Are user surveys considered "human experiments" by the U.S. National Science Foundation?
In almost all circumstances you need IRB approval to carry out the "experiments" that you are contemplating. The reason isn't that taking surveys is, somehow, dangerous, but, rather, that the data you collect might be sensitive. Moreover the subjects of the survey might have privacy concerns and expectations.
So the issue is about the data collected. How will it be maintained? How will it be disseminated? Is it (even potentially) name linked? Etc. The treatment of the data implies some treatment of the subjects.
While there are a few exemptions to IRB approval, the best place to learn if your study falls under exemption is from the IRB itself. It is better to ask and be told that approval isn't needed than to make assumptions.
For your specific situation, the fact that you may be using human models as part of the materials introduces another aspect. An IRB will want to be assured that you have proper releases from everyone.
So, ask.
Since you are concerned about a grant writing process, rather than the experiment itself, it would be wise to run a version of your proposed experiment, not just the questionnaire, past your IRB so that you can assure the NSF that you have your processes in place.
The NSF policy