The honourable physicist

CERN's Alpha Project mentioned below - image courtesy of CERN

I had a very interesting exchange with a physicist the other day which I think really helps underline the difficulties of getting science right.

I had asked this person 'do you have any personal expectations (or even hope!)' about the outcome of an experiment. His response was 'As an experimentalist, I must be completely unprejudiced about
the outcome of our experiments. So I don't have any expectations.'

Now, in answering one part of my question he was entirely correct and proper. But as far as the other part goes, assuming he is a human being and I wasn't communicating with a robot, he is fibbing. The right and proper part is having no expectations. We now recognize that it is very important that an experimental scientist does not prejudge the issue and expect an experiment to come out a particular way. This is because there is plenty of evidence of the 'experimenter effect' where an individual's expectations colour their interpretation of the data.

The archetypal experiment designed to demonstrate this kind of unconscious experimenter bias was a trial undertaken in 1963 involving albino rats and, more importantly, young scientists who believed they were experimenting on the rats, but in truth were themselves the experimental subjects.

Robert Rosenthal and Kermit Fode of Harvard University set up an experiment where twelve psychology students were given five rats each to test on a simple T-shaped maze. All the rats were from the same stock, but one group of the students was told that they had especially bright rats, naturally suited to solving mazes, while the other group was told that their rats were of a less able strain that struggled with maze solving. Those with the bright rats were instructed that their animals would show clear learning during the first day of running the maze and after that their performance would improve rapidly. The subjects with the “dull” rats were told that their experimental subjects would provide little evidence of learning.

Both “types” of rat (bear in mind that all the rats were identical in ability) did prove to have performances that improved over time, but every day over the five day trial the “bright” rats were recorded as performing better than their peers, achieving successful runs up to twice as frequently as the “dull” rats, and getting to a successful conclusion significantly quicker than their supposedly slow counterparts.

When dealing with bright rats, experimenters could have encouraged them more, given them more positive handling, which could have influenced actual performance. This is unlikely to be a problem with a physics experiment. However - and this applies to all kinds of experiment - true improved performances were not necessary as the experimenters could easily have biased the results, even though they could be totally unaware that they were distorting the data.

One possible approach that would shift the results to match the experimenters’ expectations would be if they counted borderline runs as successful with bright rats, but not with dull ones. They might also have decided to be selective about which results to record because of some apparently sensible reason (perhaps the rat was distracted by a loud noise), cherry picking positive results. And there are other, more subtle ways available for experimenters to fool themselves.

So even though some of the greatest scientists in history have had a tendency to ignore adverse results because they knew their theory was right (Newton is a good example), it is isn't an appropriate thing for a modern scientist to do. The physicist I was emailing was absolutely right to have no expectations.

However, the reason I suspect my contact of fibbing is that it's one thing not to have expectations and it's another not to have hopes. The only kind of scientist who could be totally devoid of hope and simply carry out an experiment as a wholly neutral observer has lost his or her humanity. You might think they shouldn't care about the outcome, that all outcomes should be equally interesting, but the fact is that all outcomes aren't equally interesting - at least this is true in many cases. And to pretend otherwise is a subtle form of deceit.

Take, for instance, the Alpha experiment at CERN. If everything goes to plan, phase 2 of this may well tell us in the next few years, the answer to a question that has fascinated physicists for a long time. Under the influence of gravity, does antimatter act the same as ordinary matter, or does it feel an opposing force? Would it float up rather than fall to Earth? We just don't know, because we've never had enough antimatter to measure the very weak pull of gravity.

Now, the emotionless robot scientist, the hypothetical textbook scientist, doesn't care which outcome we get. Each has exactly the same reward value - it is a result, tick, move on. But I don't think any human can honestly say they wouldn't get a thrill if antimatter acts as if it was experiencing antigravity and is repelled by massive objects. That would be just so much more exciting than if it acted like ordinary matter.

That being the case, while I can only commend the scientist I was emailing for his lack of expectations, I am very sad if he truly was without hope.