CAUSALITY

1
CAUSALITY

• A causal claim df. A claim which says or implies
  that one thing causes another.
• A causal hypothesis df. A causal claim offered
  to explain the cause or effect of something, when
  the cause or effect hasnt yet been conclusively
  established.
• A causal argument df. An attempt to support a
  causal claim or hypothesis.
• (See examples of causal claims on page 420.
  Notice how one claim is put in negative form thus
  denying that one thing has a causal relation to
  another.)

2
ONLY-RELEVANT-DIFFERENCE I

• Only-relevant-difference reasoning makes the
  following points as it concerns reaching a
  conclusion that one specific event or occurrence
  caused some other specific event or occurrence
• 1. One item has a feature that other similar
  items or things of the same kind lack. This is
  called the feature in question.
• 2. There is only one other relevant difference
  between the thing that has the feature in
  question and the other items that dont have the
  feature in question.
• 3. Therefore, that relevant difference is the
  cause of the feature in question.

3
EXAMPLE I
Similar things, or things of the same kind.
One thing.
Another thing of the same kind. (Side-by-side
comparison.)
One cut rose lasted longer than another cut rose.
The longer-lasting rose was treated with
aspirin. Therefore, the aspirin caused the first
rose to last longer.
The feature in question.
The cause.
The only relevant difference. (The two roses were
cut from the same bush, put in the same kind of
container, given water from the same source, put
in the same location so as to get the same light,
etc.)
4
EXAMPLE II
Similar things, or things of the same kind.
One thing.
The same thing at a later time after a change, or
similar things which are different stages of the
same thing. (Before-after comparison.)
Janes computer used to take a couple of minutes
to save large files, particularly those with
graphics. Then she doubled the size of the
computers RAM, and the same files now take about
one-fourth the time to save. Therefore the
cause of the faster time of saving files was due
to increasing the computers RAM.
Only relevant difference.
The cause.
The feature in question faster saving time.
5
ONLY-RELEVANT-DIFFERENCE II

• Only-relevant-difference reasoning requires at
  least two things to compare.
• The things compared must be the same except for
• 1. The feature in question that we want to know
  the cause of (e.g. the longer life of the rose,
  the faster saving time of the computer) and
• 2. The difference that produced the feature (e.g.
  the aspirin, the increased RAM).

6
ONLY-RELEVANT-DIFFERENCE III

• Note the term only in only-relevant-difference.
  For this kind of reasoning to work, it must be
  the case that what is identified as the cause of
  the feature in question is the only relevant
  difference, and not just a relevant difference
  amongst others that may contribute to the
  feature.
• MP Critical thinking requires consideration of
  important alternative differences that may have
  been overlooked.
• For instance, The rose with aspirin might have
  come from a hardier bush, or it might have been
  fresher to begin with.

7
ONLY-RELEVANT-DIFFERENCE IV

• MP Only-relevant-difference reasoning can be
  conclusive at least in experimental conditions
  (where things are very carefully controlled).
  (See the example on page 422.)
• MP Even apart from carefully controlled
  experimental conditions, this pattern of
  reasoning can yield conclusions that are certain
  by everyday standards. (See the example on page
  423.)
• However, Often we cannot be certain that the
  difference in question is the only relevant
  difference between the cases we are comparing.

8
RELEVANCE I

• A difference is relevant when it is not
  unreasonable to suppose that the difference might
  have caused the feature in question.
• Thus it is not unreasonable to suppose that the
  aspirin might have caused the rose to last
  longer, and it is certainly not unreasonable to
  suppose that the memory boost caused the computer
  to save files faster.
• MP The more you know about a subject, the
  better you are to whether a given difference is
  relevant to a feature in question.

9
RELEVANCE II

• If you know nothing about anything then you are
  not in a position to determine if a difference is
  relevant to causing the feature in question.
• Say, for instance, that when Janes computer
  received additional memory that the technician
  cleaned the casing of the computer. If you know
  nothing about anything then you could not
  conclude that cleaning the casing was not a
  relevant difference which might be the cause of
  the faster saving time of the computer.
• MP However, it is not the mark of a critical
  thinker to pretend to know nothing about
  anything.

10
RELEVANCE III

• MP To say that a factor is relevant is only to
  say that it is not unreasonable to suppose that
  it caused some feature.
• It is not unreasonable to suppose that increasing
  the RAM of a computer increases the speed with
  which it saves files. It is unreasonable to
  suppose that cleaning the computers container
  increased that speed.
• MP If you learn that, in fact, it did not
  cause the feature, that doesnt necessarily mean
  you were mistaken to think it relevant. Even
  though it did not cause the feature, it might not
  have been unreasonable to suppose that it did.
• Maybe the aspirin did not cause one rose to last
  longer than the other, maybe it had been
  fertilized before being cut and the other had
  not. Even so, it was not unreasonable to suppose
  that the aspirin was the cause of the feature in
  question.

11
ONLY-RELEVANT-COMMON-THREAD I

• Only-relevant-common-thread reasoning
  (common-thread reasoning) concerns multiple
  occurrences of something.
• For instance, a number of trees in a forest die
  at the same time. When we try to determine what
  caused the trees to die we look for the common
  thread which caused the feature in question (the
  trees dying).
• The pattern of only-relevant-common-thread
  reasoning is
• 1. Multiple occurrences of a feature (the feature
  in question) are united by a single relevant
  common thread (the common thread in question).
• 2. Therefore, the common thread in question is
  the cause of the feature in question.

12
ONLY-RELEVANT-COMMON-THREAD II

• It is possible that more than one common thread
  exists. In that case there is more than one
  possible cause.
• For instance, one common thread could be drought,
  another could be damaging winds, a third could be
  disease.
• MP Common-thread reasoning is best for forming
  hypotheses which are to be tested in some other
  way, usually through experimentation involving
  relevant-difference reasoning.
• This is because multiple occurrences of a
  feature in question could always have resulted
  from different causes.
• Different instances of the multiple occurrences
  of tree death could have had different causes
  one tree dies from drought, another from wind,
  and a third from disease.

13
ONLY-RELEVANT-COMMON-THREAD III

• Common-thread reasoning is also best suited for
  forming hypotheses because even if we are
  dealing with multiple occurrences of an effect
  somehow known to have been caused by one and the
  same thing, these multiple occurrences are likely
  to have many other things in common besides
  whatever it was that caused them.
• And common-thread reasoning cannot by itself
  tell us which of these common things was the
  actual cause of the multiple occurrences of the
  effect.
• Thus the trees in the forest which died (the
  multiple occurrences) likely had many things in
  common. Because of this common-thread is better
  for forming a hypotheses about the cause of death
  which might be identified by only-relevant-differe
  nce reasoning.

14
MISTAKES IN RELEVANT- DIFFERENCE REASONING I

• 1. The difference taken to be the cause of the
  feature in question might not be a relevant
  difference.
• (Recall that a difference is relevant when it is
  not unreasonable to think that it played a role
  in causing the feature in question. One is not
  thereby committed to saying that it definitely
  did play a role in causing the feature in
  question.)
• For instance, thinking that what caused the
  deaths of the trees was the noise from a new
  construction site nearby.

15
MISTAKES IN RELEVANT- DIFFERENCE REASONING II

• 2. The difference thought to have caused the
  feature in question might be a relevant
  difference, but might not be the only relevant
  difference.
• (Again a difference is relevant when it is not
  unreasonable to think that it played a role in
  causing the feature in question. However, that
  does not necessarily mean that it did play a
  role.)
• MP If the major difference is not the only
  relevant difference, then it may not be the cause
  of the feature in question.
• Perhaps a farmers crop production is much better
  this year than last year. And perhaps the major
  difference between this year and last year is
  more rain. But more rain may not be the only
  relevant difference perhaps there was more sun
  too, and the farmer used a new brand of
  fertilizer. Then more rain might not be the cause
  of the better crops.

16
MISTAKES IN RELEVANT- DIFFERENCE REASONING III

• 3. The difference being considered as the cause
  of something may in fact be the effect rather
  than the cause.
• For instance, George has not slept well the past
  couple of nights, and he has also been nervous
  the past couple of days. He may think that his
  not sleeping well (difference considered as
  cause) is the cause of his nervousness (the
  feature in question), while the truth is that his
  nervousness is the cause of his not sleeping
  well.
• 4. The difference might not cause the feature in
  question. Instead, the difference and the feature
  in question are each effects of a third
  underlying cause.
• It might be that both Georges nervousness and
  his insomnia are due to his feelings of guilt
  about something.

17
MISTAKES IN COMMON-THREAD REASONING I

• 1. Is the thread identified as common relevant to
  the feature in question?
• Suppose that five friends all get sick after
  eating Jeanies pot roast. A thread common to the
  group, in addition to eating the pot roast, might
  be that each just finished reading the same book.
  However, that would not seem to be relevant to
  their getting sick (the feature in question). And
  so that particular common thread should not be
  identified as the cause of the sickness.
• Better reasoning would be to suppose that eating
  the pot roast was the common thread which caused
  the multiple occurrences of illness.

18
MISTAKES IN COMMON-THREAD REASONING II

• 2. A difference taken to be the cause of the
  feature in question might be a relevant
  difference, but is not the only relevant
  difference.
• For instance, the friends who got sick after
  eating Jeanies pot roast might also have in
  common consuming a fair amount of red wine. The
  consumption of this wine might be relevant to
  their sickness, but may not be the only thing of
  relevance to their illness, since something about
  the pot roast may also be relevant.
• If there is more than one common thread, then any
  particular common thread focused on as the cause
  of the feature in question may only be one cause
  of the feature in question, and so may only
  partially explain it. (It may also have nothing
  to do with it.)

19
MISTAKES IN COMMON-THREAD REASONING III

• 3. Cause and effect may have been reversed.
• Perhaps a correlation is noted between education
  and values in that societies that spend a lot on
  education tend to have better values than those
  which do not. It is then hypothesized that
  spending more on education promotes better
  values. However, the situation may as a matter of
  fact be reversed. It may be that societies with
  better values spend more on education.
• 4. What is taken to be the common thread and the
  feature in question may have a common cause. (See
  the two examples on page 428.)

20
MISTAKES IN COMMON-THREAD REASONING IV

• 5. The feature in question might not require a
  common cause.
• Maybe the people who got ill after eating
  Jeanies pot roast all got ill for different
  reasons too much wine in one case, the flu in a
  second, an earlier lunch for a third, and so
  forth. If the causes of the common feature the
  illness were different in each case, then the
  common thread of eating Jeanies pot roast was
  not the cause of getting ill, but was merely
  coincidence.
• MP We should not unthinkingly assume that
  multiple occurrences of something have a common
  cause even if there is a common thread
  present.
• MP A common thread might induce us to assume
  that a single thing caused the feature in
  question, and that might be a mistake.

21
POST HOC, ERGO PROPTER HOC I

• Post hoc, ergo propter hoc is Latin for after
  this, therefore because of this.
• The fallacy of post hoc, ergo propter hoc df.
  Thinking that x causes y simply because y occurs
  after x.
• For instance, thinking that day causes night
  simply because night follows day, or thinking
  that because every time that the Yankees win the
  world series we have a cold winter that the cause
  of the cold winter is the Yankees winning the
  world series.

22
POST HOC, ERGO PROPTER HOC II

• MP It is a mistake to think that, just because
  y happened around the same time that x happened,
  that y happened because x happened, which is why
  post hoc, ergo propter hoc is a fallacy.
• It is true that effects follow their causes in
  time, and so it may be true that x both causes y
  and y follows x in time. It is just that, it is
  fallacious to assume a causal relation between x
  and y based only on the fact that one follows the
  other.
• You have good reason to suppose that x caused y
  if it is the only thing which accounts for y or
  if it is the best explanation of y. (See examples
  on pages 429-430.)

23
CAUSALITY VS. COINCIDENCE I

• The connection between events can be
  coincidental, not causal. Three common kinds of
  coincidence are
• 1. Two things might not be causally related at
  all, but are taken to be causally related.
• For instance, a man has a heart attack and dies
  after running. The running is taken to be the
  cause of the heart attack but in fact an autopsy
  shows that the running has nothing to do with it.
  Or people theorize that the cause of AIDS (rather
  than a cause of the spreading of AIDS) is sex.
  Both examples are post hoc reasoning.
• MP What underlies many superstitions is
  thinking of coincidental events as being causally
  related to one another.

24
CAUSALITY VS. COINCIDENCE II

• 2. Multiple occurrences of an effect are thought
  to be due to a common thread shared by all the
  occurrences when the truth is that some other
  common thread caused the occurrences.
• When this is the case then it is just
  coincidence that the first common thread is
  present.
• For instance, the people who got ill after eating
  Jeanies pot roast may have been together a day
  or two before at a party at which each was
  exposed to a common virus which took a couple of
  days to make them sick. Their shared feature of
  getting sick also had the common thread of eating
  at Jeanies, but that common thread is just
  coincidence, and, as such, is causally irrelevant.

25
CAUSALITY VS. COINCIDENCE III

• 3. It can be assumed that multiple occurrences of
  a particular effect are due to a common thread
  shared by all the occurrences when in fact the
  occurrences were not caused by a single thing.
• In the five people getting ill after eating at
  Jeanies, the multiple occurrences are the five
  cases of illness and the common thread assumed to
  be the cause of the illnesses is eating Jeanies
  pot roast. Eating Jeanies pot roast may be the
  common thread which is shared by all five
  illnesses, and so may be their cause, but it may
  also simply be a coincidence. The truth may be
  that each illness has a different cause.

26
CAUSATION IN POPULATIONS

• Some causal claims apply to populations rather
  than to individuals.
• For instance, Smoking causes cancer is meant to
  link smoking causally to cancer, not in any
  particular individual, but to smokers in general.
  Saying that smoking causes cancer means that, we
  would expect more cases of lung cancer in
  populations in which everyone smoked rather than
  in populations in which no one smoked.
• MP To say that X causes Y in population P is
  to say that there would be more cases of Y in
  population P if every member of P were exposed to
  X than if no member of P were exposed to X.

27
KNOWLEDGE OF CAUSES IN POPULATIONS

• How do we know, or what makes us think, that one
  thing is a cause of another? Or what is the
  evidence for the claim that there would be more
  cases of something (Y) in a population in which
  every person in the population were exposed to
  something (X) than if they were not?
• The first thing which argues in favor of one
  thing causing another in a population is
  controlled cause-to-effect experiments.

28
CONTROLLED CAUSE-TO-EFFECT EXPERIMENTS II

• Controlled cause-to-effect experiments involve
  randomly dividing a random sample of a target
  population into two groups an experimental group
  and a control group.
• In the experimental group all members of the
  group are exposed to something c which is
  suspected to cause something else. (For instance,
  being exposed to some mold which is thought to
  cause a certain allergy.)
• The members of the control group are not exposed
  to c. However, other than this difference,
  members of the control group are treated exactly
  the same as members of the experimental group.
  (Thus the members of this group are not exposed
  to the mold, but otherwise everything else is the
  same.)

29
CONTROLLED CAUSE-TO-EFFECT EXPERIMENTS III

• The the experimental and control groups are then
  compared with respect to frequency of some
  effect, e. (Here the occurrence of the allergy
  which the mold is thought to cause.)
• MP If the difference d, in the frequency of e
  in the two groups is sufficiently large, then c
  may justifiably be said to cause e in the
  population.
• Thus if 50 of 100 people in the experimental
  group get the allergy when exposed to the mold,
  and only 1 out of 100 people in the control group
  gets the allergy when not exposed to the mold,
  then the difference - 49 - may be said to be
  sufficiently large for the mold to cause the
  allergy in that group.

30
CONTROLLED CAUSE-TO-EFFECT EXPERIMENTS IV

• It is a question how large the difference d must
  be between the experimental group and the control
  group to say that a certain effect e is due to a
  certain cause c.
• This is determined in relation to the size of
  each group, say 100 people, at some approximate
  statistically significant level, say 0.05, which
  means that the result could have arisen by
  chance in about 5 cases out of 100.
• For instance, to think that e is due to c in a
  group of 100 people, the difference d between the
  number of people in the experimental group with e
  and those in the control group with e must exceed
  13. (For further see table on page 447, and note
  that, as the size of the population goes up, the
  figure that d must exceed goes down.)

31
CONTROLLED CAUSE-TO-EFFECT EXPERIMENTS V

• The sample of the population from which the
  members of both the experimental and control
  groups are taken should be representative of the
  target population.
• Accordingly, the sample used to construct each
  group should be taken at random.
• In addition, the assignment of a member of the
  sample to either the experimental or the control
  group should also be done at random.

32
SAMPLE SIZE

• One should not automatically assume that a sample
  size in controlled experiments is large enough
  to guarantee significance.
• MP A large sample is no guarantee that the
  difference (d) in the frequency of the effect in
  the experimental group and in the control group
  is statistically significant a particular
  experiment may not achieve statistically
  significant results even with a large sample
  size.
• MP However, the larger the sample, the smaller
  d expressed as a difference in percentage
  points need be to count as significant. (Refer
  to table on page 447.)

33
FREQUENCY DIFFERENCE

• One must also not automatically assume that the
  difference in frequency between experimental
  group and control group of an effect being
  investigated is great enough to guarantee
  significance.
• A difference may seem great that is not really
  statistically significant.
• For instance, If there are 50 rats in an
  experimental group and 50 more in a control
  group, then even if the frequency of skin cancer
  found in the experimental group exceeds the
  frequency of skin cancer found in the control
  group by as much as 18 percentage points, say 2
  rats in the control group and 11 in the
  experimental group got cancer this finding would
  not be statistically significant (at the 0.05
  level). (This is 9 rats in 50, and where we
  would expect that 2 rats 0.05 might get cancer
  by chance.)

34
ANALOGICAL EXTENSION

• MP The results of controlled experiments are
  often extended analogically from the target
  population (e.g. rats) to another population
  (e.g. humans).
• We need to be careful here, since we would need
  to know how representative the rats used in each
  group are of all rats how many features relevant
  to the experiments conclusion human beings have
  in common with rats so that the conclusion of the
  rat study can be reasonably applied to humans
  and whether there are important relevant
  differences between the target population in the
  experiment the rats and the population to which
  the results of the experiment are analogically
  extended humans.

35
REPUTABLE SOURCES

• MP In reputable scientific experiments it is
  safe to assume that randomization of the sample
  from the target, and the division of members of
  the sample into the two groups control and
  experimental has been employed, but one must be
  suspicious of informal experiments in which no
  mention of randomization is made.
• MP Any outfit can call itself the Cambridge
  Institute for Psychological Studies and publish
  its own journal. However, such an outfit
  could consist of little more than a couple of
  university dropouts with a dubious theory and an
  axe to grind.

36
NON-EXPERIMENTAL CAUSE-TO-EFFECT STUDIES I

• In a non-experimental cause-to-effect study
  members of a target population, such as humans,
  who have not yet shown evidence of a suspected
  effect e, such as allergic reaction, are divided
  into two groups that are alike in all respects
  but one.
• MP The difference is that members of one
  group, the experimental group, have all been
  exposed to the suspected cause c (e.g. mold),
  whereas the members of the other group, the
  control group, have not.

37
NON-EXPERIMENTAL CAUSE-TO-EFFECT STUDIES II

• The difference between non-experimental
  cause-to-effect studies and controlled
  cause-to-effect experiments is that the members
  of the experimental group are not exposed to the
  suspected causal agent by the investigators.
• This is because there are limits to what is
  ethically acceptable to expose humans to (and
  some other creatures?). We cant purposely
  expose human subjects to potentially dangerous
  agents.
• For instance, if smoking is suspected of causing
  cancer, we would not make people smoke to see if
  it causes cancer. What we would do is select from
  a group of people who voluntarily smoke and
  compare them with people who are otherwise like
  them who do not smoke.

38
NON-EXPERIMENTAL CAUSE-TO-EFFECT STUDIES III

• Just as with controlled experiments, experimental
  and control groups in non-experimental studies
  are compared to see how often the effect being
  looked for appears in each group.
• MP If the frequency in the experimental group
  exceeds the frequency in the control group by a
  statistically significant margin, then we may
  conclude that c is the cause of e in the target
  population.
• For instance, if more people get lung cancer in
  the experimental group by a statistically
  significant margin than do those in the control
  group, then we can conclude that smoking causes
  cancer.

39
NON-EXPERIMENTAL CAUSE-TO-EFFECT STUDIES IV

• MP Non-experimental studies are inherently
  weaker than controlled experiments as arguments
  for causal claims.
• One reason for this is that members of the
  experimental group who have already been exposed
  to the cause c in question may differ in some
  respect r from the control group in addition to
  c. And it may be that r is a cause of the effect
  being looked for.
• For instance, we may think that a fatty diet c
  causes colon cancer e. Members of the
  experimental group will have a fatty diet and
  those of the control group will not. But even if
  the experimental group has a statistically
  significant higher rate of colon cancer, it may
  be due to a different or additional factor, such
  as heavy drinking, since many people who eat
  fatty foods also consume too much alcohol.

40
NON-EXPERIMENTAL CAUSE-TO-EFFECT STUDIES V

• Another problem is incomplete knowledge of the
  causal complexity of nature and human reality.
• MP Because we do not have complete knowledge
  of what factors are causally related to what
  other factors, it is impossible to say for
  certain that all possibly relevant variables in
  such studies have been controlled.

41
NON-EXPERIMENTAL EFFECT-TO-CAUSE STUDIES I

• In a non-experimental effect-to-cause study the
  experimental group already have the effect e
  such as lung cancer being investigated.
• The experimental group is then compared with a
  control group none of the members of which have
  e, and the frequency of the suspected cause c
  e.g. smoking cigarettes is measured.
• MP If the frequency of c in the experimental
  group significantly exceeds its frequency in the
  control group, then c may be said to cause e in
  the target population.

42
NON-EXPERIMENTAL EFFECT-TO-CAUSE STUDIES II

• One cannot simply assume that, because the sample
  size seems large enough, or because the
  difference in frequency seems significant, the
  findings of the study are significant.
• One must also be careful about extending the
  results of a non-experimental effect-to-cause
  study analogically to other populations.
• For instance, we must be careful in concluding
  from a group study of something that caused a
  certain reaction in men that the same thing will
  cause the same thing in women since there may be
  relevant differences between the sexes which
  would interfere with the expected result.

43
NON-EXPERIMENTAL EFFECT-TO-CAUSE STUDIES III

• It must also be noted that the subjects in the
  experimental group may differ in some important
  way (in addition to showing the effect such as
  lung cancer) from the rest of the target
  population. (See the example on page 450.)
• MP Any factor that might bias the experimental
  group in such studies should be controlled. If,
  in evaluating such a study, you can think of any
  factor that has not been controlled, you can
  regard the study as having failed to demonstrate
  causation.

44
NON-EXPERIMENTAL EFFECT-TO-CAUSE STUDIES IV

• MP Effect-to-cause studies show only the
  probable frequency of the cause, not the effect,
  and thus provide no grounds for estimating the
  percentage of the target population that would be
  affected if everyone in it were exposed to the
  cause.

45
APPEAL TO ANECDOTAL EVIDENCE I

• If we try to argue that, because we know of a
  case or two in which one thing caused another,
  then we are appealing to anecdotal evidence in
  accounting for the causal relation.
• For instance, we might say that aunt Velma has
  begun each day with a shot of whiskey for the
  last 60 years, and then conclude that drinking a
  shot of whiskey in the morning makes you live
  longer.
• MP To establish that x is a causal factor for
  y we have to show that there would be more cases
  of y if everyone did x than if no one did, and
  you cant really show this or demonstrate that
  x isnt a causal factor for y see next slide
  by citing an example or two.

46
APPEAL TO ANECDOTAL EVIDENCE II

• If, on the other hand, we try to argue that,
  because we know of a case or two in which one
  thing failed to cause another, then we are
  appealing to anecdotal evidence in accounting for
  the lack of causal relation.
• For instance, a person might maintain that heavy
  drinking of alcohol does not cause liver damage
  since his uncle drank 10 beers a day for 60 years
  and died from getting hit by a kid on a
  skateboard. And his autopsy showed that his liver
  was normal for a man his age.