Ideally, the instructor (which is typically myself) wouldn't appear in the video and we would just have two undergrads explaining everything the student needs to know. This is true for the "move the blob" section of our hour of code activity (http://go.osu.edu/movetheblob) and it is also true for our soon-to-be-submitted hour of code activity (http://go.osu.edu/hourofcode2) and our Lunar Descent series on the youtube channel (http://go.osu.edu/STEMtube). Our goal in doing this is to use natural tendency for middle and high school students to be drawn to listen to people only a few years older than themselves (in other words their peers). Maybe this is self-evident enough that I don't need to justify it, but if you look at the most popular youtube channels that kids watch, or music that they listen to, it is typically college-aged people and young adults that are looked to as being "cool".

Having two undergrads on screen is a design goal, but often it is difficult to find a block of time when two students (at least one of them being from an underrepresented group in STEM) can meet with me for 2-3 hours to record videos, and this only gets harder towards the middle or end of the semester. So there are plenty of videos that involve myself and an undergrad on screen.

The rationale behind each video is informed to some degree by my experiences using many of these activities in introductory physics classes at OSU, which I have been doing since fall 2014. So I already have an idea of some of the subtleties that need to be communicated for each exercise. Regarding equations, sometimes it requires 15 minutes or more of explanation to justify one line of code. Our wave interference activity (http://go.osu.edu/STEMtube) requires almost 40 minutes of algebra to explain one line of code. In other activities we use much less time for pure math. The Lunar Descent video has less than 4 minutes of math. It really just depends on the video.

The youtube videos are intended to supplement the way these activities are used in a typical physics or physical science classroom. In a classroom there is a detailed step-by-step sequence of tasks, some of which involve coding while others simply involve playing around with the code (for example, change the mass of the ship in asteroids to find the value that allows you to survive in the asteroid field for the longest). You can find most of our activities at this link: http://go.osu.edu/physics_coding To answer your question, there is significant coding that is involved (not just changing parameters). In a few of the exercises, for example, students will modify a 1D code until it becomes a 2D code. There are also "verification" tests where students compare the output of the code to what one would expect from using a formula. Typically these match within a few percent.

You can find most of our activities at this link:  http://go.osu.edu/physics_coding

To answer your question, we always give students a code that has some limited functionality and they need to modify it to produce new behavior. For example, we give them a 1D code that they need to modify until the object can move in 2D.

We do try to take advantage of student's prior intuition as they modify the code. For example, the student may have a intuition for how the asteroids game should behave which is useful for completing the exercise, but we also provide links that allow the student to interact with the program at an intermediate stage (without seeing the source code) in order to check that they are on the right track. This frees up the teacher from having to micromanage each student.

The interaction between learning the physics and learning coding does not primarily go in one direction or the other in my opinion. There are some cases (for example projectile motion) where the student should be drawing from their physics knowledge to get the code to do something familiar. But other cases (like "Planetoids with Torque" or the "Particle Accelerator") the program is illustrating something dynamic that might otherwise be hard to visualize. Coding in the way that I do it is a kind of method for showing what happens instead of saying what happens. For example, adding a spring force in the "Planetoids with a spring" activity, naturally produces oscillations. The oscillations aren't something that we have to add to the code with an already fairly complete idea of how it should look in the end.

Good questions. We use STEMcoding activities in both formal settings, like high school physics, and informal settings like the ASPIRE physics camp for girls at OSU. We do have survey questions which probe attitudes about STEM that we can analyze to see if there are gender differences, we just haven't had time to dig into those data yet.

To answer your question about absolute beginner programmers, we carefully designed our hour of code activity (http://go.osu.edu/movetheblob) to be particularly straightforward, even intuitive. Code.org reviewed it as being appropriate for students as young as 6th grade. The emphasis on the physics of video games also helps novice programmers to stay engaged because they know we are going to produce something fun and interactive in the end. Finally, we have a learning management system (http://stemcoding.osu.edu) that has good bug finding and allows teachers to quickly run and give feedback on student codes. This is especially helpful for students who have never programmed before.

I'm not sure what you're asking in terms of "Do you see any transfer of what they learn by doing STEM coding in other classes?" We are working to probe gains in student conceptual knowledge of physics from completing our coding activities. We've designed everything with the classroom in mind, but the activities are also enjoyable enough that you could use them in informal education too.

Some of our preliminary findings are posted here: https://arxiv.org/abs/1701.01867