2021-2022 Exam Questions

Question 1

(a) Use one pertinent example to define and to illustrate what the novice-expert
problem is [2 points].

Then, (b) describe at least three strategies to solve this problem, and explain how each strategy may be applied to the example you chose [4 points].

Finally, (c) evaluate which one of the strategies you have described is the best for solving
the novice experts problem, explaining why it is the best [4 points].

Answer 1

A) novice-expert problem occurs when non-experts are confronted with real and factual scientific
disagreement, when they do not know who to trust. Because they are not able to asses the content of the
expert’s testimony, they thus rely on imperfect judgement to evaluate the content of the scientist. An
example of this is when a novice goes to the doctor and they have been given two options of treatment
and the doctors do not agree, the patient cannot assess which treatment is better and safer than the other.

B) Choose 3 strategies:
1. Agreement with other experts, if the majority of experts agree about a subject, we can assume they are
knowledgeable enough to make a correct decision
2. Presented arguments, Information from putative experts is widespread and easily available
3. Conflict of interest, does one of the experts have some sort of way to gain something when his side has
been chosen.
4. Appraisal by “meta-experts” / look up credentials, but novices are not always in a position to assess the
significance of one’s credentials
5. Past track record, with the help of the internet the track record of experts is easy to check

C) Agreement with other experts is the best way since past track record is not a guarantee for the future
performance. Conflict of interest is not easy to check for a novice since it would be kept secret most of the
time. When using agreement with other experts you are not using one expert his opinion but a lot of them
thus, I believe this is the best way of evaluating.

Question 2

(a) Define knowledge [2 points].
Then, (b) explain some of the main sources of evidence in virtue of which we know that human activities are radically altering Earth’s climate [5 points].
Finally, use the example of climate science to (c) explain what kind of knowledge basic research aims to acquire and what kind of knowledge applied research
aims to acquire [3 points].

Answer 2

A) Knowledge is when
● You know that something is true.
● You believe something.
● You are justified in your opinion.

B) The Paris climate agreement states that we as the world population want to keep the temperature
below 2 degrees Celsius compared to pre-industrial levels. If we do not keep the temperature under 2
degrees, sea levels will rise more than 50cm, the Sahara Desert will expand, there will be more fluctuating
weather which is disastrous for animals. Columbia University Earth Institute has published a simulation that
shows how the average temperatures are going to change in the coming years and also a graph with the
amount of energy the world is using. In 2009, 18 scientific associations pleaded for the world to take
climate change more seriously and that the greenhouse gasses humans emit are the primary driver.

C) Basic research aims to develop knowledge, theories and predictions. In the case of climate change this
is making prediction models about temperature and sea levels and also gathering information on what kind
of gasses cause the hike in temperature. Applied research aims to develop techniques, products and
procedures. In the case of climate change this would be ways to store CO2 and Methane gasses but also
more efficient solar panels and electric cars.

Answer 3

A) 1. Hypotheses: It serves as a way to derive predictions from the hypotheses about the results of future
experiments, and then perform those experiments to see whether they support the predictions.
2. Expectations: Is used as a way to after the experiment has taken place compare your expectations to the
previously made assumptions. This is important so you can reflect on what you thought beforehand
compared to what it actually was.
3. Observations: Scientists use observation to collect and record data. This is important because this
enables scientists to construct and then test hypotheses and theories.
B) Example:
The population of Udaipur amounts to more than 500,000 people. 80% live on 2$ per day. 57% report that
their household has enough income to feed their family. The vaccination rate is 6% for children.
The researchers were willing to investigate how to increase the child vaccination rate in the region and
decrease the cost per jab. Therefore, they decided to devise an experiment using three categories.
One is the control group, which is default. Another one is the first experimental group, which refers to
children from diverse villages being vaccinated in mobile clinics. The third one refers to children from distinct
villages who had the option to go to a mobile clinic and who were incentivized.
➔ Hypothesis: Lack of incentives prevents vaccination among children
➔ Expectation: Mobile clinics will increase vaccination rates
➔ Observation:A change in the immunisation rate
Conclusion: Having a statement and then having a potential test which helps find a concrete result allows us
to use all three methods together, each closely linked to the previous.
C) Expectations (combine the two below):
The reason why scientist pursue such links between all three areas is so that there is no ambiguity and
confusion as to how a final remark/conclusion has been reached since they can trace backwards from the
result to the observation which then is done by testing which happens to be based on the expectation and
that is created by using a hypothesis. This makes sure there is reproducibility, by having each of the steps
written down and executed.
Scientific experiments are conducted to examine expectations against observations to produce evidence for
the hypothesis. In this field experiment, individuals are separated into control and experimental groups to be
able to observe the observation against the expectations. Therefore, the influence of the X variable can be
examined. This is done here by separating the children into a control group, an experimental group with a
mobile clinic, and other experimental groups with a mobile clinic and incentives.

Question 4

Hypotheses, expectations and observations are all important ingredients for most
sciences. (a) Describe the importance of each of these three ingredients for scientific
reasoning in general [3 points].

Answer 4

A) 1. Hypotheses: It serves as a way to derive predictions from the hypotheses about the results of future
experiments, and then perform those experiments to see whether they support the predictions.
2. Expectations: Is used as a way to after the experiment has taken place compare your expectations to the
previously made assumptions. This is important so you can reflect on what you thought beforehand
compared to what it actually was.
3. Observations: Scientists use observation to collect and record data. This is important because this
enables scientists to construct and then test hypotheses and theories.

Question 5

Hypotheses, expectations and observations are all important ingredients for most
sciences. (b) use a real-life example discussed in our course to describe a typical way in which the three ingredients work together [4 points].

Answer 5

B) Example:
The population of Udaipur amounts to more than 500,000 people. 80% live on 2$ per day. 57% report that their household has enough income to feed their family. The vaccination rate is 6% for children.
The researchers were willing to investigate how to increase the child vaccination rate in the region and
decrease the cost per jab. Therefore, they decided to devise an experiment using three categories.
One is the control group, which is default. Another one is the first experimental group, which refers to
children from diverse villages being vaccinated in mobile clinics. The third one refers to children from distinct villages who had the option to go to a mobile clinic and who were incentivized.

➔ Hypothesis: Lack of incentives prevents vaccination among children
➔ Expectation: Mobile clinics will increase vaccination rates
➔ Observation:A change in the immunisation rate

Conclusion: Having a statement and then having a potential test which helps find a concrete result allows us to use all three methods together, each closely linked to the previous.

Question 6

Hypotheses, expectations and observations are all important ingredients for most sciences. (c) explain in some detail, and on the basis of your example, at least one scientific aim scientists pursue by using these three ingredients together [3 points].

Answer 6

C) Expectations (combine the two below):
The reason why scientist pursue such links between all three areas is so that there is no ambiguity and
confusion as to how a final remark/conclusion has been reached since they can trace backwards from the
result to the observation which then is done by testing which happens to be based on the expectation and
that is created by using a hypothesis. This makes sure there is reproducibility, by having each of the steps
written down and executed.

Scientific experiments are conducted to examine expectations against observations to produce evidence for the hypothesis. In this field experiment, individuals are separated into control and experimental groups to be able to observe the observation against the expectations. Therefore, the influence of the X variable can be examined. This is done here by separating the children into a control group, an experimental group with a mobile clinic, and other experimental groups with a mobile clinic and incentives.

Question 7

Carefully reconstruct the “Dutch book argument,” explicitly stating (a) what conclusion this argument aims to support [2 points],

Answer 7

Introduction: In what follows, I will begin by explaining what the Dutch book argument is. What arguments
it aims to support and how this argument proceeds to reach its conclusion by using an example. Moreover, I
will evaluate why the Dutch book argument is (or is not) convincing to support its conclusion.

A) The basic Idea of Dutch book arguments: the view that an agent’s degrees of belief should satisfy the
axioms of probability. There would be a problem if your degrees of belief do not conform to the rules of
probability (Axiom 1, 2 & 3)*, because there are possible betting situations where you are guaranteed to lose money. You could fall prey to a Dutch book. But since you do not want to lose money your degrees of
belief should respect the rules of probability.

Question 8

Carefully reconstruct the “Dutch book argument,” explicitly and (b) explaining – in the light of a simple
example – how this argument proceeds to reach this conclusion [5 points].

Answer 8

B) A simple example of a Dutch book is this: I show you a coin. Your degree of belief that a toss of this coin
will come out Heads is 0.6. Your degree of belief that the toss will come out Tails is 0.6. If you calculate the
probability of you would get 1.2 which is impossible according to the probability Axioms.

Suppose you are willing to take a bet on the outcome of the coin toss. A Dutch bookie offers the following
bet. a) 1.5 to 1 if the coin lands on heads or b) 1.5 to 1 if the coin lands on tails. Since your degree of belief for the coin to be heads is 0.6 and the same for tails (0.6).

Calculate P(h) = X/(X+Y) = 1.5/(1.5+1) = 0.6 so you should be willing to accept the bet. But now you have accepted both bets. If the coin lands on Heads, you win 10 euros, but lose 15 euros on the tails bet (total = -5 euros). So you are guaranteed to lose. If you have been Dutch booked.. This is based on the agent taking the bet just because the expected value is non-negative.

Question 9

Carefully reconstruct the “Dutch book argument,” (c) evaluate whether or not the Dutch book argument provides convincing support for its conclusion [3 points].

Answer 9

C) Option 1:
However, the Dutch book argument is not convincing to support its conclusion (being that there will be a
guaranteed loss). Especially betting behaviour doesn’t seem to generally be a good guide about what one
(should) believe. The idea is that every person will adhere to the probability Axioms, but with this
assumption we forget something important. The agent can always prevent a sure loss by simply refusing to
bet. Just because the expected value is non-negative does not mean that an agent has to take the bet. Also
gambling might not be a good way to prove this theory, because many other factors play a role: How much
you like betting, emotional value attached to money and/or beliefs in the outcome of the event.

Option 2:
To conclude the Dutch book argument is convincing to support its conclusion. If a to make a rational decision
the agent should stick to the probability Axioms. Even though an agent could always prevent a sure loss by
simply refusing to bet, they could not be aware of the fact that their logic is flawed.

• Axiom 1: All probabilities are numbers between 0 and 1
• Axiom 2: If a proposition is certainly true, then it has a probability of 1. If certainly false, then it has prob. 0.
• Axiom 3: If h and h* are exclusive alternatives (they cannot both be true at the same time),
  then P(h or h) = P(h)+P(h)

Question 10

(a) Describe the problem of confounding in the light of an example mentioned in our course [2 points].

Answer 10

A) There can be variables that are not considered, which can influence both the independent and
dependent variable. These confounding variables can be a common cause between X and Y. In the example
in the lecture, an increased ice cream consumption (X) leads to more people drowning (Y). However, this
does not take into account the confounding variable ‘hot weather’ (variable Z) which influences both (for
example how many people go swimming).

Question 11

(a) Describe the problem of confounding in the light of an example mentioned in our course [2 points]. Then, (bi) define randomization [1 point] and (bii) explain how it differs from random sampling [1 point].

Answer 11

B) bi) Randomization: is the process of making a process of group random, often two distinct groups: a
control and an experimental group. People are assigned to a group to evenly spread out factors that could influence results of an experiment.
bii) Random sampling: A random group is selected from a larger group so that it is a random sample from
the overall population. It also differs from random sampling because each potential participant has an equal chance of being selected for the control or experimental group.

Question 12

Finally, evaluate whether or not randomization is (ci)
necessary for solving the problem of confounding, supporting your position with an argument [3 points], and whether or not randomization is (cii) sufficient for solving the problem of confounding, supporting your position with an argument [3 points].

Answer 12

C) ci) Randomization is necessary for solving the problem of confounding, because it helps to keep variables
constant. Essentially, the researchers should distribute participants with certain characteristics equally
among the control and the experimental group. This decreases the exposure to confounding variables.

cii) Randomization is necessary because it prevents selection bias and makes a distribution process
more equal and fair. A thorough random process filters out possible confounding variables. This reduces
potential for confounding by generating groups that are fairly comparable with respect to known and
unknown confounding variables.

Randomization minimizes the effect of confounding. However, randomization alone is not sufficient to
eliminate the problem of confounding variables. As if, for example, the target group was already chosen with
some bias, dividing participants into two groups randomly would not solve the issue. There would be a
chance that if we divide participants randomly, one group would have more biased selected target group
participants. Also, if the distribution is done once, it doesn’t guarantee the elimination of extraneous
variables. So, randomization by itself is not enough to eliminate the confounding variable problem.
Researchers could apply random sampling, restriction, matching and also manually make adjustments to the
group. Therefore randomization is necessary, but not sufficient for solving the problem of confounding.

Question 13

Define what a scientific model is in general [2 points].

Answer 13

A scientific model has the aim to make a particular part of the world easier to understand, define or
visualize by referring to usually commonly accepted knowledge. It is often a simplification of a real world
scenario shaped by constraints or assumptions.

Question 14

(b) describe Schelling’s model of housing segregation and its main result [2 points].

Answer 14

Thomas Schelling created the model of housing segregation. The model shows how segregation is
caused by minor preferences of people to have “a like-neighbors”. Schelling used simple tools to illustrate
how it develops. By using a checkerboard and two different types of coins. By placing coins randomly on
the board and leaving empty spaces that serve as free living space. Inhabitants of the filled spaces move
away to a random free living space if a certain percentage of surrounding squared are inhabited by a
different type of coin (than they themselves are).

Various assumptions are made. Firstly, there are only two types of agents. Secondly, they live on a
two-dimensional grid. Thirdly, at the start agents are placed randomly on the board. Furthermore agents have an unknown preference regarding their neighborhood and are only satisfied if a certain ratio of its neighbors is in the same category as they are. Lastly agents act according to a simple rule; move location randomly whenever they are dissatisfied with their neighborhood.

The main result is that segregation can emerge even where agents do not have racist attitudes (and even
do not mind being a minority in a neighborhood as long as there are a % of similar agents). Small
preferences for like neighbours result in massive segregation.

Question 15

(ci) Define idealization in general [2 points], and (cii) explain what idealizations this model makes and for what purposes [2 points].

Answer 15

The Schelling’s model contains idealizations (the representation of something as ideal or perfect). The
goal is to get rid of everything that’s not essential to making a point.

Cities are not a perfect grid like the checkerboard. Not all people share same preference for like
neighbours, and they might not even know each other. People do not randomly move house, or any time
they are unsatisfied. There is lots of other factors involved especially economical and ecological factors.

The purpose of idealizations is to make the model easy to construct, manipulate, analyse, and run on a
computer. So people can focus on the important aspects of the phenomenon. Schelling wanted to make
clear that individual choices can lead (under specific conditions) to significant unintended consequences
for larger groups.

Question 16

Schelling’s model of housing segregation and its main result, (d) evaluate what one can learn (if anything) about the real world from this model [2 points].

Answer 16

Schelling’s concept can be applied to real life situations. Even though scientific models are simplified in order to make it understandable. Schelling’s point still stands: Individual choices can lead (under specific
conditions) to significant unintended consequences for larger groups.

Question 17

(a) Carefully reconstruct the no miracles argument and the pessimistic meta-induction argument, and state what conclusions these two arguments aim to support [5 points].

Answer 17

I will start my answer by explaining the No Miracles Argument which aims to establish that our current
best scientific theories and models are most likely true and therefore our current best theories mark a
progress compared to previous, probably false ones. After that, I will describe the Pessimistic Induction
Argument which looks at the history of science and argues that if past successful and accepted scientific
theories were found to be false, we have no reason to believe the scientific realist’s claim that our currently
successful theories are approximately true and therefore we have no reason to believe that our knowledge of reality gets more accurate over time.

Firstly, the No Miracles Argument uses two premises:
The first one is that our best scientific theories are massively predictively successful, and facilitate
incredible technological innovation. This means that science practically has succeeded and that nowadays
we have benefited from it with immense innovations, such as in the technology sector.

The second premise describes that the best explanation for these successes is that our best theories are true.

From these premises, the No Miracles Argument concludes that our best scientific theories are true. Therefore, our understanding of reality is progressing.

There are two problems with the No Miracles Argument. The first problem is the use of abductive inference. It is hard to define what explanations are ́better ́ than others without imposing some criterion on the basis of which the judgment is made. The second problem is that this argument fails to situate science in a historical context, it is naïve to judge that successful predictions in theories make other theories possible as times are always changing and you would have to look at each specific theory within their own ́historical context ́.

The Pessimistic Induction Argument also uses two premises:
P1: There have been many empirically successful theories in the history of science that have subsequently been rejected as false.
P2: Our current best theories are no different in kind from those theories that were rejected

Therefore we have reason to believe our current best theories will be rejected as false. Hence we should not
think that our current theories are true. This argument is inductive since it generalizes an observation from
the past to all our current best theories.

Question 18

the no miracles argument and the pessimistic meta-induction argument, (b) critically evaluate which one of these two arguments is the most convincing, supporting your position with one reason [5 points].

Answer 18

I believe the Pessimistic Induction Argument is more convincing, mainly because of the two problems of
the No Miracles Argument: the use of abductive inference and a failure to situate science in a historical
context, even though a theory can be believed as true at a given time (following the No Miracles Argument), a new theory can come which shows the old accepted theory to be false. false. This is what the Pessimistic Induction Argument uses as a first premise and combining this with the premise that our current best theories are indifferent in findings from those old accepted theories that used to be accepted, it concludes that we have no reason to believe our current best theories are true.

I would like to mention that this does not mean that we should not trust our current best theories.
Science is still our best way at generating knowledge and the use of science to learn about our world is
unlikely to be surpassed by better ways of generating knowledge even if some scientific theories are
sometimes abandoned for new ones that.

To conclude, after explaining the No Miracles Argument and the Pessimistic Induction Argument I argued that the Pessimistic Induction Argument is most convincing in the light of the two problems of the No Miracles Argument using the example of Newtonian mechanics. After this, I emphasized that this does not mean that we should no longer trust science using the same example.

Question 19

(a) What’s robustness analysis [3 points]? (b) What is robustness analysis useful for [3 points]? (c) Give one example to Illustrate your answers to (a) and (b) [4 points].

Answer 19

Introduction: I will start my answer by describing robustness analysis, which is one way in which
modelers can learn about what features have consequences for the study of a certain target system. After that, I will discuss various reasons why robustness analysis is useful for scientists. Finally, I will elaborate more on this with the example of co2 levels in the atmosphere.

Robustness analysis is one way in which modelers can learn about what features have consequences for
the study of a certain target system. First you start with building small, slightly different models of the same
target. And by building these models with each slightly different aspects scientists can manipulate them in
comparable ways to find out about the kind of output they give. After that you can compare these models
results and draw conclusions about what kind of assumptions and features matter.
‘
Moreover, to elaborate on what robustness can be used for, but especially what it is useful for is to find
out about what idealizations matter and what idealizations don’t matter. If a model, given a small change in a parameter, gives me widely different results and given that these results compared to our observations are off might be one reason to
reject the initial idealization, because it leads to a very inaccurate result with respect to our target system. So
the first point of robustness analysis is assessed to what extent yields results that are invariant with the
results that do not change when you change small parameters of a model.

Robustness can be helpful when it comes to identifying model features responsible for certain results. If you can show that certain variation in a certain territory does not lead to a variation in the target system the model can identify what idealizations matter and what idealizations do not matter.

To evaluate what similarities and idealizations matter to learning about the world. Say that you have
smaller incomplete models of a specific feature of the climate and when you compare different results,
accounts of these slightly different models you can get a more comprehensive picture of the complex
system.

C) To give an example about robustness analysis we can look at various industries and phenomena. However, it is especially evident in Climate science where different research groups build slightly different models of the same target system, for instance the level of co2 in the atmosphere (target system) and by building models of the atmosphere considering different parameters and values scientists can manipulate them in comparable ways to find out about the kind of output they give. You then compare those results and then you can draw conclusions about what kinds of assumptions and features matter when it comes to representing co2 levels in the atmosphere. If, for example, you can show that variation in a certain
parameter in a certain territory, let’s say rain, does not lead to any variation in co2 levels in the atmosphere.
Then that is one reason to conclude that level of rain in a certain region is not responsible for the final co2
level that the model is giving you. Therefore we can analyse what idealization has an impact on our target
system (atmosphere) and what idealizations not (rain).

In conclusion, I have elaborated that robustness analysis is one way in which scientists can learn what
features have consequences for the study of a certain target system with small, slightly different models. It is
useful for identifying significant idealizations and insignificant ones and what specific features lead to what results and to see what similarities and differences there are between these

Question 20

9. Consider these three ideas:

(i) Scientific models contain falsehoods.
(ii) Knowledge requires truth.
(iii) Scientific models allow scientists to acquire knowledge.

(a) Explain each of the three ideas [4 points], and (b) evaluate whether or not, and in what
sense, those three ideas can all be true at the same time [6 points]. Refer to at least one actual scientific model discussed in our course to illustrate and support your answer.

Answer 20

Firstly, to answer this question, it is important to know what a model is because then we can look at the
ideas that are associated with a model. A model is an idealized representation of a complicated situation or
phenomena (the target system), which has the goal to make the target system easier to understand and
track. This means that a model stands in for something else. To illustrate the concept of a model, it is best to
use an example.

i) All models contain falsehoods. The model contains idealizations; it contains distortions or simplified
matters. The first means that models do not include or leave out certain features that the target system does
have, the latter means that the model includes certain functions that the target system does not have. The
intention of these is twofold: to make the model easy to understand and construct, as well as to focus the
attention on core factors. Bay Model this means that for example the number of sailboats was excluded from the model (a distortion) but that the weather would be the same at all times (a simplification). These idealizations are made such that the model doesn’t become too complex.

ii) The second idea is that knowledge requires truth. Knowledge actually contains this in its definition: a
belief that is true and justified. Truth means that a belief must not be false, that is it must be a fact. Again,
we can illustrate this with the Bay area. It is true that water flows in the bay area, it is not true that it stands
still. This is influenced by many factors, including the weather, boats that make the water move etc. I can believe that the water stands still, but as it is clearly observable that the water does not stand still, this belief is not true and this belief is does not lead to knowledge.

iii) The third idea is that scientific models allow scientists to acquire knowledge. This idea contains elements from both ideas. However, the first two ideas are in contradiction. The first idea stated that all models contain falsehood and the second idea states that knowledge contains truth.

B) The idea of a model is to simplify reality, in such a way that a real situation or phenomena can be
explained. For this, idealizations are made, containing falsehood. Still, one can learn from a model, as long as
the model is similar to the target system in relevant ways. So by including factors that matter, scientists can learn about a certain aspect of a target
system, without replicating the entire target system.

Question 21

Should rational decision making be modelled as if humans always seek to maximize their utility through increasing their income and consumption? Provide an argument in support of your answer based on some specific model of rational choice and on some
concrete example of how humans actually choose discussed in our course [10 points].

Answer 21

Rational choice theory states that individuals use rational calculations to make rational choices and achieve outcomes that are aligned with their own personal objectives. These results are also associated with maximizing an individual’s self-interest. Rational decision making should indeed be modelled to
maximization, however, we should not forget that these models are idealized. In an ideal world, the model would hold true, however, in the real world this is almost never the case. Economists argue that costs that have already been incurred and cannot be recovered, sunk costs, are irrelevant to rational decision making.

An example is if you have started a graduate program at a university, and in the last year you feel like you are
taking the wrong path. A rational choice could be to change program or do something else that you feel
really passionate about. To change is the rational thing to do. This rational choice is not leading to the
maximized utility through increasing income or consumption, therefore we can conclude that the rational decision-making model is idealized and is different from the real world.

Question 22

(ai) Define inductive inference [1 point], and (aii) explain how it differs from deductive
inference [2 points].

Answer 22

(Ai) Inductive reasoning is reasoning based on an observation, often a sample. Induction refers specifically to inference of a generalized conclusion from particular instances.
(Aii) The difference with deductive inference is that induction is based on general observations while
deduction is based on facts or universal premises.

Question 23

Then, (bi) give one example of inductive inference [1 point] and (bii) one example of deductive inference [1 point].

Answer 23

(Bi) For example, 25 of your 30 classmates order the same burger at a hamburger restaurant. From your
observation, you induce that the particular burger is good, otherwise, your classmates would not order it.
(Bii) A deduction would be that you know that that specific hamburger gets ordered most often or it is
widely accepted that that is the best burger.

Question 24

Finally, (ci) describe the problem of induction [2
points], and (cii) evaluate whether or not inductive inference is the best we can 3 reasonably
hope for when it comes to making reliable predictions and generalizations [3 points].

Answer 24

(Ci) The problem is that no inductively strong argument, no matter how strong, guarantees the truth of its conclusion. If you base your reasoning off of 9 out of 10 people saying the same thing, those 9 people could still be completely wrong.
(Cii) This problem leads to the question of whether inductive inference is the best we can reasonably hope
for when it comes to making reliable predictions and generalizations. As stated prior to this conclusion,
inductive inference is, by definition, not the absolute truth. However, one could assume that if a lot of
people state a premise or give a certain answer, reliable predictions and generalizations could be made, since predictions and generalizations are based on generally accepted premises. Hence leading to the conclusion that inductive inference is the best we can reasonably hope for, if we are not aiming for absolute truth but for reliable predictions and generalizations.

Question 25

12. (a) State the value-free ideal of science, explaining whether it makes an empirical claim
    about how science actually works or a normative claim about how science should ideally
    work [2 points].

Answer 25

A) The value-free ideal of science states that science should be free from the influence of any value, since
any science influenced by values in any way is bad science. This is a normative claim since this claim asserts that science ought not to be influenced by any value, otherwise it instantly would be bad. The value-free ideal suggests that science is a source of objective knowledge to the extent it is free from values.

Question 26

(b) Describe at least three ways in which values can legitimately factor into science [3 points].

Answer 26

Three legitimate roles of values in science are:
- Choice of research questions.
- Based on the topic and/or purpose of a paper/research, values can influence the choice of
the research questions.

• Choice of how much evidence is needed before accepting or rejecting a hypothesis.
• Values can lead to a different measure of how much evidence is necessary before
  researchers accept or reject a hypothesis.
• Honesty
• Honesty is a value also beneficial and influential for papers. Researchers can choose how
  ‘honest’ they are by choosing to leave certain truths or conclusions out of the research.

Question 27

(c) In your view, does the fact that values sometimes factor legitimately into science constitute a violation of the value-free ideal, or is it consistent with that ideal?
Give an argument in support of your answer, illustrating it with one example covered in our
course. [5 points]

Answer 27

Whether it’s violated depends on the values exhibited: In my opinion, it doesn’t constitute a violation, if
the right values are followed. The important idea is to ensure that science is affected in the correct ways and
that science successfully deters adverse factors from becoming dominant.

E.g exclusions of members of certain groups from scientific societies and institutions (e.g., women and
people of color in science) this shouldn’t matter at all because…

Question 28

(ai) Define big data [2 points], and (aii) describe one example discussed in our course involving big data-driven research [3 points].

Answer 28

A) i) Data that is nowadays described as ‘Big data’ consists of large datasets that nowadays are typically in a
digital format and that are analysable with computational tools in a reasonable amount of time.
ii) An example discussed in our course involving big data-driven research is the one into data-driven hiring
systems. These data-driven hiring systems were there to assist in the recruitment process and to review
applicants. The data used in these systems were CV’s submitted by applicants over the last 10 years along
with the career dynamics of company employees. Based on this data an algorithm computed if a new
applicant was a good fit for the company.

Question 29

(b) Is big data-driven research always free
from any bias? Support your answer by referring to one pertinent example of your choice
and explaining different meaning of “bias” [5 points].

Answer 29

B) A bias is an inclination or prejudice for or against one person or a group, especially in a way considered to be unfair. Big data-driven research can also be biased, even if it is not meant to be that way. An example is the data-driven hiring system discussed in this course. This system, based on CV’s submitted by applicants over the last 10 years along with the career dynamics of company employees, was also biased. In this case CV’s with the word “woman” were penalized and this algorithm ranked women for less pay than it ranked men. This is a bias based on gender towards men that is based on big data.

Question 30

(a) Refer to each point on the checklist for evaluating whether a practice qualifies as scientific, which we discussed in our course, to assess whether economics is a genuinely scientific field [6 points].

Answer 30

In what follows, I will begin by providing a checklist which is used to assess whether a practice qualifies as
scientific. Then, I will assess whether economics is a genuinely scientific field based on that checklist. Finally, I will illustrate and explain whether economics is a genuinely scientific field by giving a specific example of economic theory.

A) To begin with, an idea or project is qualified as scientific when it is in accordance with the following
characteristics:

● It aims to provide natural explanations of natural phenomena (naturalism)
● It puts forward ideas that can be tested with empirical evidence (empirical investigation, falsifiability)
● It updates ideas based on available evidence (evidentialism)
● Would abandon any idea that was thoroughly refuted (openness to falsification)
● Employs mathematical tools appropriately when they are useful (mathematical techniques)
● Involves the broader scientific community (social and institutional structure)

When looking at economics with these characteristics in mind, economics is not really a genuinely scientific
field because it is not in accordance with some points. The first one is that economics is not a natural
phenomenon. Economics can be regarded as a social science since it is about ownership, use and exchange
of scarce resources, and it is looking to explain the behavior of individuals, groups and organizations.
While science is based on evidence, economics for a big part is not based on evidence. Many concepts in
economics are based on assumptions without evidence for those concepts to be true. Examples are perfect
competition, perfect capital markets, rational behavior and so on. So these assumptions are not based on
available evidence and there is even evidence that those assumptions are not in line with reality. Yet these
assumptions are still being used and taught at schools and universities, while a genuinely scientific field
would abandon ideas that are refuted.

Question 31

(b) Illustrate and explain your answer with at least one specific example of your choice of economic theory [4 points].

Answer 31

A specific example of economic theory could be ‘Keynesian Economics’ which implies active government policy to manage aggregate demand in order to prevent economic recessions. The checklist points out that in order for an idea to be genuinely scientific something should provide natural explanations for natural phenomena. The government actively trying to control aggregate demand of the economy is not really a natural phenomenon. For this reason economics cannot be labeled as a genuinely scientific field.

In conclusion, I provided a checklist to assess whether a practice qualifies as scientific and came to the
conclusion that economics is not a genuinely scientific field based on the checklist. Finally, I provided a
specific example of economic theory to support this conclusion.

Question 32

19. My financial advisor tells me: “There is a 80% probability that your stocks will go up by
    the end of this year.” First, (a) explain the differences between objective and subjective
    interpretations of probability [3 points].

Answer 32

(A) The first perspective to look at probability is the epistemic view which judges probability based on a
rational belief. The concept of this subjective interpretation of probability implies that everyone has a certain degree of belief that an underlying hypothesis is true or false. The numerical formula for calculating the degree of belief is “X/(X+Y)” in which X is the amount that is lost if your claim about the hypothesis is wrong and Y is the amount you can win.

If we speak about the rational degree of belief then this adds the proposition that the assumption is
coherent with the three axioms of probability by Kolmogorov.If we observe a probabilistic judgment objectively we can proceed using a view based on frequency or propensity. The concept of frequency implies that an experiment is repeated within a long period of trials in order to obtain, for instance, a maximum likelihood of an event. Propensity interpretation looks at a probability through causal dispositions. For example being strong is a causal disposition of a bodybuilder or being hot is a causal disposition of an oven. In other words, it is the tendency of a system to behave in a
certain way.

Question 33

(b) describe the problems for each of the
interpretations of probability you have distinguished [3 points].

Answer 33

Subjective: difficulty: subjectivism refers to rationality based on Kolmogorov axioms but it can still be odd
—> if YOU believe world will end tmr is 90% it is your rational belief but still very unlikely
Objective frequency: is good for social sciences examining large populations but not for single events such as
the rising stock price
Objective propensity: they cannot capture the inevitability of an event (drink —> hangover)

Question 34

(c) provide one argument for whether I should interpret the probabilistic judgement of my advisor
objectively or subjectively [4 points].

Answer 34

If we are supposed to decide which interpretation of the advisor’s probabilistic judgment is
recommended, I would say it makes sense to apply an objective view based on propensity (maybe)…

Question 35

Explain in brief (ai) the classical approach [2 points] and (aii) the Bayesian approach to statistical testing [2 points].

Answer 35

The first option is that the classical method of statistics is characterised through so-called null hypothesis
testing (NHST) which is a statistical version of the HD method. It contains four main steps: Firstly, formulating a null hypothesis. Secondly, developing expectations in the form of probability. Then search for data that can give information about the relevance and credibility of the probabilities and lastly draw an inference from this comparison.

The second option is the use of Bayesian statistics which is the statistical version of the inductive method
and, in brief, it provides evidence of the observed data that either contributes to the hypothesis being true
or false. The procedure contains three steps which are firstly formulating competing hypotheses, secondly
gathering data about the target variable that can confirm or disconfirm the hypotheses and lastly inductively updating the probabilities about the probabilities.

Question 36

Then, (b) identify and critically evaluate at least two different types of problems for the Bayesian approach to statistics [3 points],

Answer 36

However, the Bayesian method reveals two main limitations when it comes to evaluating probability
related to hypotheses.
The first problem is the choice of prior probabilities. In particular, if the subjective degrees of belief differ for
specific hypotheses then this can lead to a falsification of the according probability as a result.
The second problem is that the Bayesian statistical approach is often not compatible with other sciences. For
instance, bayesian statistics cannot explain to what degree a certain hypothesis is correct or incorrect.

Question 37

(c) three different advantages that the Bayesian approach has over the classical approach to statistical
testing [3 points].

Answer 37

The first one is that it can be used to quantify the probability of a hypothesis being true. In the classical
approach, for instance, the NHST method only indicates if a certain hypothesis shall be rejected or not based on the p-values without any quantification.

The second one is that with the help of bayesian statistics we can now collect information for and against the hypotheses through posterior and conditional probabilities. An easy example to explain such a scenario is for instance the research about efficiency of a certain drug. If we want to quantify a certain hypothesis with the addition of a second variable

The last advantage is the fact that NHST only tells us that something is false. Bayesian statistics tells us more than a rejection through the characteristic of quantifying probability and giving conditional information that helps us to understand more of the prior events happening. For example if the event of someone dying, we can evaluate this event better if we for instance knew about the prior event of this person being a smoker for over sixty years.

Question 38

In 2017, a group of 72 authors proposed to enhance the reproducibility of the social
sciences by changing the p-value threshold for statistical significance from 0.05 to 0.005.
(ai) Define what a p-value is [1 point], (aii) what reproducibility means [1 points] and (aiii)
why it is important [2 points].

Answer 38

In null hypothesis significance testing, the p-value is used as a determining factor whether a
hypothesis can be rejected or accepted. In the case where you set your p-value threshold at 0.05 and the
p-value that comes out of your testing comes out to 0.04, you can reject the null hypothesis with a 95%
certainty. To simplify: If your p-value is below the threshold value, you can reject the null hypothesis. For
example: You conduct an experiment where you want to find out whether the temperature on a given day
is correlated with the amount of hours people watch Netflix on a given day. For your null hypothesis you
state that there is no correlation between temperature and watching Netflix. For your alternate
hypothesis you state that there is a correlation between the outside temperature and hours of Netflix
watched. You collect your data and analyze it in a computer program like Excel or SPSS. Your p-value of the
correlation comes out to 0.008. Considering the p-value threshold of 0.05, you can reject your null
hypothesis and state that the correlation is statistically significant.

Aii & Aiii) The reproducibility of an experiment states how easy it is to replicate an experiment. For
example: The aforementioned experiment concerning the correlation between the outside temperature
and hours of Netflix being watched, is easily reproducible. You only have to make a new survey or
questionnaire and send it out to another sample of people. However, not all experiments are as easy to
reproduce due to high costs, financial incentives, or other reasons. Reproducibility is important because it
solidifies findings made in previous experiments. An experiment done three times over with the same
results is more trustworthy than an experiment that has only been done once.

Question 39

Then, (b) describe one consequence of this proposal, if it were implemented [3 points]. Finally, (c) do you think this consequence would help with the “replicability crisis.”? Why, or why not? [3 points] (the reproducibility of the social sciences by changing the p-value threshold for statistical significance from 0.05 to 0.005.)

Answer 39

B & C) The proposal to reduce the p-value threshold from 0.05 to 0.005 to enhance the reproducibility of
the social science experiments and help deter the reproducibility crisis will have the consequence of
making the problem of P-hacking way more prevalent. The reproducibility crisis is the situation wherein
experiments are not redone and when they are redone, it is usually not done in a good way. This has
multiple reasons ranging from scientists with wrong incentives to do research because of their employers.
P-hacking is the practice of nit-picking or adding more data to generate a preferred p-value. Getting a
p-value of below 0.005 is more difficult than getting a p-value below 0.05. Because of this reason, I think
this consequence will only worsen the reproducibility crisis. A scientist driven by financial and/or social
incentives might be tempted to employ P-hacking to get their p-value below 0.005 to be able to get their
research published and recognized.

Question 40

(ai) State the value-free ideal of science [1 point], (aii) explaining whether or not this ideal accurately describes how scientific practices work [2 points].

Answer 40

A) The value-free ideal of science is the view that societal values should not have any interference in
the process that encompasses science. The polar opposite of the value-free ideal is a world wherein
all science is subjective and not driven by pure logic and rationality. In the real world, science falls
somewhere in between these two extreme opposites. Science is in some ways driven by logical and
rational factors and not fully influenced by values. However, scientists are humans, therefore it is
also rational to assume there is some degree of influence which societal and personal values have on
the outcome of scientific experience. Humans are not fully driven by logic, emotions and values
always play a role in judgement and decision making, albeit a small role in some situations.

Question 41

State the value-free ideal of science - use one example
covered in our course to describe one way in which some specific value illegitimately
influences science, (bi) explaining what the social value in your example is [1 point], (bii)
what its influence on science is [1 point], and (biii) in what sense this influence is
illegitimate [2 points].

Answer 41

B) Bi) An example given to us in the course in which some specific value influences scientific
outcomes is Big Oil. Big Oil is a term used for the few biggest oil companies around the world. These
companies have had enormous amounts of power and money at their disposal for decades. They
funded active lobbies and research against climate change. Scientists, funded by Big Oil, were tasked
with research proving that oil was not as polluting as thought by the general public. This is an
example of the so called ‘funding bias’, wherein a scientist, hired by an institution, has to prove
something that is in the interest of the institution.

Bii) The social value in this example for both the scientists and the big oil companies is money, by
proving the big oil companies right, the scientists get paid and the companies get to continue their
profitable operations. These biases have a negative impact on the integrity of science as a whole, as
it brings in the illegitimate value of money and profit in the process of research. This value is
illegitimate because it makes the researchers biased, they set out with the goal to prove that
something is true, and not to research whether something is true or not. Because of this, scientists
might be tempted to employ tactics like p-hacking and image manipulation to prove their point. In a
world where we would adhere to the value-free ideal of science, this would not happen. The
research would still be done, but in an unbiased manner. The researchers would not have any
incentives to do the research any other way but to research the topic without any influence from
anyone.

Question 42

(a) Define and illustrate with examples covered in our course the nomological, pattern
based, and causal conceptions of explanation [5 points].

Answer 42

A) According to the nomological conception of explanation, a scientific explanation mentions a law that can be relevant for the explanation of a phenomenon. Based on the pattern concept, explanations fit particular statements about phenomena into a more general framework of laws and principles. According to the causal conception, explanations appeal to causes that bring about the phenomenon to be explained. For instance, a rising sea level is explained by climate change. Glaciers and ice sheets worldwide are melting and adding water to the ocean.

Question 43

nomological, pattern based, and causal conceptions of explanation [5 points]., (b) identify and critically
evaluate at least one problem for each of these three conceptions of explanation [5 points].

Answer 43

B) A key problem for the nomological conception is that a lot of adequate explanations are not relevant to
any laws. There have been cases where revolutionary explanations have been successful without appealing
to any laws. For instance, dinosaurs went extinct thousands of years ago due to a volcano eruption or an asteroid hit the earth. Neither of those explanations is linked to a scientific law.

A key problem for the pattern conception is that it refers to explanations that fit a phenomenon into a wider pattern. But some explanations are highly specific. Explanations may be only relevant to certain situations. This normally includes events which solely occur once and therefore, the explanation is only used once. Consider the explanation for how the human heart pumps blood. This explanation may not apply to the function of the hearts of other kinds of organisms.

A key problem for causal explanations is that there is no consensus regarding the nature of causation.
Essentially, there are occasionally disagreements regarding whether a given explanation captures genuine causes, meaning whether a certain explanation causally explains a consequence.

Question 44

(a) Define diversity in your own words [1 point]. Focus on one of these three characteristics of people—ie, either their gender, skin colour, or political views—and

Answer 44

Diversity refers to the practice or quality of including or involving people from a range of different social
and ethnic backgrounds and of different genders, sexual orientations.

Question 45

(b) describe one step that you think could be taken to increase diversity in science with respect to that characteristic [3 points].

Answer 45

One step to increase diversity in science is to eliminate barriers that generate discrimination against
women in scientific careers and decision-making. This can include formal laws, but also combating
negative stereotypes, which aim to maintain gender roles. This is particularly important since researchers
in science may be biased, so they might pick for a sample solely people that are similar to them, meaning
with similar demographic characteristics. With a higher share of women as researchers, there might be a
higher chance of more

Question 46

do you think that scientists’ diversity in that characteristic would contribute positively, negatively or would make no difference to their pursuit of knowledge? Give one argument in support of your answer and use one example from any field of science you’re familiar with to illustrate your position [6 points].

Answer 46

Certainly, scientists’ diversity in that aspect is valuable. Scientific research could be of higher accuracy
when gender is taken into consideration, especially in samples. Not addressing sex and gender in
investigation may limit the advantages of contemporary science. This is due to the fact that something which might be applicable to men, for instance, might not be applicable to women. The majority of scientific research doesn’t consider women, since men are considered the default, which may result in results which are applicable to only men, not women. This could include medication testing. Men might experience certain or no side effects based on the trials. On the other hand, women might not experience or experience other side effects. The conclusions of such trials, basically, could be limited since an overrepresentation of men was present. Therefore, the research might not be relevant for women.