decision making? Agents can make intuitive decisions using System 1. Agents behave in a random manner. Agents usually make choices in a consistent way. Agents usually make choices in an unpredictable way.

Agents usually make choices in a consistent way.

Which of the following is INCORRECT according to the Appraisal Theories? Appraisal Theories consider emotions stative mental states. An infinite number of combinations of appraisal programs is possible. Emotions are governed by if...then... rules. Appraisal Theories are a kind of computational theory of emotions.

Appraisal Theories consider emotions stative mental states.

Which of the following is INCORRECT concerning addiction and SDIs? Chronic SDIs have less DA receptors. Chronic SDIs are more sensitive to future reward than control subjects. Chronic drug abusers are less attracted by natural rewards such as money. Without drugs, chronic SDIs generally feel more depressed due to low DA level.

Chronic SDIs are more sensitive to future reward than control subjects.

Which of the following is INCORRECT according to the computational theory of emotions? Emotions are caused by environmental stimuli. A "fear" program can cause communication process changes (e.g. you stop talking). Emotions are submodules of other modules of the mind. Emotions involve information processing.

Emotions are submodules of other modules of the mind.

The value of a state is equal to the reward received in that state T/F

False

Which of the following is NOT a reason why researchers argue for Universal Moral Grammar? Intuitive moral judgement is a module similar to language Intuitive moral judgement is domain-specific Intuitive moral judgement is controlled Intuitive moral judgement is innate

Intuitive moral judgement is controlled

Why did people usually do better in the following modified Wason Selection Test (only above 18-year-olds- can drink beer. find out if this rule has been violated) Moral judgement is domain-general Moral judgement is domain-specific Moral judgement is innate Moral judgement is dissociable

Moral judgement is domain specific

In Q(St, At) = Rt + Y1R(t+1) + ... a. Y is the learning rate b. Rt+1 is the current reward c. At is the current state d. Q(St, At) is the cumulative discounted future reward

Q(St, At) is the cumulative discounted future reward

According to the reading Gibney (2015), what are the three big ideas in recent AI? a. Search b. Neural Network c. Computation theory d. Reinforcement learning

Search Neural network reinforcement learning

Concerning objective values, which of the following is INCORRECT? a.One common objective value is monetary value. b. In gambling, a goal of the gambler is to maximize the expected objective value. c. The decision-making theory based on the objective value can explain why Batman tried to save Rachel in The Dark Knight. d. The objective value is not the utility.

The decision-making theory based on the objective value can explain why Batman tried to save Rachel in The Dark Knight.

Which of the following is INCORRECT regarding morality? moral judgement has its innate basis moral intuition can is influenced by social factors The joker's social experiment failed because social factors does NOT have an effect on moral judgement

The joker's social experiment failed because social factors does NOT have an effect on moral judgement

If a patient with split-brain sees an image of a key in the left visual field, what will the patient report? The patient will say s/he saw nothing. The patient will say s/he saw a key.

The patient will say s/he saw nothing.

In reinforcement learning, Y must be a number between 0 and 1 because we value the current reward more than the reward we may receive in the future

True

Only when a function can be implemented with a step-by-step sequence of operations can it be implemented by a Turing machine

True

which of the following is INCORRECT about the universal Turing machine (UTM) a. the instructions of a Turing machine are the transition rules of that Turing machine b. UTM stores both instructions and input in its tape c. UTM reads the instructions and input in its tape d. UTM can implement any Turing machine e. UTM can implement any conceivable function

UTM can implement any conceivable function

Concerning the learning rate in neural networks, which of the following statements is correct? a. If the learning rate is too big, the model converges slowly. b. The learning rate is constant across different neural networks. c. We can decrease the learning rate gradually in the training of a neural network. d. If the learning rate is too small, the model cannot converge.

We can decrease the learning rate gradually in the training of a neural network.

Which of the following is INCORRECT for the question: When does our behavior appear "irrational"? When the subjective value is much bigger than the objective value. When the objective value is much bigger than the subjective value.

When the objective value is much bigger than the subjective value.

C(w) = 1/2∑i(target(i)−output(i))2C(w) = a cost function a sigmoid function a learning algorithm an activation function

a cost function

Which is true concerning the gradient ∇C(w)? a. It indicates the fastest way to maximize the cost function. b. It indicates the fastest way to minimize the cost function.

a. It indicates the fastest way to maximize the cost function.

under which of the following circumstances will the perceptron learn from the prediction error? select all that apply a. target=0, output =0 b. target=1, output =0 c. target=0, output =1 d. target=1, output =1

a. target=1, output =0 a. target=0, output =1

Kahneman's Nobel lecture said, "We are not accustomed to thinking hard and are often content to trust a plausible judgment that quickly comes to mind." Which idea of the Prospect Theory did Kahneman stress? Reference point Representativeness Availability Accessibility

accessibility

Computation involves... a. symbolic output b. an algorithm c. symbolic input d.a goal

all

Which of the following is incorrect concerning Turing machines? a. a function is computable if it can be implemented by a Turing machine b. all conceivable functions are computable c. turing machines defines computation d. turing machines defines what is computable

all conceivable functions are computable

Which of the following can be a common scale for objective value. Money Gold Shells All of the above

all of the above

In the literature, emotions were defined as feelings motivations evaluations All the above

all the above

a Turing machine may NOT have a. a finite set of tape symbols b. a finite transition table c. an infinite tape d. an infinite set of states

an infinite set of states

Which of the following brain areas are crucial to decision making? Ventromedial prefrontal cortex Orbitofrontal prefrontal cortex Both Neither

both

Appraisal Theories of emotions can explain why we have subjective feeling of emotions.

false

Substance abusers are not sensitive to immediate punishment.

false

System 1 is fast and controlled.

false

The Prospect Theory is a theory mainly explaining the heuristics in System 2.

false

The human mind is identical to a digital computer, and the digital computer can be reduced to a universal Turing machine

false

The learning rate is constant across different neural networks

false

The lecture shows that the graphic representation of a Turing machine is derived from the transition table

false

The reference point means losses are more significant to us than gains.

false

connectionist neural networks use transition rules

false

different functions can be implemented by a single Turing machine

false

Δwj = 𝛼 (output(i) − target(i) ) xj(i) is the perceptron learning rule.

false

Which of the following is the hard problem of consciousness? Are feelings and consciousness correlated? How is consciousness implemented in the brain? What is the neural basis of consciousness? How does the physical brain give rise to consciousness?

how does the physical brain give rise to consciousness?

why the following quote is counter-intuitive: "Rocks are smarter than cats because rocks have the sense to go away when you kick them? a. it ignores the symbols used by the agents in their responses b. it does not consider intelligence as information processing c. it does not consider the goals of the agents d. it takes into account the rules that have led to the responses of the agents.

it does not consider the goals of the agents

which is the most important feature of a perceptron? a. it can perfectly complete a categorization task b. it is able to learn the right weights and bias c. it can process information automatically d. it can generate an output

it is able to learn the right weights and bias

which of the following is INCORRECT regarding the single-layer perceptron? a. it uses the sigmoid function as the activation function b. Frank Rosenblatt created the first single-layer perceptron c. the output of the perceptron is determined by the eights and the bias d. none of the above

it uses the sigmoid function as the activation function

The rationalist model of moral reasoning suggests that... moral reasoning can be captured by mathematics moral reasoning is intuitive moral reasoning involves emotions moral reasoning is part of system 1

moral reasoning can be captured by mathematics

is the following a sigmoid function? linear looking graph

no

Which of the following is INCORRECT according to the computational theory of mind? a. Turing machine sets up the theoretical basis of the computational theory of mind b. This theory proposes that all mention is computation c. computation involves the processing of symbols using rules d. each fodorian module is a turing machine e. none of the above

none of the above

In reinforcement learning, the purpose of learning is to... a. obtain the highest immediate reward b. obtain the most distance reward c. minimize the loss d. obtain the cumulative future reward

obtain the cumulative future reward

A key property of many approaches to choice in cognitive science, psychology, economics, computer science, is... explaining how choices are neurally represented in the brain representing options with a common scale representing options with multiple scales explaining how the agent learns

representing options with a common scale

Utility is expected future reward subjective value objective value expected utility

subjective value

Expected utility is sum of [probability * subjective value for all possible outcomes] sum of [probability * future reward for all possible outcomes] sum of [probability * objective value for all possible outcomes] discounted future reward

sum of [probability * subjective value for all possible outcomes]

What can the following figure tell us? (duck/bunny illusion where you can switch what the painting is) The brain entertains one interpretation at a time. The brain entertains two interpretations simultaneously. The brain generates subjective experience solely based on the visual stimuli falling on the retina. All of the above.

the brain entertains one Interpretation at a time

which of the following is INCORRECT regarding connectionism? a. connectionism is neurally inspired b. the input of a perceptron or a neural network is a number c. the output/activation of a unit is either 0 or 1 d. the output of a perceptron or a neural network is a number

the output/activation of a unit is either 0 or 1

popular activation functions include... a. the sigmoid function b. the step function c. any linear function d. any non-linear function

the sigmoid function the step function

"How are different types of information integrated in the brain?" is an easy question of consciousness.

true

A subset of substance abusers are not sensitive to future punishment

true

According to Appraisal Theories, emotions are special computational programs.

true

According to Appraisal Theories, emotions include at least an appraisal program and an affect program.

true

Anticipatory SCR (skin conductance response) measures subjects' responses to expected punishment.

true

Blindsight and binocular rivalry suggest that the mind is an active interpreter.

true

Change of state means the change of wealth is more significant than the state of wealth.

true

Different neurons are fired to encode the chosen value and the offer value.

true

In the perceptron learning rule, the bias is treated as a special weight associated with an input that is always equal to 1.

true

Libertarian paternalism argues that we can influence other's decision making but meanwhile ensure the freedom of choice.

true

Loss aversion means we assign a bigger utility to losses than to gains

true

Moral judgement is modulated by emotion

true

Neural networks can approximate any computable functions.

true

Neural networks resort to learning from big data, therefore they generally assume less innate knowledge than the computational theory of mind.

true

Some substance abusers are similar to patients with ventromedial frontal cortex dysfunctions.

true

The algorithm must be clearly specified in a Turing machine

true

The left hemisphere may house an "interpreter" function that gives rise to the conscious self.

true

Turing machines are driven by transition rules

true

a Turing machine can be implemented in different physical materials

true

connectionism uses a mathematic model that does not necessarily resort to symbolic processing

true

w′= w + Δw = w − 𝛼 ∇C(w) indicates how a weight is updated.

true

Which is INCORRECT concerning Δw = −𝛼 ∇C(w)? a. 𝛼 is the learning rate. b. Δw is opposite to the gradient that scaled by the learning rate. c. with Δw we can increase the cost function. d. Δw is the incremental change on weights.

Δw we can increase the cost function.

Concerning Δwj = 𝛼 (target(i)− output(i)) xj(i), which of the following is correct? a. 𝛼 is the weighted input. b. xj(i) is the weight associated with wj. c. (target(i)− output(i)) is the desired output. d. Δwj is the incremental change on wj.

Δwj is the incremental change on wj.