One of the advantages of doing something in a low stakes way is that you can try something and no big deal if it is a total screw up. And if you do try out things this way, then an awful lot of the time it will turn out to be just that, a total screw up. So if your goal is to do something in a professional manner, to be well regarded for the service you are providing, and to get others to utilize that service, the "right approach" is to quickly identify best practices perhaps by scanning the prior research on the subject, embrace those, and then professionalize the way that embrace occurs as soon as possible. Most of us, I believe, do this with our teaching. And to a large extent, those of us who support instruction do likewise.
Some time ago I read A Life in School by Jane Tompkins. The book was recommended to me on the particular issue of where instructor ego belongs in teaching. Tompkins was a Professor of English (and I believe is the wife of literary theorist Stanley Fish). The book is simultaneously engaging and unsettling. Afer being completely miserable about her own teaching, Tompkins came to the conclusion that she was getting in the way of her students' learning. She kept modifying the approach, producing some interesting outcomes but never ones that satisfied her that she had "found it," that right way to conduct a class.
For Tompkins experimentation was a kind of penance. For me it's a form of self expression. I don't think it is fundamentally different for me to scheme up an experiment with teaching method than it is for me to design a module in Excel that presents Econ concepts in a novel way, or for that matter experimenting with a theme for this blog. I thrive on trying things. I'd much rather learn that way, at least at this point in my life, than reading the literature and accepting best practice.
Part of the reason is skepticism. People miss things, or frame things slightly differently and then come up with conclusions other than what I would have come up with had I been in their shoes - not always to be sure but enough of the time that I want to try it for myself.
The thing is, I think that experimentation, as an overall approach rather than focusing on one particular experiment, improves instruction. Frankly, too much teaching is dull. It is safe and it is dull. There is more excitement and stimulation when the outcome is unknown. And if the outcome is unknown by the instructor, then surely it will be unknown to the students as well. It helps, in addition, for the outcome to matter to both the students and the instructor.
Apart from Tompkins, who teaches this way? Are there particular institutions or departments within institutions that encourage it? One of the reasons I got into the technology support business is because I thought that adaptation necessitated some form of experimentation - instructors had to try out the technology enough to have it fit what they are trying to accomplish. When that happens, I think things will typically turn out really well (though the instructor might bitch about putting in too much time).
When it doesn't happen --- well that's why some focus so much on what the technology will do. The cool technology let's the educational technologists experiment. That helps their learning. It keeps them sharp. But does that lead to related experiments by the faculty? In most cases, I don't think it does.
In the political economy of ed tech, there are two main problems right now. First, faculty development is being squeezed. It is being squeezed by tight budgets but also because CIOs focus on the technology itself rather than its use. But second, we treat faculty development as promoting best practice rather than as encouraging experimentation.
I don't know how to get out of this bind, but I do feel we're not in the right place.