What is a scientific prediction?

Backreaction:

 After my bachelor's degree I changed field from mathematics to physics. I wanted to understand, at least to some extend, the world around me. Mathematics seemed to entail an infinite amount of possibilities whose each and every relevance wasn't clear to me while physics is tied to reality by experiment. Basically the reason why I'm a phenomenologist today is that I know how easy it is to get lost in the mathematical universe, and that this getting lost has clearly addictive qualities.

In the last decade in high energy physics one could notice a trend towards more phenomenology. While I welcome this for obvious reasons, here as in any aspect of life one can desire too much of a good thing. I've read quite a few of papers where the word "phenomenology" was used merely as decoration, and in other cases "phenomenological" is essentially an excuse for inconsistency. Such fashion trends in the community and their side-effects however aren't really surprising. What is surprising though is that the demand for "predictions" has been picked up by the public and has been used sometimes inappropriately as a measure for scientific quality. Thus I thought it would be worth clarifying what a scientific prediction is and isn't.

1. A scientific prediction is a statement about a future event.

That is to say the prediction was made without knowing whether it is correct. Strictly speaking this doesn't necessitate it to be about a future event, but it's hard to reliably show one didn't know about a measurement that was already made (a possible scenario is that available data wasn't analyzed with regard to a specific hypothesis). If you calculate the cosmological constant to be in agreement with the today measured value it's not a prediction, it's a postdiction. While it is certainly preferable to calculate the outcome of an experiment before it was done to avoid confirmation bias, this isn't always how it works. Sometimes theory is ahead, but sometimes the data is in already and awaits a theoretical explanation. Science is in the first line about understanding. Explanations, even if not predictions, are valuable. However, if there is something genuinely new about an explanation it will typically also imply new predictions.

2. A scientific prediction is based on a scientific theory. That means in particular it is reproducible (by everybody with the appropriate education), consistent, and the theory it is based on is not in conflict with available data already.

If you dreamed a meteoroid will crash into the White House on New Year's day, that's a prophecy, not a scientific prediction. Same for the recurring remark that the LHC might create angels at 14 TeV collision energy. That's funny, but not a scientific prediction. You may find it inconvenient that your theory be reproducible because this means other people must be able to understand it without your help. However, if you aren't able to communicate how your theory connects to state-of-the art science, it's your fault and not everybody else's fault. Likewise, if your theory comes with a prescription only to use it for this effect, but not for this effect because it doesn't work there for reasons only you understand, that's not a scientific theory. If your theory predicts a fourth lepton generation but has the side-effect that atoms are unstable, tough luck. See here for what it means for a theory to be consistent.

3. A scientific prediction is falsifiable. In practice this means often it's implausifiable.

Falsifiable means that your prediction can be shown to be wrong. This typically though not always implies the prediction has to be quantitative. "You will die" is not a scientific prediction: if you're still alive in 200 years, it could still be you will die someday. "At least 99.99% of people your age who smoke 1 pack per day will be dead 80 years from now," is a scientific prediction because 80 years from now you can look at the data and see whether I was correct (or rather somebody else will have to look, cough).

In physics, scientific theories often contain parameters and a measurement does not indeed falsify the theory but constrain the parameters until they are constrained so much it's point- and useless to consider a theory further. A good example is Brans-Dicke theory. If there are deviations from general relativity of the Brans-Dicke type, they are so small you can forget about them. Same for violations of Lorentz-invariance, time-variation of the structure constant, and so on. These are not falsified but tightly constrained. Reason why in high energy physics new theories are often not actually falsifiable is that for a new theory only small deviations from already extremely well confirmed theories are allowed. We know that our present theories are correct to high precision and new theories cannot differ by much or they are false already. If the deviation is too small however, it becomes unmeasurable.

As you can guess, implausibility is not binary but a continuous scale, thus people frequently disagree on exactly when to discard a theory. (MOND anybody?) As far as I am concerned everybody can decide for themselves how to waste the time of their life, as long as they don't waste other people's time.

That a theory is implausified rather than falsified is quite common if very good theories are available already as in theoretical physics. But in other fields falsification is easier. The dopamine level might just not correlate with schizophrenia. Holy water doesn't sanitize your hands. The world is older than 10,000 years, etc.

A statement is a scientific prediction if all three above explained requirements are fulfilled. If you you have suggestions for improvement of my definition, please leave in comments.

What you should have learned from this post:

• Not every statement about a future event is a scientific prediction. More commonly it is a prophecy.

While making predictions is a merit of a theory, it's neither a substitute for scientific quality nor an indicator for promise.

Explanatory power of a theory can be valuable even if not predictive.

Theories in physics often have free parameters that can be constrained, rather than an ansatz being generally falsified.

"You do not really understand something unless you can explain it to your grandmother." ~ Albert Einstein

Posted via email from Jim Nichols