Correlation vs Causation

126 Replies, 13821 Views

Hoping to draw more attention to this, but I've long pondered the use of "correlation" over "causation" in the case of discussing neuroscience, and the motives behind making such alternative explanations? One author seems to think it's almost out of desperation

Quote:Thus for instance, ‘correlation is not causation’ is countered by the observation that the effects of other organs – the kidney’s role in filtering toxins, for instance – is not disputed, and that it’s highly selective to apply different reasoning to the brain (p. 102). (Who now continues to resist the implications of the correlation between smoking and lung cancer?) To insist otherwise, is a ‘fallacy called moving the goalposts: an utterly unreasonable person pretends to be reasonable, if only more evidence, impossible to obtain, were available’

Honestly it raises a good point, but at the same time I feel the comparison is lacking.

My sediments to everyone here, I'm quite on a fence myself as of recently on the whole matter, so I don't come looking to damn the skeptics or the proponents here. As loose as those titles are here. But, to anybody who argues against the idea of the mind not being a product of the brain, why should "correlation not causation" be taken seriously as a explanation? And likewise
The source of the quote - https://www.amazon.com/Myth-Afterlife-ag...op?ie=UTF8

I think the question is pretty moot. "Correlation" seems appropriate when talking about observational associations (for example, much of the fMRI studies). But once you have solid interventional research, where the presence or absence of the factors studied are controlled by the researcher, we have moved into "causation" (for example, studies on anaesthetics).

And I think most researchers, including proponents of "mind", accept this. Even idealists try to take this into account with the "brain as a filter" idea, for example.

A comment on the two examples given...kidney's role in filtering toxins is not based on "correlation". But the association between lung cancer and smoking is, in a way. That is, there aren't interventional studies in humans (for obvious reasons - it would be highly unethical). But even so, it serves as an example where the strength of the correlation exceeds that seen for causation.

Linda
(This post was last modified: 2018-02-01, 11:46 AM by fls.)
[-] The following 1 user Likes fls's post:
  • stephenw
(2018-02-01, 11:40 AM)fls Wrote: The source of the quote - https://www.amazon.com/Myth-Afterlife-ag...op?ie=UTF8

I think the question is pretty moot. "Correlation" seems appropriate when talking about observational associations (for example, much of the fMRI studies). But once you have solid interventional research, where the presence or absence of the factors studied are controlled by the researcher, we have moved into "causation" (for example, studies on anaesthetics).

And I think most researchers, including proponents of "mind", accept this. Even idealists try to take this into account with the "brain as a filter" idea, for example.

A comment on the two examples given...kidney's role in filtering toxins is not based on "correlation". But the association between lung cancer and smoking is, in a way. That is, there aren't interventional studies in humans (for obvious reasons - it would be highly unethical). But even so, it serves as an example where the strength of the correlation exceeds that seen for causation.

Linda

I appreciate the methodological approach in your take on the semantics of the two terms.  In practice, science has two stages - gathering and recording data and then - the logical analysis of the meaning of this data.  Correlation is a term of data comparison and contrast.  It belongs to the empirical side of the work, so data correlation can be rock solid in revealing real-world structures and patterns.  The meaning of the data correlation is open to interpretation.

Causation is a term of logical analysis and interpretation is a tool of the job.  It rests on correspondence between data patterns and rule-based model building.  Manipulating variables is the science method to probe relationships experimentally.  In the modern day, causation rests on process models and the simulations of the process models. The reason that this has become THE METHOD is that changing variables in a model is lot easier than in an experimental set-up. 

The process models for electro-chemical activity in the neural net are physically sound and when modeled, their simulations show we have a decent grasp of the presentment of complex patterns of electronic signaling.

On the other hand - the current state of affairs on the process models for how understanding and deeply felt meaning comes from these electrically charge patterns, is (in my opinion) dumb-struck.  The quality of the just-so-narratives are good, its just that they don't lead to working models for mind (but do generate academic and publishing paychecks).  

Darwin and his protege G. Romanes believed that the evolution of mind and mental capability was an important part of his theory of evolution.  What the hell happened - to that part of the theory of evolution?  Darwin was fascinated by animal instinct and its inheritance.  Science --- after a hundred years of ignoring mind in evolution --- is finally returning to the observation of inherited behavioral traits through epigenetics.
(This post was last modified: 2018-02-01, 02:42 PM by stephenw.)
[-] The following 3 users Like stephenw's post:
  • Kamarling, Valmar, Typoz
(2018-02-01, 02:39 PM)stephenw Wrote: I appreciate the methodological approach in your take on the semantics of the two terms.  In practice, science has two stages - gathering and recording data and then - the logical analysis of the meaning of this data.  Correlation is a term of data comparison and contrast.  It belongs to the empirical side of the work, so data correlation can be rock solid in revealing real-world structures and patterns.  The meaning of the data correlation is open to interpretation.

Causation is a term of logical analysis and interpretation is a tool of the job.  It rests on correspondence between data patterns and rule-based model building.  Manipulating variables is the science method to probe relationships experimentally.  In the modern day, causation rests on process models and the simulations of the process models.

I don't know what you are referring to here. But that may be because you seem to be referring to "causation" differently than it is used with respect to scientific research. The research in the field of medicine/physiology/anatomy does not rest on process models and simulations, but direct experimentation. Models are only as good as the extent to which they correspond to those real world results, after all.

Linda
(This post was last modified: 2018-02-01, 04:53 PM by fls.)
(2018-02-01, 04:53 PM)fls Wrote: I don't know what you are referring to here. But that may be because you seem to be referring to "causation" differently than it is used with respect to scientific research. The research in the field of medicine/physiology/anatomy does not rest on process models and simulations, but direct experimentation. Models are only as good as the extent to which they correspond to those real world results, after all.

Linda

I suspect some folk aren’t happy unless you can present a causation chain all the way back to to the Big Bang. Thus philosophy.
[-] The following 1 user Likes malf's post:
  • stephenw
(2018-02-01, 06:36 PM)malf Wrote: I suspect some folk aren’t happy unless you can present a causation chain all the way back to to the Big Bang. Thus philosophy.

That doesn't seem fair when talking about consciousness.  At least to me.

With as little as we know about it, debating whether neuroscientific study is more correlation vs causation seems relevant.  At least from my more layman perspective.
(2018-02-01, 06:36 PM)malf Wrote: I suspect some folk aren’t happy unless you can present a causation chain all the way back to to the Big Bang. Thus philosophy.

Then, what caused the Big Bang?
[-] The following 1 user Likes Typoz's post:
  • malf
(2018-02-01, 04:53 PM)fls Wrote: I don't know what you are referring to here. But that may be because you seem to be referring to "causation" differently than it is used with respect to scientific research. The research in the field of medicine/physiology/anatomy does not rest on process models and simulations, but direct experimentation. Models are only as good as the extent to which they correspond to those real world results, after all.

Linda
We strongly agree about models and empirical results.

We strongly disagree about process models.  Process models go where experimentation can't.  Process models are quality control on causal analysis!!  This is my main point.  If you have a counter-example were a process model won't help test a biological thesis - please offer it.
Quote:Causal Models

In some fields, such as macroeconomics, epidemiology, and sociology, experimental manipulation is simply not feasible, and causal relationships must be inferred from observed correlations. Beginning around 1990 has been an explosion of interest in developing causal modeling techniques to facilitate such nonexperimental causal inferences. Two important works that have garnered a substantial amount of attention from philosophers are Causation, Prediction and Search(2000), by the philosophers Peter Spirtes, Clark Glymour, and Richard Scheines, and Causality: Models, Reasoning, and Inference (2000) by the computer scientist Judea Pearl. Both frameworks employ graphs to represent causal relationships among sets of causal variables. The variables in a set V form the nodes of a graph, and certain pairs of variables are connected by edges in the graph. In a directed graph, the edges take the form of arrows, which point from one variable into another. If a graph over the variable set V contains an arrow from the variable X to the variable Y, that indicates that X is a direct cause of Y (also called a parent of Y ): the value of X has an effect on the value of Y that is not mediated by any other variable in the set V.

The causal structure represented by a directed graph is connected to a probability distribution over the values of the variables by the causal Markov condition. This condition states that, conditional upon the values of its direct causes, the values of a variable are probabilistically independent of the values of all other variables, except for its effects. In other words, a variable's parents screen off that variable from all other variables, except for its effects. (The causal Markov condition is closely related to Reichenbach's common cause principle, discussed above.) 

As to process models in Physiology - Denis Noble's models of heart cells and heart cell function are the standards in modern medicine.  I don't know what you mean by "rest on" - but medicine is heading toward bioinformatics as a primary tool.

and my favorite  
Quote: “The sciences do not try to explain, they hardly even try to interpret, they mainly make models. By a model is meant a mathematical construct which, with the addition of certain verbal interpretations, describes observed phenomena. The justification of such a mathematical construct is solely and precisely that it is expected to work - that is correctly to describe phenomena from a reasonably wide area. Furthermore, it must satisfy certain esthetic criteria - that is, in relation to how much it describes, it must be rather simple.” J. Von Neumann
(This post was last modified: 2018-02-01, 07:37 PM by stephenw.)
(2018-02-01, 07:36 PM)stephenw Wrote: We strongly agree about models and empirical results.

We strongly disagree about process models.  Process models go where experimentation can't.  Process models are quality control on causal analysis!!  This is my main point.  If you have a counter-example were a process model won't help test a biological thesis - please offer it.

As to process models in Physiology - Denis Noble's models of heart cells and heart cell function are the standards in modern medicine.  I don't know what you mean by "rest on" - but medicine is heading toward bioinformatics as a primary tool.

and my favorite  
No idea what you’re going on about. I’m talking about direct experimentation, not playing around with models. For example, we think aspirin prevents some heart attacks because of randomized controlled trials, not because somebody plugged some numbers into a model of a cardiac cell.

Also, the ‘causal inferences’ you refer to are attempts to infer causation from correlation when you lack the ability to do what I described in my first post - perform interventional experiments.

Linda
(This post was last modified: 2018-02-01, 09:01 PM by fls.)
(2018-02-01, 06:36 PM)malf Wrote: I suspect some folk aren’t happy unless you can present a causation chain all the way back to to the Big Bang. Thus philosophy.

Don’t you just have to show that interventions on the brain change ‘mind’? What am I missing, here? If I can show that H. pylori can cause gastric ulcers, why am I not allowed to show that syphilis can cause dementia (ethics aside, of course).


Linda
(This post was last modified: 2018-02-01, 09:08 PM by fls.)
[-] The following 2 users Like fls's post:
  • stephenw, malf

  • View a Printable Version


Users browsing this thread: 1 Guest(s)