Assessing survey validity is much more complicated than asking whether we "think" it is valid (although I suppose you can argue that would be a form of "face validity").

There are several steps to this process, and there are much literature on survey construction and testing. First, you should generally ask experts in the particular field to review your question-set. Here, I'd argue that I don't know what you mean by "human behavior" or how these particular questions address the multifactorial nature of human behavior? So you'd have to consider adding additional questions and dropping irrelevant questions.

Assuming you make it that far, you then need to consider the number of responses. Here you offer a 4-point scale. Generally, an odd number of responses is recommended so that responders have a middle score to choose from.

Next, you'd want to have a sufficient number of people respond to the survey so that you can then investigate the psychometric properties of the survey.

You'll want to perform factor analysis or item response theory as well as test for reliability.

These are only a few of the steps you'll need to perform, but they give you flavor of what to consider. 

There is a large literature in survey methodology on asking behavioral frequency questions.  Journals like Public Opinion Quarterly regularly publish papers on this.

Thanks Ariel! Thanks for your answer. You were right, I wanted to ask experts here. You were the first :) Perhaps my question was not clear enough, so I amended it. I think I am very well aware of methodological and statistical methods. Therefore my my question rather addressed the RESEARCHER EXPERTISE and EXPERIENCE. I did not make that clear enough. Beg your pardon!

Cheers Daniel, I did not know that. Thanks for the hint, i did not know what to "google", but I think I will have a look at the journal. In case you had some papers at hand, I would really appreciate your help. But anyway, thank for the tip!

I did a couple a long time ago, but have not been working in that area for a long time so not qualified to give more up-to-date info. I do have a couple of papers that I will attach, but they are old

Article How response alternatives affect different kinds of behaviou...

Article How much is 'Quite a bit'? Mapping between numerical values ...

Article Telescoping of landmark events implications for survey research

Article Intensifiers in Behavioral Frequency Questions

Article Survey Questions About the Frequency of Vaguely Defined Even...

Alexander, it seems to me that questions 1 to 3 may be easily answered by your "totally agree -- totally disagree"-response set. The reason is: agreeing to some vague numerical concept ("regular basis", "regularily", "easily") seems reasonable because there is something beyond mere frequency in the meaning of these questions. Q1: social norms on "healthy behavior" require physical exercise, Q2: Participation of students into "topics" represents an accepted standard of didactics.Q3: To get too often into panic is a matter of mental health and therefore touches the connotation of stigma. Asking your questions by the exact wording you mentioned avoids being to strict and leaves some room for interpretation. Therefore "agree/disagree" per se is not a behavioral report but more like a goal for "good behavior" that your respondent agrees on.

By contrast: dog walking is a behavior with a very sharp difference between "daily" and "less then daily", simply from its numerical  meaning. In the same line, smoking "7 or more"  cigarettes daily is also a strict numerical question, especially because you make this clear by the supplemental wording  "or more". Therefore my impression is that "totally agree -- totally disagree" is not the best fitting repsonse set here. I would prefer a simple "yes/no" format.

Kind regards

Ulrich Frick