Dear Chuck, Ken and all,
Thank you all for the interesting discussion.
It's tempting to think of correlation and explanations of causality being independent. Ken provides some examples of why it might be so.
In many ways, it is better to think of them being related in the same manner as evidence and representations, evidence and models or evidence and research theories.
A more general picture than that which Ken presented is to view all situations in terms of a given, and always finite, amount of relationships, often dynamic.
Behaviours of parts of the situation or the situation as a whole depend on the relationships and their dynamics.
Theories of causality attempt to explain behaviours of parts of such situations in terms of the more dominant dynamic relationships that apply to those behaviours. Such theories of causality are always incomplete, and always incompletely represent the behaviours and the causes of them .
Measures can be taken of the behaviours. The values of such measures depend of course on the relationships. The correlation between such measures in turn depends on the causal relationships.
The question is HOW MUCH we can infer of the causal relationships from the correlations between the measures of behaviours?
That depends on the complicatedness of the situation, i.e. the number of relationships and variables, the degrees of freedom of the behaviour of variables, the amount of variety, the amount of orthogonality (linked to degrees of freedom, variable boundaries, behaviours of boundaries and a number of other related factors, AND the kinds of theories of causality.
In the limit, it is possible to shape aspects of a situation, for example to reduce the degrees of freedom (in practice or in theory) to the point where correlation exactly maps to causality. At that point it becomes what is often loosely called 'evidence'. Physics experiments are examples of this. For example, correlations between measurements of nuclear crashing and bashing can, if the situation is shaped to reduce the degrees of freedom, indicate (say) the presence or absence of Higg's Bosa.
At the other end of the scale, correlations between measures of behaviours of systems with very high levels of degrees of freedom between variables may indicate almost nothing about causal relations.
For all situations, correlations between behaviours of variables indicate *something* about causality and provide a sound basis for some kinds of theories of causality
Whether this is sufficient often depends more on the type of causal theory that is required. Economic theories are a sweet example of this issue.
So, as Don might say, this requires a more nuanced understanding.
Best wishes from the North West Fringes of Holland and its wonderful flat bottoms.
Dr Terence Love
IEED, Management School
Lancaster University, Bailrigg, UK
Love Services Pty Ltd
PO Box 226, Quinns Rocks Western Australia 6030
Tel: +61 (0)4 3497 5848
Fax:+61 (0)8 9305 7629
[log in to unmask]
From: [log in to unmask] [mailto:[log in to unmask]] On Behalf Of Ken Friedman
Sent: Tuesday, 17 February 2015 6:38 PM
It’s true that no one has said much about correlation. For me, the reason is that any one of dozens of processes suggest possibilities. Earlier, you mentioned correlation and causality: the reason I avoided talking about correlation is that there is generally no relation between correlation and causality. As with abduction and metaphor, correlation is valuable and suggestive in the logic of discovery, but there is much more require to establish causality.
The famous example is the case of a rooster who crows every day at sunrise. This does not means that the rooster’s call forces the sun to rise. As best I know, every Nobel Laureate in physics holds a PhD. (I am not sure that every PhD is a physics degree — some may hold a PhD in mathematics.) There is 100% correlation between winning the Nobel Prize in physics and holding a PhD. There is an magnificently low correlation between earning a PhD and winning the Nobel Prize in physics.
As with abduction, correlation points to interesting possibilities, but it does nothing to establish truth value, validity, or correct conclusions. Correlation is one of several dozens mechanisms that inspire creative and valuable ideas. Because we can build on these in many ways, there is more to be said than I felt I could manage in a short post. There are many books about correlation across many fields, but these tend to be technical and careful in delimiting the issues they address. There might be something in them, though, so Amazon is worth searching. The Stanford Encyclopaedia of Philosophy has a great deal on correlation tucked away with articles on other topics. But I felt that I couldn’t add much.
The case of abduction is different. It is a definable process — and in a sense, one can address abduction in a more limited scope. I entered the conversation to make a collection of articles available and to offer a few apposite thoughts.
Ken Friedman, PhD, DSc (hc), FDRS | Editor-in-Chief | 设计 She Ji. The Journal of Design, Economics, and Innovation | Published by Elsevier in Cooperation with Tongji University Press | Launching in 2015
Chair Professor of Design Innovation Studies | College of Design and Innovation | Tongji University | Shanghai, China ||| University Distinguished Professor | Centre for Design Innovation | Swinburne University of Technology | Melbourne, Australia
PhD-Design mailing list <[log in to unmask]>
Discussion of PhD studies and related research in Design
Subscribe or Unsubscribe at https://www.jiscmail.ac.uk/phd-design