Improving the scientific hypothesis
The scientific hypothesis is very important. In is a major reason why
the Western world overtook China in the last millennium. 2000 years ago,
the Chinese were more advanced than Europeans. Scientific discovery
stagnated in China, when scholars believed you had to move from
certainty to certainty. Ed has written this in a number of books.
Essentially the hypothesis approach is starting with a conclusion, then
seeing if there is evidence for the conclusion.
For example, a person first hypothesizes that smoking causes lung
cancer. A study takes place in which data is collected and carefully
analyzed, to see whether the statistics support the hypothesis.
This approach has a major weakness though. There is a tendency to
discover what you are looking for. (One of several reasons for this is a process called sensitization, which Ed has written about.)
When a person has a hypothesis, sometimes their ego (or financial interests) can get involved. Experiments can be inadvertently tainted to produce the
desired result. A more blatant case is to
simply repeat the experiment until the desired effect is produced. There is the old story about Pasteur inoculates one dog, 2 dogs are infected they both die. No problem, do the experiment again. This time the inoculated one dies and the other lives. No problem, he did the experiment a third time till he got the result he wanted. (Not sure of the exact details, but its to make a point)
One experiment was done by someone who wanted to prove
oxygen was bad for you. He put some flies in a box which contained pure
oxygen (An absurd artificial situation). These lived longer than the control group (the undesired result)
so he redefined his aim and concluded that oxygen reduced the quality
of life on the flies.
If I wanted to prove "Blondes have more fun", I could influence the
results of a study, by how blonde or fun is defined, how the data is
sampled etc. etc. etc. There are myriads of ways of doing it.
Instead of hypothesizing "smoking causes lung cancer" the question could be asked "Is there a relationship between smoking and lung cancer or not?" A negative result should be regarded as just as valuable as a positive result. This small step could help the experimenter be more objective and get more studies (with negative results) published.
Another limitation is that many things are discovered accidentally, that
is when looking for something else.
My proposal is that an alternative to the hypothesis, is to observe a phenomenon and explore its causes, effects and relationships with other
phenomena.
For example, instead of hypothesizing "smoking causes lung cancer" ask
What are the long term effects of smoking on the lungs? (or the whole
body, if the study was large enough).
This is more neutral and objective. Its also a lot more efficient.
Accidental discoveries need to be noted, also there are approaches that
can increase the probability of finding accidental discoveries. Ed writes about this in his books on creativity.
I can go into more detail, but the post is probably long enough already.
the Western world overtook China in the last millennium. 2000 years ago,
the Chinese were more advanced than Europeans. Scientific discovery
stagnated in China, when scholars believed you had to move from
certainty to certainty. Ed has written this in a number of books.
Essentially the hypothesis approach is starting with a conclusion, then
seeing if there is evidence for the conclusion.
For example, a person first hypothesizes that smoking causes lung
cancer. A study takes place in which data is collected and carefully
analyzed, to see whether the statistics support the hypothesis.
This approach has a major weakness though. There is a tendency to
discover what you are looking for. (One of several reasons for this is a process called sensitization, which Ed has written about.)
When a person has a hypothesis, sometimes their ego (or financial interests) can get involved. Experiments can be inadvertently tainted to produce the
desired result. A more blatant case is to
simply repeat the experiment until the desired effect is produced. There is the old story about Pasteur inoculates one dog, 2 dogs are infected they both die. No problem, do the experiment again. This time the inoculated one dies and the other lives. No problem, he did the experiment a third time till he got the result he wanted. (Not sure of the exact details, but its to make a point)
One experiment was done by someone who wanted to prove
oxygen was bad for you. He put some flies in a box which contained pure
oxygen (An absurd artificial situation). These lived longer than the control group (the undesired result)
so he redefined his aim and concluded that oxygen reduced the quality
of life on the flies.
If I wanted to prove "Blondes have more fun", I could influence the
results of a study, by how blonde or fun is defined, how the data is
sampled etc. etc. etc. There are myriads of ways of doing it.
Instead of hypothesizing "smoking causes lung cancer" the question could be asked "Is there a relationship between smoking and lung cancer or not?" A negative result should be regarded as just as valuable as a positive result. This small step could help the experimenter be more objective and get more studies (with negative results) published.
Another limitation is that many things are discovered accidentally, that
is when looking for something else.
My proposal is that an alternative to the hypothesis, is to observe a phenomenon and explore its causes, effects and relationships with other
phenomena.
For example, instead of hypothesizing "smoking causes lung cancer" ask
What are the long term effects of smoking on the lungs? (or the whole
body, if the study was large enough).
This is more neutral and objective. Its also a lot more efficient.
Accidental discoveries need to be noted, also there are approaches that
can increase the probability of finding accidental discoveries. Ed writes about this in his books on creativity.
I can go into more detail, but the post is probably long enough already.
Replies to This Discussion
-
Permalink Reply by Kim Jones on -
It gets even worse when a hypothesis is developed into a theory. The very term "theory-dependence of observation" is in fact the greatest fear of scientists but the paradox is that you cannot really observe anything at all without some kind of theory that something is there to observe in the first place.
As I am fond of saying "Believing is seeing" (not the other way around).
I recently lost my car keys (as you do). I turned the whole apartment upside-down in a frantic attempt to find them as I was running late. My "theory" was that I must have dropped them somewhere and I was only going to find them if I crawled around on my hands and knees and scrutinised the floor. I wasted fifteen minutes in this state. Just as I gave up the search and reached for the phone to call a taxi I saw them sitting by the phone. In other words I had convinced myself that I had actually lost them, so I went about proving this fact to myself rather stupidly. In other words, I "found what I sought" (which was to convince myself that the keys could not be easily found.) My perception - my very eyesight - was addled by this stupid theory. I must have walked past the phone twenty times and could have found them just by looking in a place that my theory had not addressed.
-
Permalink Reply by tzujan on
-
what about the null hypothesis? aren't experiments supposed to be proven wrong?
-
Permalink Reply by Kim Jones on
-
Well, the Null Hypothesis is a perfectly sensible starting point but if the experiment confirms statistically significant correlations in the data observed then your null hypothesis is refuted and you are free to build a tentative theory on the basis of your experiment.
Experiments are supposed to be proven wrong, yes, because this then falsifies the hypothesis which is a great service done in the direction of finding the most correct theory.
A hypothesis is to a theory as a concept is to an idea. The one is the starting point that (hopefully) leads to the other. In other words, the goal of forging a concept is to embody it in an idea which will put the concept to work. The goal of having a hypothesis is to end up with a beautiful theory that explains what it is you are observing. Very occasionally this happens.
If your experiment(s) refute the hypothesis then it would probably be better to come up with a different hypothesis rather than keep on tweaking the experimental setup to confirm the hypothesis as Jack has noted in the case of Pasteur.
The hypothesis that light could only travel through some kind of medium (air, water, whatever) led to the erroneous theory that all of space was filled with a substance called "the Ether". Since no experiment ever found evidence for such a thing, the theory was put to bed. The assumption that light propagates as a wave was refuted by the new Quantum theory (which has not yet been refuted and probably never will) which indicated the corpuscular nature of electromagnetic radiation.
The point is, they had to look in an entirely different direction to the one suggested by the initial hypothesis to come up with a more correct theory.
-
Permalink Reply by Kim Jones on
-
To return to Jack's proposition:
Instead of hypothesizing "smoking causes lung cancer" the question could be asked "Is there a relationship between smoking and lung cancer or not?" A negative result should be regarded as just as valuable as a positive result. This small step could help the experimenter be more objective and get more studies (with negative results) published.
This could further be refined by asking "Smoking WHAT causes lung cancer?". Nobody just sits there and smokes like Morticia Adddams in The Addams Family. If one smokes, one smokes something. This means we should perhaps spend more time looking for substances that are smoked that don't cause lung cancer.
It's all about framing the question, in other words. Whatever question you ask of Nature or Reality, you will get an answer to the question asked. That does not of itself mean you have somehow stumbled upon "the truth".
The following is just a bit of an excursion:
Posters and propaganda that state simplistically "Smoking Kills" should perhaps be taken down or - best of all - modified to read more accurately as "Smoking Nicotiana Tobacum Kills".
Looking for or allowing for the smoking of beneficial substances, or, at the very least, low-risk substances exhibiting substantial benefits in other directions seems to be logical and practical.
Aerosol delivery of asthma palliatives is similar in concept (and much safer) than smoking a product which would do the sam job, basically because simpler and less risky. Why bother with rolling a cylinder of paper and stuffing it with whatever, striking a match (and possibly burning your fingers) etc. when you can simply raise the gadget to your face and inhale.
Those of you who have seen Edward's movie 2040 Possibilities will recall the section where he talks about recreational drugs. His idea is that by 2040 recreational drugs will be seen as largely beneficial because they actually do help people be happy. Clearly the only thing to overcome would be engineering out the harmful side effects. One of these is the ever-present spectre of addiction. By 2040 then, we could have hand-held aerosol guns (just like asthma inhalers) which dose a measured amount of tetrahydrocannabinol for example, and of a type that will not cause the body's tolerance of it to rise. This would mean you would only ever get affected to the extent of the single dose, no mater however many successive puffs you took to try to increase the high. In this way, people might enjoy all of the benefits of recreational drugs without most of the dangers.
A drug might be designed, for example, specifically for long-haul truck drivers that they could absolutely rely on - OK, you say - relying on a drug is addiction, but then if a certain drug addiction means traffic accidents are reduced then addiction in that specific direction is perhaps something of a good thing.
Back to the main topic of improving the scientific hypothesis.
-
Permalink Reply by Raymund Kwok on -
One important concept in scientific method is that everything can change. There is no absolute truth. Existing theories are being improved or disapproved while new one comes. This concept makes the world dynamic. China became less dynamic when Confucianism was used as the official ideology.
One of my learning as a scientist is the importance of first proposing hypothesis with big impacts, then perform careful experiments to refine the hypothesis. This approach increases the level of dynamic in the scientific community. So, I prefer crazy ideas followed by highly disciplined work supported by logic and mathematics. I prefer not to add restriction to the phase of hypothesis generation.
-
Permalink Reply by Raymund Kwok on
-
"Dynamic" here means changing. Scientific method allows improvements and breakthroughs to be made continuously. Now, the growth is exponentially.
As mentioned in Prof. de Bono's book, China once was a world superpower. But when Confucianism was adopted as the official ideology, the change becomes very slow. While in Europe, the use of scientific method allowed Industrial Revolution, which allowed many countries to by-pass China.
-
Permalink Reply by Raymund Kwok on
-
The scientific community is composed by people. The human factor is very important in the scientific community. I read and encountered many many stories of wrong doings. But this is not the deficiencies of scientific method.
Scientific method itself does not contain a value system. Scientific method does not tell us what is the purpose of using scientific method.
If the goal is to increase a country's GDP, I believe the following are important :
- institutions allow high impact ideas to be generated
- funding mechanism to support projects that will positively impact GDP (means idea with high impact to GDP and team of people that shows high probability of success in the project)
- a network of institutions that allows quick implementation of ideas to real products
-
Permalink Reply by Free Willy on -
Raymund. Can you enlighten us?
Are you trying to say that scientific method is not a living as of a living person but is evolving because people are interacting and finding ways to improve the scientific method. And it is the interaction of people who try to manipulate the outcome of scientific result even method used to gain competitive advantage for self.
In the sense of increasing a country's GDP, what is measured got done at the expenses of society, environment, etc. And by virtue of Confucianism it should not happen (therefore China would be less dynamic). Rapid development happened because overall it create greater value and securing the China superpower status (to create a balance of power). It is the people who have value system and not scientific method which is just a tool.
-
Permalink Reply by Raymund Kwok on
-
Scientific method as a method is simple and not really have any change. But people's thinking can change. One example is the science policy changed from focus more at pure science in 70s to currently more at applied research.
I am not saying it is moral for a scientist to manipulate the results. Manipulate here means providing false results that the scientist already knew they are false. Scientific method allows people to make mistakes. It allows people to interact and finding better theories.
It is possible to interpret the data differently. People can base on the same experimental data but give different theories. This is common in the scientific community. This encourages more experiments for supporting the theories.
It is not making sense to have a pure objective view on the experimental results. Let's assume from a set of experiments, statistical results show:
(a) smoking is related to lung cancer
(b) smoking is not related to skin cancer
(c) smoking is not related to X1 cancer
(d) smoking is not related to X2 cancer
..
(..) smoking is not related to X98 cancer
Let's compare the following three conclusions:
(1) "smoking causes cancer" and further conclusion can be "smoking should be banned".
(2) "The effects of smoking to 100 types of cancers was studied. Lung cancer was find to be linked to smoking while others not. "
(3) "Statistically, from the 100 types of cancers studied, smoking showed insignificant effect".
For (2), the reporting scientist can treat this as pure objective findings and let the policy maker to make decisions.
For (1), the scientist provide logical suggestions based on the experimental results.
For (3), this is a manipulation of data and trying to hide important findings with a lot of unrelated data.
-
Permalink Reply by Raymund Kwok on
-
We agree very well that scientific method is just a tool. It is the people to decide what value system to be used. When people value sustainability, more resources can be put on sustainability projects. When people value reducing number of poor people, more resources can be put into research for production technologies that allow competitiveness in factory production for job creation.... Some countries focus renewable energy, some on LED for lighting, some on software ....
Confucianism is not really practiced in China now. The government recently tried to promote it for improving people relationship.
-
Permalink Reply by Franis on -
Interesting ideas. Believe it was Issac Asimov who said something such as, "It is when you hear laughing that you know that a discovery has been made, not, Eureka! I found what I expected!"
Indulge me to put in my own words how I understand that you are saying. Are you saying that the shaping of the question in creating the hypothesis directs the attention toward specific results - and therefore would discount unpredictable results?
Can you say more about how to shape a constructive hypothesis?
Seeking relationships and describing qualities is more constructive and leads to greater inventions? As opposed to the tendency to "match for results" that shapes many hypothesis?
-
Permalink Reply by William Jack Jordan on -
Returning to some of black hat thinking re scientific method. Some points are:
Publication bias: Studies that support the null hypothesis (i.e. don't support the hypothesis are less likely to be published in journals etc. and less likely to be read.
More sensational results are more likely to be published, even if they are flawed.
Media publication: Some studies go straight to the media and bypass peer review, again they may be sensational but flawed.
Flawed studies: Flawed studies are more common than people may think, especially those regarding human behaviour. Flaws include selection bias, small numbers, lack of adequate controls, failure to deal with confounding factors, faulty maths etc etc
Other types of flaws include the Hawthorne effect (attention effecting subject's behaviour), Placebo effect, and failure to create a double-blind experiment (one of the best features of the scientific method) etc
Reference to other studies: Once a study is published it may be treated like it is cast in stone. (Rock logic comes to mind) This is very different from what Raymund was saying about change. Studies have reference to flawed studies, rendering them flawed.
Inadequate training: Studies may be done by people with little training in the stats etc. side of things. Medical studies sometimes have this difficulty. Sometimes studies involve specialist staticians + the experimenters, and the 2 groups don't communicate well.
Taking off my black hat, in some cases an approach of:
(a) Observing what is (without modification)
(b) Looking for 'real' patterns
(c) Designing a way forward
could be more efficient than the hypothesis method.
Imagine if we had 2 towns A and B. One has a higher level of something undesirable than the other
e.g. a disease, crime etc.
We could hypothesis it was because Town A had a Lead Smelter there, or the men wore nylon underpants or whatever.
If the hypotheses kept proving to be null, it might take a month of sundays to discover the cause.
It would be more efficient to observe (and rank) the differences between A and B and see if there was a pattern.
Samir said Experiment developes hypothesis ,hypothesis develops experiment.
The thought for an hypothesis comes from somewhere, this would be worth exploring.
Also still idea leads to multiple hypotheses. I can imagine there are whys of testing several hypotheses at once.
Studies are typically done with little suggestion of how the results are to be used.
To be practical the information needs to be applied.
© 2013 Created by Administrator.
Powered by
