Sunday, January 29, 2006
Causal Relationship
In numerous discussions with my roommates, I have held that I like research in physical sciences because it heavily depends on deduction rather than induction like life sciences (which interests them). In physical sciences, once you know the universal laws you can reliably and accurately predict the behavior of any system; knowing the general rule you can apply it to a specific instance. And the laws or rules are derived from a relatively few postulates, making study of physical sciences less memory intensive. The 'truth' in physical sciences is deterministic, or atleast has a steady state probability distribution. On the other hand, life sciences (like behavioral & social sciences) make a large number of observations and tend to generalize them to a rule based on statistical induction. Thus, the 'truth' is rarely deterministic, and many times has a dynamic distribution, with the probabilities changing with each new study.
A few days back I was complaining to a friend about my back pain. I said it must be because I am not getting enough sleep and rest. He said, my not getting sleep (Y) must be caused by my thinking about something (X) (he actually said somebody :-)) I had also reached the same conclusion independently, but since he put it in words I saw a flaw in the reasoning. Based on observations, everytime I can't sleep Y, I think of X. Thus, there is high correlation between X & Y, but can we conclude that X causes Y? It is equally possible that since I can't sleep and have nothing better to do I think of X. Is there anyway in which we can prove a causal relationship between X & Y? For all we know, both X and Y may be caused by a third variable Z (e.g. age), resulting in high correlation between X & Y. All we can do is notice the correlation and speculate on the causes. We cannot even reliably say whether the cause of insomnia is psychological or physiological. The same argument applies to sleep and back pain - we cannot say which causes the other. Thus, proving causal relationship (what is called "kaarya - kaaran siddhanta" in sanskrit) is a major pitfall in life sciences, making its discoveries as speculative and reliable as price discovery in stock markets.
A few days back I was complaining to a friend about my back pain. I said it must be because I am not getting enough sleep and rest. He said, my not getting sleep (Y) must be caused by my thinking about something (X) (he actually said somebody :-)) I had also reached the same conclusion independently, but since he put it in words I saw a flaw in the reasoning. Based on observations, everytime I can't sleep Y, I think of X. Thus, there is high correlation between X & Y, but can we conclude that X causes Y? It is equally possible that since I can't sleep and have nothing better to do I think of X. Is there anyway in which we can prove a causal relationship between X & Y? For all we know, both X and Y may be caused by a third variable Z (e.g. age), resulting in high correlation between X & Y. All we can do is notice the correlation and speculate on the causes. We cannot even reliably say whether the cause of insomnia is psychological or physiological. The same argument applies to sleep and back pain - we cannot say which causes the other. Thus, proving causal relationship (what is called "kaarya - kaaran siddhanta" in sanskrit) is a major pitfall in life sciences, making its discoveries as speculative and reliable as price discovery in stock markets.
Comments:
<< Home
o boy, what a blog!! very interesting and rewarding to read.
Earlier, I too had a similar preference and opinion about life sciences; but now I believe that the glamor is largely due to the [i]state[/i] of a science, rather than an inherent nature of a science. Any science in its early stages is mostly empirical, based on little deduction but a huge number of observations. For example, before Euclid's elements, math was mostly a mess made of large look-up tables and approximate (thus inaccurate) formulas induced by experimental results. I believe that even Da Vinci's work on developing instruments employed less deduction and more induction. On the other hand, vedic concepts of medicine were much more organized compared to the mess we have now. Somehow the popularity of Ayurveda didn't survive the clash and merger of cultures, and the modern life sciences had to start all over.
I agree that life sciences, as of today, are not as causal and deductive as physical sciences. There are many two-way connections, for ex I learned in a seminar that if one keeps a smiling face for sometime, he/she will eventually feel happier!! But the same is true for many laws of physics, for ex PV = nRT interrelates pressure, volume, and temperature; the answer to "which is cause and which is effect" depends upon what you control at a particular moment vs what you measure.
No science can be completely deductive, or else there won't be any need for research. To the extent that a researcher's job is to advance the state of science, I see more opportunities in life sciences, thus weakening the preference for physical sciences.
May be some day there will be another Euclid, this time for biology, to put all the pieces together and produce a cohesive picture :)
Post a Comment
Earlier, I too had a similar preference and opinion about life sciences; but now I believe that the glamor is largely due to the [i]state[/i] of a science, rather than an inherent nature of a science. Any science in its early stages is mostly empirical, based on little deduction but a huge number of observations. For example, before Euclid's elements, math was mostly a mess made of large look-up tables and approximate (thus inaccurate) formulas induced by experimental results. I believe that even Da Vinci's work on developing instruments employed less deduction and more induction. On the other hand, vedic concepts of medicine were much more organized compared to the mess we have now. Somehow the popularity of Ayurveda didn't survive the clash and merger of cultures, and the modern life sciences had to start all over.
I agree that life sciences, as of today, are not as causal and deductive as physical sciences. There are many two-way connections, for ex I learned in a seminar that if one keeps a smiling face for sometime, he/she will eventually feel happier!! But the same is true for many laws of physics, for ex PV = nRT interrelates pressure, volume, and temperature; the answer to "which is cause and which is effect" depends upon what you control at a particular moment vs what you measure.
No science can be completely deductive, or else there won't be any need for research. To the extent that a researcher's job is to advance the state of science, I see more opportunities in life sciences, thus weakening the preference for physical sciences.
May be some day there will be another Euclid, this time for biology, to put all the pieces together and produce a cohesive picture :)
<< Home
