There are at least two good subjects here, which is at least one too many for a 1,000-word essay. There's a question of constitutional interpretation: is it correct to say that the meaning of the word "commerce" is what changed from 1918 to 1995 in a series of Supreme Court cases, including—but hardly limited to—the three cases you cited? Then there's the quite separate question, familiar from our original discussion of realism: how do we predict what judges will think? (There's a third question, how to teach constitutional law to first-year students, but, subject to your correction, I don't think that's your actual topic.) The first step in improvement, I think is to decide which of these is the bearer of your central idea, and to focus on it.

I think we can agree that the Commerce Clause cases you mention (along with the minimum wage cases, the NLRA cases, Panama Refining, the "sick chicken" case, and many others) are not really about defining "commerce." They are about setting limits to the use of a textually-plenary power as a basis for comprehensive federal statutory regulation of American society. (The quintessential illustration of which is the use by the Justice Department of the Commerce Clause, rather than section 5 of the Fourteenth Amendment, as the primary constitutional authority for the Civil Rights Act of 1964.) This is why, from a realist point of view, discussion about words rather than things is itself both an analytic error and a political position. If what is changing is the political context rather than the meaning of the word, then we are never as certain about our law as we like to pretend.

Which is what rightly connected this inquiry to your other question: if we have to understand the socio-political context in which cases are decided to understand what the very words of the law mean, how can we predict how judges will decide cases? In two sense, this course is designed to be an answer that question, and it should afford you a number of starting points for writing about it. The starting-points should be numerous because the premise of our inquiry is that there are always multiple viewpoints from which we try to understand social phenomena, and it's the consilience of those different views that gives us confidence in our interpretations or predictions. Judges are individual human beings, so we gain some insight from personality psychology, but our material for each individual judge is very limited and unlikely in most cases to offer substantial individual insight. We can read what judges write, however; we can talk to former clerks who have spent much professional time with them; and we can study the practices they had before they went on the bench. We can put those forms of information to work in submissions in particular cases, as I tried to show with respect to one amicus brief in oner of my cases, analyzing the arguments presented to the Justices in light of what the lawyers knew about the individuals making the decision. That might provide further starting-points for the questions the second half of the current draft raises.