Causality 400 864 RMQQ - Lectures on Causalty PDF

Preview

Summary

Lectures on Causalty...

Description

400864 Research Methods Causality study notes These study notes examine the concept of causality and what’s meant when we say that one event causes another. They look at how we decide whether one thing causes another, whether and how we can be sure that events really are causally joined, and the different ways that one thing may cause another. The relevance to healthcare and evidence-base practice is demonstrated throughout. Knowledge goals •

Terminology: causality, causal relationship, causal inference, epistemology;

•

Hume’s theory of knowledge and causation;

•

Mills theory of causation;

•

Types of causal relationships;

•

Relevance of causal inferences to evidence-based healthcare practice and policy.

1

“All for the want of a horseshoe nail” – a nursery rhyme •

For want of a nail, the horseshoe was lost

•

For want of a horseshoe, the horse was lost

•

For want of a horse, the rider was lost

•

For want of a rider, the army was lost

•

For want of an army, the battle was lost

•

For want of a battle, the war was lost

•

For want of a war, the kingdom was lost

•

And all for the want of a horseshoe nail

Did the horseshoe nail cause the kingdom to be lost?

For Western Sydney University classes taught by Dr John Bidewell

1

Causality

The nursery rhyme is traditional from the Mother Goose collection for children. It shows how one event can lead to other events. Even a minor event can make something big happen. When one event makes another event happen, we say that the first event causes the second event. When we see this happen, we observe a causal relationship or causality. A chain of causal events is possible, with one event leading to a set of events because of it. Relevance to healthcare Identifying causal relationships is important for healthcare because we want to know: • What causes health disorders, so we can stop the cause and prevent the disorders or at least reduce their risk. • Whether treatments cause improvements in patients’ health, so we know which treatments to use. • Whether treatments cause adverse side effects, so we better understand the risks of some treatments. For the same reason, casual relationships are important for evidence-based practice because we want reliable and valid evidence about: • The causes of health conditions and disorders. • The effectiveness of treatments. • Adverse effects of treatments.

2

Fooling ourselves about causality

Sometimes we mistakenly believe that one event causes another. Something may appear to cause something else when it fact it doesn’t. These mistakes can happen in health care. •

We might believe that something causes a health disorder when it doesn’t.

•

We might believe that our treatments work when they don’t. If the patient improves after treatment, it could be because of something else happening in the background that we don’t know about. It’s easy to be fooled by appearances.

400864 Research Methods (Quantitative and Qualitative)

2

Causality

As well as needing to know when one thing causes another, we need to know when one thing doesn’t cause another. It’s hard to be certain about causality. Is it correct to say that: •

The sun causes day?

•

An iceberg caused the Titanic to sink?

•

Gravity causes objects to fall?

•

Mosquitoes cause malaria?

•

Smoking causes lung cancer?

•

Advertising causes people to buy things?

•

Fatty food causes obesity?

•

Petrol causes a car to move?

•

Drink driving causes road crashes?

•

The assassination of Archduke Ferdinand in Sarajevo in 1914 caused World War I, as historians often claim. The same historians may identify other factors that were responsible for the conflict. The assassination merely started the war.

Health researchers are interested in causality for the same reasons that health practitioners are. Health practitioners want to be sure that: •

Causes of health disorders are known so that the disorders can be prevented by stopping the cause.

•

Treatments cause only beneficial effects, without adverse side effects or other unwanted results.

To be sure that such claims are true, empirical evidence is needed, but it has to be good quality evidence and the evidence must be correctly interpreted.

3

Cause and effect in everyday life Cause and effect happens in everyday life as well as health care practice. In the

news we often hear causal statements, for example, that increasing atmospheric carbon dioxide from human activity is causing global warming, which will cause environmental, economic and social problems if we don’t prevent it. We also make causality happen. Every time we operate an on-off switch we are causing something happen. When we make something happening, we say that we are causing it. 400864 Research Methods (Quantitative and Qualitative)

3

Causality

In healthcare, as in everyday life, we believe that we understand causal relationships more more than we really do. We’re over-confident about causality. It’s too easy to assume causality from casual observation. It’s tempting to believe that one thing causes another whether or not there is good evidence for it. Establishing whether one thing really does cause another can be surprisingly difficult.

4

Causal inferences

Claiming that one thing causes another (such as a treatment helping the patient) requires more than just observation. It requires our personal decision that a causal relationship exists. This decision is known as a causal inference. To understand this term, it’s necessary know what’s meant by the term inference. •

An inference is a decision or a conclusion that something is true or has happened, based on some evidence or observation.  If we arrive at the car-park to find an empty space where we parked our car, then based on that observation we might infer that the car has been stolen.  If someone nearby reports having seen someone loitering near our car, then from that information we might infer that the loiterer drove away in our car.

•

A causal inference is a conclusion that one event causes another. If we treat a patient and observe that the patient recovers, we could infer that the treatment caused the improvement. That inference will be a causal inference.

Every time we conclude that a treatment improves the patient’s condition, or that we know what brought about the patient’s condition, we are making a causal inference. Thus, causal inferences are common in healthcare. The question is, when, if ever, can we trust causal inferences to be true? How can we be sure that our causal inferences are valid? To answer these questions, we have to think about what it means to make something happen. Fortunately, a lot of helpful thinking has already been done.

400864 Research Methods (Quantitative and Qualitative)

4

Causality

5

Theories of cause and effect A theory is an explanation about how or why something

happens. For centuries, philosophers have written theories about causality. We’ll start with the Scottish philosopher, David Hume (1711-1776), who is regarded as one of the greatest Western philosophers partly because of his work on causality. Hume’s ideas remain influential hundreds of years after he wrote them. His ideas about causality can affect how we think about healthcare practice and policy.

5.1

David Hume

Hume’s epistemology

Hume had many interests. He wrote about morality (ethics), politics and religion. He is well known for his writing about knowledge, or more technically, epistemology, which is the philosophy of knowledge. Epistemology is about what knowledge actually is. The questions epistemologists ask are these: •

How do we know things?

•

How can we truly know anything?

•

How can we be certain that what we know is correct?

•

Can we be certain about anything?

The opinion that real knowledge is impossible, that we can never be sure of anything, is called scepticism. A sceptic is a person who doubts whether something, or anything, is true. Hume has a reputation as a sceptic but we’ll see that he wasn’t a total sceptic. He allows for knowledge but within limits. Those limits are relevant for researchers and health practitioners today. Extreme scepticism, the belief that we can never truly know anything, is a useless idea in healthcare and everyday life. If possible, we’d like to think we know something. The question is, how to avoid scepticism. Hume deals with scepticism by looking at the different types of knowledge, their strengths and limitations. Hume’s two major works are A Treatise of Human Nature (1739, 1740) and An Enquiry Concerning Human Understanding” (1748, 1777). Hume’s ideas on causality are contained in these two major works. The Treatise is the main work but it was unsuccessful when published, so Hume summarised it in the Inquiry. Both works are still considered important. Both are readily available free on the Internet but you don’t have to read them. Hume’s epistemological writing includes a theory of

400864 Research Methods (Quantitative and Qualitative)

5

Causality

cause and effect. That theory describes what in our everyday lives we mean by the term causality, and whether we can prove causal relationships to be true. 5.1.1

“Hume’s Fork"

Hume proposed two types of knowledge, now known as Hume’s Fork. The term fork is appropriate because a fork is uncomfortable to sit on. Hume’s Fork has uncomfortable implications. Unlike most forks, but like the fork illustrated here, Hume’s Fork has only two prongs. Each prong refers to a type of knowledge. The two types of knowledge, according to David Hume, are: •

Relations between ideas.

•

Matters of fact. 5.1.1.1

Relations between ideas

Relations between ideas involve only our thoughts. Relations between ideas have nothing to do with observation or experience. They have nothing to do with the external world that we understand through our senses. They are not empirical. We do not see, hear or feel relations between ideas. We can only think about them. Examples of relations between ideas include geometry, mathematics and logic exercises such as chess and sodoku. Relations between ideas can be true or false. They are proved true or false always by reasoning and logic and not by observation or experiment. The statement, “A triangle has three sides,” must be true because it’s a logical fact that doesn’t need checking in the real world. A four-sided triangle is logically impossible so there is no point travelling around the word looking for one. A triangle having three sides is true for all triangles, past, present and future. The problem with relations between ideas is that they say nothing about whether one thing causes another. A health professional might say, “A virus causes the common cold.” Even if the statement is true, and viruses really do cause the common cold, that is not a logical truth in the same way as triangles having three sides is a logical truth. According Hume, the causes of diseases, and whether or not our healthcare treatments work would not be logical issues. They are not relations between ideas. Instead they belong to the other category of knowledge. To summarise: Relations between ideas are about logical truths only, and say nothing about cause and effect in the physical world.

400864 Research Methods (Quantitative and Qualitative)

6

Causality

5.1.1.2

Matters of fact

According to Hume, matters of fact are the other type of knowledge. Unlike relations between ideas, matters of fact do concern the real, outside physical world that we see, hear and touch. Matters of fact consist of knowledge from observation and experience, also called empirical knowledge. The statement, “A clover has three leaves,” is matter of fact because it is based on observation and experience. We can prove or disprove three-leafed clovers by travelling the world, finding clover patches and counting leaves. A clover having three leaves is an empirical truth, not a logical truth. Unlike with the four-sided triangle, it is perfectly possible to imagine a four-leafed clover. There is nothing illogical about a four-leafed clover. We can even picture of a fourleaved clover. It’s just that we’ll never find a four-leaved clover in the real world. That the world is apparently without fourleafed clovers is an empirical rather than a logical fact. Notice how questions about the number of leaves on a clover are answered by real-world evidence whereas questions about the number of sides on a triangle are answered in our heads. Matters of fact are supported by observation and experience. Matters of fact also refer to causality, which is Hume’s big idea. Saying that a fatty diet can clog a person’s blood vessels, or that lack of exercise combined with excessive weight can lead to Type 2 diabetes are matters of fact about whether one thing causes another. These claims are testable from real-world evidence rather than just thinking about them. To summarise: •

Relations between ideas are about reasoning, not observation and experience. We don’t have to go around looking at triangles and counting their sides to know that all triangles have only three sides. That is a logical rather than an empirical truth, and it has nothing to do with cause and effect.

•

Matters of fact are about observing and experiencing the physical world. They include statements of cause and effect. Someone does have to go around collecting and analysing real-world data to show that:  Excess energy in a person’s diet causes them to increase weight.  Excess weight causes chronic health conditions.

400864 Research Methods (Quantitative and Qualitative)

7

Causality

 An exercise program combined with an energy-controlled diet will cause a person to lose weight. Questions about the causes and treatments of disease and other health conditions are empirical. They are answered by real-world observation, by research evidence and not by reasoning alone. That is the basis for evidence-based practice and the Western medical tradition over the last several hundred years. 5.1.2

Hume’s principles for inferring causes

Hume went on to question whether matters of fact can be proved in the same way that relations between ideas can be proved. He described the problem in terms of billiard balls on a billiard table. We can observe billiard balls striking each other and causing movement. One ball hits another ball and seems to make it move with a predictable speed and direction. These observations are matters of fact. They are empirical. We infer that the motion of one ball causes the motion of another ball, another matter of fact. From empirical observations with billiard balls, Hume devised principles for inferring causality. These principles describe how we decide that one event causes another. According to Hume, we follow three rules for inferring that one event causes another. 1. Contiguity in time and space – the cause and the effect share the same time and place. When one ball hits another and causes the second to move, both balls are on the same billiard table at the same time. 2. Causes precede effects – the cause must happen just before its effect. The first ball must hit the second ball a moment before the second ball moves. 3. Constant conjunction – the observed relationship between the cause and effect happens consistently over repeated observations. The cause and the effect are always observed to occur together in the same way. The motion of billiard balls is predictable. One ball hitting another will have the same effect on the speed and direction of the first ball is the same each time that the balls collide in the same way. If that were wrong, no one could play billiards, snooker or pool! Thus, according to Hume, we infer that the first event causes a second event: •

If the two events happen in the same time and place and;

400864 Research Methods (Quantitative and Qualitative)

8

Causality

•

If the first event happens immediately before the second and;

•

If the two events always happen together.

We can easily convert these principles into healthcare scenarios. For deciding that that a treatment cures a patient, Hume’s rules about causal inference will apply when: •

The treatment and cure happen together. We say that a treatment works when we treat a patient and he or she improves soon afterwards. We are unlikely to say that a treatment works when the patient’s condition only starts to improve the next year, or instead a different patient on the other side of the world gets better.

•

The treatment has to be given before the patient improves It would be incredible to say that a treatment can be effective before it has been tried. If a patient improves before you treat them, you can’t claim your treatment as the cause the benefit.

•

We expect the same improvements to occur when the treatment is given to patients under similar conditions. If treatment outcomes were unpredictable, that would be very worrying. Think about it: If anything could happen to a patient after receiving treatment, and anything could cause any illness, then healthcare would become impossible. Professional healthcare is possible only because causes and effects are predictable.

Repeated observation reveals cause and effect and suggests that causal relationships work reliably. If sailing rough seas has made us seasick in the past, we can expect it to make us seasick in future. More happily, we can expect our cars to start next time we turn the ignition key because our cars have started when we turned the ignition on previous occasions, except when something else like a flat battery prevents the car from starting. Relevance to healthcare Evidence-based practice uses reasoning similar to Hume’s. If enough participants in a randomised controlled trial improve soon after receiving treatment, we’ll infer that the treatment works and assume that the treatment caused the improvement. We shall expect the treatment to work for other patients from the same population in future. We shall conclude that the treatment caused the beneficial change. Likewise, if people who smoke have a consistently much higher rate of respiratory disorders than people who don’t smoke, we shall conclude that smoking causes respiratory disorders.

400864 Research Methods (Quantitative and Qualitative)

9

Causality

5.1.3

Proving causal inferences

It would useful if we could prove statements about one thing causing another. Then we would know for sure that treatments cause beneficial effects or harmful side effects. Hume has plenty to say about that. He described the problem of whether the past predicts the future. Just because “constant conjunction” has always happened in the past doesn’t logically show it will always happen in future. Causal inferences t...