Psych and Social sciences are the "harder" science

Pythagorean

Fanelli D (2010) “Positive” Results Increase Down the Hierarchy of the Sciences. PLoS ONE 5(4): e10068. doi:10.1371/journal.pone.0010068


http://www.plosone.org/article/info:...l.pone.0010068

Pyrrhus

I have been discussing this repeatedly, and extensively every few months on this forum, and elsewhere. Someone wrote it in a paper, and published it. Now, I can send them this paper, haha.

Pythagorean

Do you any anecdotes to offer? I feel like I've had some pretty robust physics professors, but I've read very little of their published work and I'm probably not capable of an objective critique. I have been considering a psychology PhD for sometime (I did my undergrad in physics; finishing an MS in computational neuroscience currently).

D H

Six comments.

1. Someone will inevitably post the XKCD comic on this topic. I'll leave that task to someone else.

2. This proves the adage "Politics ain't rocket science. It's much harder."

3. On a more serious note, there are a lot of different ways to interpret this paper. One is that the soft sciences are indeed soft. They are much more subject to confirmation biases than are the hard sciences.

4. "This study analysed 2434 papers published in all disciplines and that declared to have tested a hypothesis." They did this by searching papers (titles only? titles+abstracts? full text?) for the phrase "test* the hypothes*". I don't recall seeing that phrase in the articles I read. I just looked at several 'test of physics' type papers as a test of the hypothesis "physicists don't use the phrase 'test* the hyothes*'". That phrase is just not there for the most part.

5. Some disciplines use the phraseology tested for in this paper, others don't. Some may use it preferentially to report positive results, others tend to use it more when reporting negative results. There is no accounting for this potential bias in the paper.

6. Data. Where's the data? They analyzed 2434 papers. Which ones, and how did they classify them?

Pyrrhus

Good comments.

I have done meta-analyses in the past. In fact, I published one too. The main problem is selection bias. The data is selected by the researcher, and this is very crucial. The researcher must be careful to have searched the literature in depth, and even if done so quite exhaustively, there are also "grey publications" that are likely to be ignored. However, I suspect that even if bias is present is likely too not throw the results to far off.

Pyrrhus

Yes... Search Mathematical Models in Economics in this forum. Most posters seem to think that Math only "works" in Physics. A very ridiculous claim. I don't know about PhD in Psychology. The only area of psychology that I read periodically is psychometrics. I may read a bit about perception.

Pythagorean

I heard a Nature podcast from an economist criticizing the way economists use mathematical models (compared to physicists). The nature of the criticism was basically that economists don't have much for controlled experiment, so elegance becomes a determining factor of what kind of math survives, which isn't scientifically sound.

I am not familiar with modern economic theory, so I have no opinion, just regurgitating the podcast hearsay.

Pyrrhus

Noted, and I disagree. Not willing to fight this mathematic modeling argument anymore. Also, economists do have controlled experiments for some cases, but for most we are left with estimating models by controlling for as much as we can...

M65

The issue of controlled experimentation can be dealt with in two major ways I can think of.

One is agent-based modeling. Here you can create and recreate an environment as many times as you like, with controls. Here's a great example of a very successful agent-based sim (sorry can't post links yet):

www (dot) u (dot) arizona (dot) edu/~mlittler/artanasazi.htm

The other, more obvious one, is to do experiments with real people à psychology.

More generally speaking, this artificial distinction between "hard" and "soft" sciences on grounds of whether experiments can be done, complexity, etc. is nonsense. "Real" physical systems can easily become as complex as more well-known social complex systems. (See e.g. the three-body problem.)

To paraphrase Duke Ellington: there are two kinds of science: good science and bad science. As the late Hans Eysenck pointed out, there are concepts in psychology, like reinforcement learning, that are on far sounder empirical footing than some of the more speculative areas in modern physics.

So trying to rank sciences in terms of "hardness" seems pretty silly.

ps first post here; hello

FredoX

Yes please!
We need more physicists and mathematicians in psychological science.
But be prepared to have papers rejected because you used "vague terms like stability, self-organization and attractor landscape" (yes this is an actual quote from a reviewer on a paper I co-authored). I have seen papers by colleagues rejected in the Journal of Mathematical Psychology because "the mathematics you use is too difficult for our readers" when basic differential calculus was used. Yes you read correctly. That journal actually has mathematics in its title.

I believe psychology as a science is in serious trouble and I will provide some arguments below. This serious trouble is slowly being revealed to the general public and it is probably best captured by a satirical piece published in the high impact journal: perspectives on psychological science by Arina Bones (We Knew the Future All Along: Scientific Hypothesizing is Much More Accurate Than Other Forms of Precognition—A Satire in One Part
Perspectives on Psychological Science May 2012 7: 307-309, a poster version is available at her website: http://www.projectimplicit.net/arina/B2012.pdf )

It discusses the paper that started this thread and links it to some recent painful events for psychological science being the publication of the paper by social psychologist Daryl Bem in which he claims to have found evidence for the existence of Psi (http://dbem.ws/FeelingFuture.pdf ) and the largest scientific fraud committed by Dutch social psychologist and media darling Diederik Stapel.

My concerns:
- If you are a physicisist, read the paper by Daryll Bem, published in one of the highest impact journals of psychology and try stay sane while wondering how such a paper could ever be published in the name of science. Bem claims to have found evidence that his students can feel the occurence of future events above chance, especially men detecting erotic events. He claims this ability may have evolved because there is an evolutionary advantage for males to know where future possibilities for reproduction may occur. (no I am not making this up).
The editors claim the paper has been rigurously reviewed and it has been accepted because there seemed nothing wrong with the methods used. The fact that these results and their interpretation would turn almost all scientific disciplines from physics to evolutionary theory upside down (tear down that LHC!) is apparently not important. The paper should have been rejected on theoretical grounds. This concern may be summarised as: there is NO theory evaluation in psychology, all so called theories are actually theories of construction (that save the phenomena) they are not theories of principles (in the sense that Einstein proposed).

- if psychological science is indeed is the harder science, then a first year of psychology at graduate level should at least include a course in the major achievements of physics so we could learn from their efforts to understand the behaviour of non-living matter. Instead, most schools of psychology do not venture beyond teaching probability theory and inferential statistics dating back to the 1900s. The study of behaviour is the study of change, why not at least start to teach the mathematics of change? Summarising: many subdisciplines of psychological science are completely detached from contemporary developments in philosophy, mathematics, physics and biology.... And this cannot be maintained for much longer. Engineers and computer scientist will take over most of the study of human behaviour.

- psychological science as a discipline is not capable to initiate a Solvay-like conference in which a consensus is achieved on a formalism describing what it actually is the discipline is studying. What is cognition, what isn't? What are the relevant levels of a system one should study? What are the interesting phenomena one should study? Currently (and unknown to many psychologist) humans are studied as ergodic systems. Certainly a questionable premise. It may be hard to believe for a physicist, but in psychological science I can claim to study mental representations of letters hypothesised to be involved in reading, without having to specify what these representations constitute of, how they relate to neural activity, or how they may be described or modelled mathematically. Interpreting results as if they existed is sufficient. Summarising: without a consensus formalism like the quantum formalism about the system psychology is studying, the interesting levels in the system that should be studied psychological science will not advance to produce theories of principles like the physical sciences have.

So is it a harder science in terms of object of study? Yes it probably is, but few working in the field realise what that actually entails: studying and learning from other disciplines of science. At least embrace mathematics as a tool to formalise theories and evaluate their predictions and empirical accuracy.

Any tips on how to achieve this are very welcome :)

chill_factor

I read outside interpretations of Daryll Bem's work and it seems he is making the case for predictive "modeling" and extrapolation by the brain subconsciously, not that people can actually see the future.

If that's the case, would his argument be more secure?