Ethics of scientific studies of humans

[perpetualponder]

First I'll discuss an article on the ethical issues involved with conducting experiments to see how many people are fooled by phishing attempts. Issues discussed include:

1. Is it ethical to send forged emails to students in your university in a pretend phishing attempt to see who falls for it?
2. After such a study should you tell the participants what happened? Such a "debriefing" is normally required. According to the paper the primary risk of harm to the subjects is their reactions to being told about it, so some of the studies they described did not include a debriefing step.

My answer to #1 is that conducting such a study is ethical since it's exposing people to something they often receive anyway. Real phishing attacks risk substantial harm, but fake ones do not. However I strongly disagree with their answer to #2, which essentially amounted to eliminating debriefing after it was found to piss off a lot of subjects that learned they'd been fooled. Their decision to skip debriefing sounds to me like it was motivated by reducing risks to themselves, not reducing risks to the subjects as claimed. Secrecy played a large part in historical unethical studies, so I would want more compelling arguments before I would support skipping debriefing.

I'm also concerned that the paper I linked to was co-authored by a member of the IRB and one of the researchers. I would have expected the IRB to stay at arms length from the researchers they are reviewing to ensure impartiality. Publishing papers together creates the appearance of dubiousness.

The second article describes how there's some question as to whether or not flu vaccines and medications actually prevent deaths, but no one knows for sure because most people believe in efficacy and therefore no one could get approval for such a study on ethical grounds. I admit that some possible studies should be unethical on grounds of denying people treatment, but if a substantial, well-informed minority does not believe a piece of medical dogma it should be tested.

The reason for this controversy is that the existing studies of flu vaccine efficacy do not use placebos. It's really hard to deduce anything from a study without randomized controls and placebos.

Overall I'm a bit concerned about the quality of scientific evidence, especially in medicine. It seems to me that the gold standard of experimental evidence should include:

1. Hypotheses to test chosen and announced _before_ study is conducted. (If you test 20 hypotheses and aren't careful, one of them will seem to be true by chance alone.)
2. Experimenters promise to publish the results regardless of outcome to help make meta-studies meaningful. (It can happen that 20 separate teams study some question, 19 get an unsurprising result and move on to something else, while the 20th gets exciting results that are statistically significant to 95% level and publishes them. Net result: the record of publications includes exactly the wrong study.)
3. Experimenters have no conflicts of interest regarding the results, not even the possibility of publication in prestigious venues.

If I recall correctly the medical community is making some steps towards #2, but as far as I know there isn't much of #1 or #3. I wonder if it would make sense to designate around 10% of experimental science funding towards conducting experiments where we think we know the answer, with a requirement that studies funded by such a program must be announced in advance (to address #1) and be published in one particular journal (#2 and #3).