[I have literally no qualifications to have an opinion on this and everyone should take it with many grains of salt.]
If I ran the world, there are two things I would do to make research on human beings more representative and have higher sample sizes.
First, there are lots of cases where there are multiple treatments for a particular condition, but there isn’t any really solid evidence about which one is the best. (Autism treatment, of course, is an example close to my heart, as is antidepressants.) Right now, people are prescribed a variety of treatments, but the data is pretty much useless, because maybe doctors who have rich patients favor Treatment X and doctors who have poor patients favor Treatment Y and then it’s no surprise that people who get Treatment X fare better.
Instead, we should use this as an opportunity to run a randomized controlled trial! If you’re depressed, your doctor should press a button and get a random SSRI to prescribe you. There would be no extra burden on the patient, because you could use the information about effectiveness and side effects that should be in the patient’s medical records anyway. Result: enormous sample sizes with plenty of real-world validity.
Research should of course be consensual, so the patient can opt out and rely on their doctor’s clinical judgment. I think this should require signing a consent form that says “I UNDERSTAND that there is no clear evidence that one treatment is better than another. I UNDERSTAND that my doctor’s clinical judgment may be influenced by many factors, including but not limited to advertising from pharmaceutical companies.”
This would be easiest, of course, to do in a country that, unlike the US, has a national health system. But I think it would be possible in the US as well.
Second, we would recognize that participating in human-subject research is a necessary part of contributing to society, the same way that jury duty is; everyone has to participate in creating science that benefits everyone. Researchers with an academic affiliation, or who are sponsored by a researcher with an academic affiliation, have the right to pull a random sample of people from the voters’ or drivers’ rolls. If they wanted to research a subpopulation of people, they could send out questionnaires (with an online option, because we are in the 21st century here); only the people who are eligible for the study need show up at the study site.
People would have a right to refuse to participate in any studies with an actual risk, such as drug trials, weight loss studies, or studies which involve sensitive issues like violence or stigmatized traits. However, certain studies would be certified by the institutional review board as minimal risk: for example, spending an hour talking with an interviewer about household chore division or filling out a new personality instrument or having your blood pressure taken before and after you watch a video about puppies. IRBs are currently absurdly strict about ethics, so I expect they will not certify anything as minimal risk unless it is actually minimal.
Once again, the result is studies with large, random sample sizes and a complete end to psychology actually being the discipline of Introduction to Psych Student Studies.