| Part of the series on|
Logic and rhetoric
Causality in its simplest form is the relationship between cause and effect. In science, one's main objective is to find causal relationships, or in simpler terms "This causes that". Most of modern science is based on causal relationships and they are the core pillar of good science. The old mantra "correlation does not imply causation." is often what separates the science from the pseudoscience, the scientists from the cranks, and the medicine from the bullshit; causality is the glue that holds rational thought together..
History of causality
One of the first quotes about the concept of causality comes from Plato.
“”Now everything that becomes or is created must of necessity be created by some cause, for without a cause nothing can be created.
|—Timaeus by Plato|
Aristotle would expand upon Plato's idea of causality in Physics and Metaphysics where he argued that there were four causes, namely the material, formal, efficient, and final cause. Thomas Aquinas would later argue that "from every effect the existence of the cause can be clearly demonstrated, and so we can demonstrate the existence of God from His effects" and the Kalam cosmological argument uses a similar line of reasoning. Both Aquinas's argument and the Kalam cosmological argument expand upon Plato's belief that everything has a cause which is a philosophical position called universal causality. Whether universal causality is true is debatable. The philosopher Wes Morriston offers a detailed counterargument in his paper Must the beginning of the universe have a personal cause? where he argues that "when applied to the beginning of time, the principle that whatever begins to exist must have a cause is not clearly true."
Wes Morriston is hardly the only skeptic when it comes to causality. David Hume was chief among the philosophers to challenge the nature of causality promoted by the Ancient Greeks, Aquinas, and Aquinas's scholastic colleagues. He called into question the very ability of the human mind to understand it.
“”We have sought in vain for an idea of power or necessary connexion in all the sources from which we could suppose it to be derived. It appears that, in single instances of the operation of bodies, we never can, by our utmost scrutiny, discover any thing but one event following another, without being able to comprehend any force or power by which the cause operates, or any connexion between it and its supposed effect... All events seem entirely loose and separate. One event follows another; but we never can observe any tie between them. They seem conjoined, but never connected. And as we can have no idea of any thing which never appeared to our outward sense or inward sentiment, the necessary conclusion seems to be that we have no idea of connexion or power at all, and that these words are absolutely, without any meaning, when employed either in philosophical reasonings or common life.
|—Of the Idea of necessary Connexion Part II by David Hume|
But as skeptical as he was, Hume did not believe causal relationships were beyond rational comprehension. Instead he concluded that humans were able to induce cause and effect relationships through careful observation of contiguity, succession, and constant conjunction. He would also formulate the problem of induction when he noted that "our reason fails us in the discovery of the ultimate connexion of causes and effects" as it is "impossible for us to satisfy ourselves by our reason, why we shou’d extend that experience beyond those particular instances, which have fallen under our observation."
In more modern times, the discovery that time is relative has had profound implications for the understanding of causality. In physics, the space time interval is what defines the flow of causality and reversing its direction causes the direction of causality to be reversed. For things going slower than the speed of light, such a reversal is impossible unless one gets sucked into a black hole.
Causality in science
Causal inference is the process in which someone can use data to claim there is a causal relationship. This is central to most of science, and it is literally science at its core. Some people seem to forget about the part where the data has to support a causal relationship and not just a correlation between the data points. Causal inference is often very important in statistical data, as you are taking a large pre-existing dataset to come to a conclusion, and not a controlled test environment.
- Time precedence — Time precedence refers to the notion that for something to be considered a cause, the cause must have occurred before the effect. For example, if one suspects that hot temperatures are a cause of sweating, hot temperatures must have been present before the onset of sweating for it to be considered a cause, not after. Variable X simply being before effect Y is not sufficient for X to be considered a cause of Y and a failure to recognize this fact can lead to the post hoc, ergo propter hoc fallacy.
- Relationship — For X and Y to be in a relationship, they must be variables, not constants, and they must have codependency, not independence. In our sweating example, both hot temperatures (variable X) and sweating (effect Y) are variables, and they are codependent as people exposed to hot temperatures (X) should sweat (Y) more often than people exposed to cold temperatures. As temperatures above the mean (X) would be expected to induce sweating above the mean (Y), hot temperature and sweating are said to have positive covariance. If instead variables X and Y have a negative covariance such as cold temperatures and sweating, an entity above the mean for variable X would be expected to be below the mean for variable Y. If X and Y have a covariance of zero, then X is not predictive of Y and vice versa. Variables with a covariance of zero are independent of one another.
- Non-spurious — Even if variable X has both relationship and time-precedence with effect Y, this still does not mean a causal interconnection has been established as the interaction between X and Y must also be non-spurious. For example, if ice cream eating and pool attendance have positive covariance and ice-cream eating occurred before going to the pool, this does not mean that ice cream eating caused one to go to the pool. Instead both X and Y could be caused by the third variable of hot temperature (Z) and therefore the interconnection between X and Y is spurious and not causal. Failure to recognize spurious interactions can lead to the cum hoc ergo propter hoc fallacy, a.k.a. correlation does not imply causation.
In Multilevel modeling of social problems: A causal perspective, R. B. Smith identifies a stable association between variables and the successful elimination of external factors as being two necessary components for the establishment of linear causality.
Two structural models used with causality are hierarchical and nonhierarchical models. The difference between them is that hierarchical models lack feedback loops whereas nonhierarchical models have them. Feedback loops occur when X causes Y and Y causes X, or when they effect each other through one or more other variables. For example, X causes Y, Y causes Z, and Z causes X. A statistical methodology called path analysis can be done with standardized hierarchical models but the tracing rule of path analysis does not apply for nonhierarchical models.
If there is a causal relationship between variable X and effect Y, then X can be one of a couple different kinds of causes.
- Necessary causes — With necessary causes, cause X must be present for effect Y to occur. For example, a naturally conceived birth (effect Y) in humans necessarily requires a sperm to fertilize an egg (cause X). The presence of cause X, however, does not necessary require effect Y to occur. For example, the female may be on birth control thereby preventing cause X to lead to effect Y.
- Sufficient causes — With sufficient causes, the presence of X necessarily implies the occurrence of effect Y, but effect Y does not imply the presence of X. For example, chopping a person's head off with a guillotine (cause X) necessarily implies the occurrence of death (effect Y). However just because someone died (effect Y) doesn't mean that their head has been chopped off (cause X) as they could also have been killed by hanging, firing squad, electric chair, lethal injection, stoning, or burning at the stake.
- Contributory causes — With contributory causes, the presence of cause X makes the presence of Y possible but not with 100% probability. For example, eating refried beans (cause X) could be a cause of flatulence (effect Y), but this effect could also have been due to the soft drink one drank with it.
The Bradford-Hill Criteria
In epidemiology, causal relationships can be determined via the Bradford-Hill Criteria. There are 8 parts of this criteria, each one strengthens the possibility of causal relationship between the cause of the disease and the effects of it. The criteria are as follows:
- Strength of the association — The stronger the association the more causal the outcome.
- Consistency of findings — Finding must apply in different conditions.
- Specificity of findings — There must be a one-to-one ratio of cause and effect.
- Temporal Sequence of association — The cause must precede the effect.
- Biological Gradient — The more exposure, the higher the disease rate.
- Biological Plausibility — Is there a biological mechanism for what is happening?
- Coherence — Does it match up with what is already known about the disease?
- Experiment — Does removal of exposure change the outcome?
- The arrow of time — Human perception of causality always flows with the direction of time as causes always precede effects. Whether the present moment is physically distinct from the past and future or is merely an emergent property of consciousness has been one of the major unsolved problems in physics. Research into quantum computing, still very much in its preliminary stage, suggests that the arrow of time may be an emergent phenomena. Jayne Thompson, a researcher of quantum computing, had the following to say about causality and time.
“”The most exciting thing for us is the possible connection with the arrow of time. If causal asymmetry is only found in classical models, it suggests our perception of cause and effect, and thus time, can emerge from enforcing a classical explanation on events in a fundamentally quantum world.
- Determinism versus indeterminism — Many physicists hold to the belief that quantum mechanics is non-deterministic and for phenomena such as radioactive decay only a probabilistic description of causality can be given. In other words, Aquinas was wrong about causes being clearly demonstrated for every effect. Sometimes events just happen due to chance and therefore are not deterministically caused. The non-deterministic nature of quantum mechanics is not universally accepted however. Einstein famously dissented and promoted a hidden variable theory that would explain why quantum mechanics appeared to be non-deterministic. Louis de Broglie would also initially dissent in 1927, but he would never fully develop his theory. David Bohm would resurrect it in 1952. Known as pilot wave theory, the de Broglie–Bohm theory, or Bohmian mechanics, the theory is a deterministic interpretation of quantum mechanics that some physicists are actively researching.
Causal fallacies (Non causa pro causa)
Given the inherent difficulty of establishing causal relationships, people make causal fallacies all the time. This is particularly true of the social sciences where ethical prohibitions on human experimentation may prevent the elimination of external factors that corrupt the data. Even B. F. Skinner limited his experimentation to animals (despite rumors to the contrary).
The latin term cum hoc, ergo propter hoc is often used to describe the fallacy that occurs when one incorrectly uses correlation as the basis for causation. The fact that there are at least five different ways A could be correlated with B only one of which is A causes B contributes to the prevalence of this fallacy. If one assumes that A is the cause of B when B is actually the cause of A, then one has committed a reverse causation fallacy. If one assumes A is the cause of B due to correlation when both A and B are caused by variable Z, then one has committed a fallacy due to a confounding variable. If one assumes A is the sole cause of B when B is caused by a complex variety of factors, then one has committed the fallacy of causal oversimplification. Even if A primarily causes B, the failure to comprehend that B may also cause A can lead to fallacies. For example, an example of oversimplification of predator-prey relationships would be believing that more predators (A) will cause the numbers of prey (B) to be reduced but failing to recognize that lower numbers of prey (B) also leads to lower numbers of predators (A). Another fallacy is assuming that A is a cause of B when the correlation could simply be due to coincidence. This is a fallacy that commonly occurs when people mine data looking for statistical correlations in which case it would be an example of a post-designation fallacy.
The term post hoc, ergo propter hoc is often used to describe the fallacy that occurs when one assumes that because A preceded B, A caused B. This fallacy occurs because some people place too much emphasis on time precedence while pretty much ignoring the other requirements for causality. Andrew Wakefield is a perfect example. Just because vaccinations occur before the discovery of autism in a child does't mean vaccines cause autism as you also have to have relationship and non-spuriousness which Wakefield failed to prove. In fact, vaccines have been extensively studied and there is no positive covariance between vaccines and autism. But do you know what does have positive covariance with autism? Rubella.[note 1] You know, one of the diseases the MMR vaccine is supposed to prevent.
- Correlation and Causation by David A. Kenny - There are also some links to other statistical books on the page. These books contain math which is unavoidable given the complexity of the subject.
- The Geometry of Causality by Matthew O’Dowd - Causality as explained by an astrophysicist.
- Are Space and Time An Illusion? by Gabe Perez-Giz - Another good explanation of causality by an astrophysicist.
- Reversing cause and effect is no trouble for quantum computers - An article that describes how quantum computing may provide insight into the arrow of time.
- Is This What Quantum Mechanics Looks Like? by Veritasium - A good video of a macroscopic model of pilot wave theory using oil, water, and a speaker (pilot wave theory is deterministic).
- It also has time precedence and arguably is not spurious.
- Timaeus by Plato, 360 B.C.E, translated by Benjamin Jowett, MIT's Internet Classics Archive
- Plato's Timaeus by Donald Zeyl and Barbara Sattler, Stanford Encyclopedia of Philosophy, first published Oct 25, 200, revised Dec 18, 2017
- Aristotle on Causality by Andrea Falcon, Stanford Encyclopedia of Philosophy, first published Jan 11, 2006, revised Mar 11, 2015
- The Summa Theologica by Thomas Aquinas, translated by fathers of the English Dominican Province
- Must the beginning of the universe have a personal cause?: A critical examination of the Kalam Cosmological Argument by Wes Morriston, Faith and Philosophy, Vol. 17, No. 2, April 2000
- Of the Idea of necessary Connexion Part II by David Hume, from An Enquiry Concerning Human Understanding, The Harvard Classics, 1909–14
- A Treatise of Human Nature, Book I, Part III: Section XIV by David Hume
- Causality by F. Daniel Hidalgo amd Jasjeet S. Sekhon, UC Berkeley
- The Problem of Induction by Leah Henderson, Stanford Encyclopedia of Philosophy, Mar 21, 2018
- The Geometry of Causality, Space Time by Matthew O’Dowd, PBS Digital Studios, Feb 2, 2017
- Causal inference in statistics: An overview by Judea Pearl Statistics Surveys Vol. 3, 2009, 96–146.
- Correlation and Causation by David A. Kenny, March 2004
- Change, Self‐organization and the Search for Causality in Educational Research and Practice by Matthijs Koopmans, Complicity: An International Journal of Complexity and Education Volume 11, 2014, Number 1, pages 20‐39
- Causal Reasoning iSTAR Assessment: Inquiry for Scientific Thinking and Reasoning, filed in Dimensions of Scientific Reasoning on Apr.11, 2011
- Reversing cause and effect is no trouble for quantum computers, July 20, 2018, National University of Singapore
- Causal Determinism by Carl Hoefer, Stanford Encyclopedia of Philosophy, first published Jan 23, 2003, revised Jan 21, 2016
- Pilot Wave Theory and Quantum Realism, Space Time by Matthew O’Dowd, PBS Digital Studios, Nov 30, 2016
- Is This What Quantum Mechanics Looks Like? by Veritasium, Nov 2, 2016
- Fallacies, Internet Encyclopedia of Philosophy by Bradley Dowden
- Fallacy Files by Gary N. Curtis, accessed August 2018
- Kimberly Moekle, Conspiracy Theory handout, Stanford University, Fall 2008
- Dr. Robert Bennett, Logically Fallacious, accessed August 2018
- Does Rubella Cause Autism: A 2015 Reappraisal? by Jill Hutton, Frontiers in Human Neuroscience, 2016, 10: 25.