How Scientists Know An Idea Is a Good One

Physicist Chris Lee explains one of the toughest judgment calls scientists have to make: figuring out if their crazy ideas are worth pursuing. He says: "Research takes resources. I don't mean money—all right, I do mean money—but it also requires time and people and lab space and support. There is a human and physical infrastructure that I have to make use of. I may be part of a research organization, but I have no automatic right of access to any of this infrastructure. ... This also has implications for scale. A PhD student has the right to expect a project that generates a decent body of work within those four years. A project that is going to take eight years of construction work before it produces any scientific results cannot and should not be built by a PhD student. On the other hand, a project that dries up in two years is equally bad. ... the core idea also needs to be structured so, should certain experiments not work, they still build something that can lead to experiments which do work. Or, if the cool new instrument we want to build can't measure exactly what I intended, there are other things it can measure. One of those other things must be fairly certain of success. To put it bluntly: all paths must lead to results of some form."

Re:For certain values of "good"

[Anonymous Coward]

I can only speak for my own field (physics), but the national average length of a Ph.D. is almost 7 years. This is according to an AIP study I read about 8 years ago. There is a large spike at 5 years (theorists) and a long tail on the high end (experimentalists). Also, during the first two years before you qualify for candidacy you are rather inundated by classwork. In which case, aiming for 4 year project sounds about right. It allows for a bit of a buffer for when things break.

Re:4 years..

[dkf]

A PhD student has the right to expect a project that generates a decent body of work within those four years.

Four years? Ha! That's a good one!

The easiest way to enforce that is for the awarding institution to say that if it isn't done in 4 years, it will be taken as a complete failure. Suddenly, people find that it is possible to write up in time. (Seriously, if you can't stop pissing around "doing just one more experiment" or "reading just one more paper" and write up your thesis, you're a failure as a researcher and should be publicly branded as such.)

Good ideas are discovered after the fact!

[SirAstral]

Good ideas are hard to determine, and sometimes you find out something was actually a really bad idea only after several years like trans fats, or saccharin.

The results of scientific discovery are diminished by classifying them as success/failure. The only 2 classifications should be "A Truth Discovered" or "Pseudo Science".

Any lab experiment which is conducted to seek the truth even if it does not yield a commercially viable result is still a truth discovered. A so-called failed experiment still is a success at discovering a method which does not work to achieve desired results, and discovering what does not work in some cases can be more important then finding out what does and is an actual truth discovered.

Any experiment performed to skew results in a particular direction, or where evidence is tossed that does not agree with your idea's is nothing but pure Pseudo Science. Unfortunately we have so much of this it has made people distrust scientists because they have proven that they are just as opportunistic as normal people and will do just about any dishonest thing for a buck! True Science be damned!

Basically they don't. At least they shouldn't

[140Mandak262Jamuna]

If you know `a priori whether an idea is "good" or "bad" it will bring prejudice that will taint the results. One of the famous example is a naive Indian astrophysicist on his first trip outside India met Eddington on eve of his big presentation, in 1929. That guy explained in great detail his idea and Eddington not only dismissed it, he was scheduled to present just before this paper from that young man. He trashed the idea so much that the young man abandoned that field altogether and chose to pursue a different field [*]. Most others dismissed that paper too. It eventually garnered that young man a Nobel Prize in physics, and is the foundation of what is known as Chandrashekar Limit that tells you if a star is big enough to go supernova. That paper was discovered about 15 years later, after WW II. So in theory they should not know if an idea is good or bad.

But that is theory. In practice, having some realistic goals based on available resources of money and time is common to all fields, not just science.

[*] Chandrashekar was not bitter about Eddington, he credits being forced to change fields in his late 20s, taught him how to learn and he deliberately abandoned his field of study about every ten years, he continued to be productive into his late 70s. If you find the spoof paper written in his style The Imperturbability of Elevator Operators, by S Candlestickmaker, by one of his grad students, it makes hilarious reading for the geeks. ]

yet Math is applied Logic

[Eric Coleman]

To continue the CKXD comic,

Math is applied Logic
Logic is applied Philosophy
Philosophy is applied Sociology

and "the circle is now complete."

Re:Read the literature... or not

[ColdWetDog]

Anyway, I was taught early on this is one of the main reasons to attend conferences -- after seeing an interesting presentation (or even poster) about stuff close to yours, you go for a beer or two with the presenter and hear all the failures they suffered and the wrong turns they took on the way. And share your own, too.

And that's just one of the reasons I left academic science - people quit doing that. As funding dried up, people dried up. In fact, there were labs who had a reputation of getting it's post docs and grad students to 'hoover' the conference looking for ideas, strategies, concepts and bringing them back and working on some of the more likely leads. If that lab has eight post docs and 10 grad students, they can generally beat your solo effort if they so chose. So you didn't say much. Not much fun.

That and the beer. Man, I hate beer.

Re:Luck...

[ColdWetDog]

But seriously, if you work with old-school biologists, do the world a favour, and teach them that a Gaussian error on a number of cells is dumb and wrong.

I think that entry into either medicine or the biological sciences should require a passing grade on a graduate level statistics course. Only then do you stand a chance in hell to start moving away from a century of misconstrued numbers. In medicine, it's still painfully obvious that most researchers couldn't get past Stats 101. And that is even after they have the manuscript reviewed by a biostatistician (who is probably shivering in a basement closet hoping that the next group of researchers gives up looking for him and goes to a bar.)

Of course, I'd still be fixing cars for a living, but that might have been a better outcome for myself and society....