Someone asked a question about the ethics of approaching a panhandler and sharing information about a research study. I don’t know all the details, but apparently, this study was examining veterans of the Iraq war, and this panhandler was holding a sign saying something like please help a veteran of the Iraq war. There was some concern about whether the monetary incentive would be disproportionate for someone who had to beg for a living, or it might be a problem if the panhandler was given money and a flyer about the research study at the same time. I discussed some of my concerns about this study, but it was from the perspective of statistical validity rather than from an ethical perspective.
I suspect the researcher is trying to help out someone who obviously needs help, but that adds a lot of complexity to the research design just to gain one additional patient. Go back and look at how the convenience sample is defined in the original protocol. It probably mentions getting patients through a clinic referal or something like that. If you change the recruitment process for one individual, that is a protocol violation and it also harms the representativeness of the sample (though a convenience sample is never all that representative to begin with). The other problem is that if the researcher adapts the recruitment process in a haphazard way, that makes the research less repeatable.
Now it may be that the protocol never mentioned how subjects were going to be recruited. If so, shame on you for approving such a vague protocol. But if there is some detail about the recruitment process, it surely did not specify random encounters with panhandlers. So hold this researcher to his/her original plan. It may or may not be ethical to recruit panhandlers, but it fails from a statistical perspective, so that in itself is enough justification to say no.
You can find an earlier version of this page on my original website.