Thanks for your questions, Jeff. At a high level, the Minecraft commands have some neat functionality that can be hard to replicate using blocks. Particularly, the in-game commands can live in the game, as opposed to being tied to individual players. Another factor is that the block based environment does not allow for undo actions after the code has been executed.

Relatedly, we find that many students seem more interested in doing coding with commands than with block coding or Python. 

With regard to analyzing spatial reasoning, most of that has been qualitative, but we are in the process of doing some automated work with eye tracking data and also using log data. We have used some automated approaches of log data to locate important moments within Minecraft summer camps, but are just getting started with using automated approaches.

+Jeff and +Marcelo this is a new idea for me, can you please say more about how teachers are/would be involved in the process of automating measures of engagement, collaboration and spatial reasoning? This is something that I did at the qualitative level as a classroom teacher, and wonder how automation might support(or hinder) this task if I were still in the classroom.

Good question Pati. I think one challenge with automation is that it can over simplify what students are doing. When we reduce the extensive building activity to a small set of numbers about how many blocks moved or destroyed we might over look a lot of the rich ways that students are thinking and engaging in the activity. In my work, I've mostly used the log information to see when students may be deviating from their typical build patterns. So, there might be moments where students are building much faster than they normally do, or much slower than normal. This might key me into important moments to look at qualitatively.

I tend to think about automation as something that can help us see different points of salience in our data, and less in terms of producing labels about students being "productive" or "disengaged".

Using something like eye tracking data, I might get a better window into what students are paying attention to during the building process. In our work, we see this in the form of attentional or visual anchors. As we get better at computationally analyzing this data in ways that don't oversimplify user behavior, we might start to see different ways that students are demonstrating changes in spatial reasoning practices, for example.

@Pati 100% agreed with Marcelo. I'd add that I think automation or AI-driven components should be tools to be leveraged by the teacher - a lot of the time the interest in them is to replace the teacher to cut costs and speed up exposure to material. If we went in the opposite direction, and saw automation as a way to help teachers find more venues for assessment and feedback to deepen project-based learning we'd be in better shape. The kind of R&D time required for it I've only witnessed at the private and university level, or in informal learning settings, unfortunately.

(also Marcelo I just noticed your advisor was Paulo Blikstein - I was the director of the Champaign-Urbana Community Fab Lab for half a decade - the world is a small place :)

We are in the process of doing additional testing to see what types of features people would like to have. In parallel, we are exploring the use of gesture-based input (using a standard camera or a wearable device with an inertial measurement unit). The other major modification that is still in development in allowing participants to build using standard wooden blocks.

Broadly speaking we have seen Minecraft integrated into several different learning contexts. Within the current project we support middle school classrooms with lessons and activities related to coding, architecture, science (mainly sustainability), and pre-algebra. In some cases teachers are looking for connection points to specific learning standards, while in other cases it is used more as an enrichment. For example, we also partner with the local library and the YMCA to offer summer programming with Minecraft.

On the topic of learning standards, other practitioners and researchers have worked to identify some standards that are met within existing Minecraft Education Edition lessons. Seeing teacher creativity in coming up with ideas that align with different countries' standards has been particularly neat.

Peers teaching each other is one aspect of what we look at across the various contexts that we work in. We have seen less of this during Covid-19 with students being remote, but it does certainly take place. We also see students using a number of web-based resources created by other kids to facilitate their learning of new approaches. Beyond this, our focus has largely been thinking about how youth engage in peer learning related to computational thinking. For example, when do students share code with one another versus explaining how to do something? When do students simply do something for someone else versus pointing them to a web-based resource? And when do students ask an adult to assist? Anecdotally, many of the younger students that we have worked with (K-2) seem more likely to want help from an adult than from a peer. For middle school students they have seemed more willing to accept help from a peer.