Thursday, 28 March 2013

Scientist Makes Mistake!

University tests prove conclusively that scientists are sometimes wrong.

Does this mean we should disbelieve everything scientists say?  Should we toss out of the window widely-attested scientific findings about such important things as the age of the earth, the health effects of tobacco or the processes of climate change?

Funnily enough, no.  We should believe them when they are right and disbelieve them when they are wrong .  The more difficult question is: how can you tell?

The answer to this is in two parts.  The first is, you need to ask other scientists.  Scientists themselves have two terms for this - peer review, and replication.  Peer review is where you get other scientists to look at a work of science and check that the methodology is sound, the evidence has been properly gathered and supports the conclusion, and so on.  Replication is what you do when results are tentative, or based on small samples - you run the experiment or test again in slightly different circumstances to see if you get the same result.  If you keep doing these two things consistently, then over time scientists will correct each others' mistakes.

This leads to a follow up question - how can we trust the integrity and independence of the scientists on whom we are relying?  How do we know they're not lying or fudging their results to favour a particular point of view?

Surprisingly, the answer to this question turns out to be not public floggings but public funding.  This is because of what economists refer to as the Principle of Funder Directed Note Sequencing.  First identified in relation to players of wind instruments, this Principle has been found to be applicable to a wide range of professions.

It suggests that scientists who are employed by particular organisations have a strong financial incentive to report findings that serve the interests of their employers.  Scientists paid by drug companies will be tempted to find that more expensive drugs will improve health.  Scientists paid by coal mining companies will be sorely tempted to find that carbon emissions have minimal effect on the earth's climate.

If follows, then, that if you want scientists to act in the public interest, it will be easiest for them to do so if paid by the public.  Ideally they will be tenured and protected from direct political interference so that their findings are not open to political manipulation.  Their data and research will be freely available for anyone to check, review, replicate or just muck about with to ensure they are not deceiving us.  They will be encouraged to speak their mind even if their findings are not popular.  They will be encouraged to research things that are important for the wellbeing of society even if there is no money to be made out of their findings.

You pay peanuts, you get monkeys.  Monkeys are great experimental subjects but very poor at scientific method.  For that you need highly trained scientists, and if you want them you have to pay what they're worth.

No comments: