Hi Lorna,

We've looked at preservice teachers at a few different points in their program and have observed a few patterns (some more informal some published or in review). Much of this supports the prior scholarship on teacher noticing and eye-tracking with video generally. First, the less experienced (i.e., experience with actual students), the more people tend to look around (almost as if they are in awe of being in a classroom). Oddly, many of these individuals report a higher sense of presence than others with more experience (not that the latter reports low presence, but the former reports a sense of 'hyper-presence'). As folks gain more experience, they tend to slow down their head (or Field of View) movements and begin focusing more and more on teachers and students, before transitioning to students particularly. One interesting observation we made (just by looking at the x- and y- coordinates of the gaze data) is that experienced teachers look lower with their eyes than novices. This is because they look at the children's work more whereas novices who have gotten into being more student-centered look at the children's faces more than the tables where they are working, writing, counting, etc.

For the eye-tracking, we are working to combine observational with a form of AOI (areas of interest / regions of interest if you prefer) that uses machine learning. Right now, we have it where we can distinguish between teachers and students, but are fine tuning the AI to differentiate between specific students. This is also important down the road as we hope to use this with multi-perspective 360 video (so we can see if they are looking at the same or different students at different positions in the classroom). Our plan is to use a few different statistics with the student gaze distributions (including the Unalikeability coefficient - we find it to be slightly better than Gini, Hoover, or Theil). For using eye-tracking in an observational way, we are in the midst of qualitative analysis of those videos and hope to provide a better description of what the statistics tell us (based on the qual). 

I will note that we have used pre-defined AOIs when we analyze 360 viewings with the field of view (no eye-tracking). For example, our platform gathers this data for when people watch on their laptops at home. It's not as precise as eye-tracking, but does give a very nice approximation (particularly when considering most of a person's eye gaze tends to be within a particular region of the center of their field of view). 

Great questions! 

We tend to notice more movement with the flat screen than the headsets (see Kosko et al., 2021 in Journal of Teacher Education). We have one paper we are finishing that does a similar comparison and confirms this a bit. Part of it is that there is a greater sense of immersion with the headsets than flat screen and we do not tend to look around as haphazardly if we are more immersed. The big question for us has been how much the benefit of the headset matters (regarding cost, etc.). No clear answer yet there, but technology advances may make the question moot.

We haven't noticed anything with IPD, but none of our 360 videos are 3D. I wonder if that is an aspect to consider. 

Regarding presence, we had wondered about this but there aren't any observable differences across all the data we have collected (we did gauge perceived tech savviness, prior use with VR, etc.). Tech savviness sometimes came out as a factor but not consistently across studies. 

Regarding approaches other than 360 video, yes they are considered but some of the costs need to come down to make them a viable alternative for us. There are scholars doing great work with VR and digital agents, and we are dabbling with holograms, but nothing significant to report just yet. 

We are working on data from an eye-tracking study now that shows the differences between experienced and novice. We also see differences between preservice teachers with and without certain field experiences (essentially upper elementary vs not as that relates to how they look at children's fraction reasoning).

As far as differences between folks who have background teaching marginalized populations - we have not looked at this, but it would be an excellent study. I could see how the technology (and Praxi as a research tool) could be used to examine differences in where people turn the perspective, how they frame students, etc. Although we did not conduct a study to look at such factors, we did find evidence of gendered noticing (unexpectedly as it wasn't a focus of the study) in our Kosko et al. (2022) paper in Computers & Education.

Hi Zohreh,

That's an interesting question. Currently we are focusing on teachers' professional noticing, which is a skillset involving being situationally aware to the point of attending to certain events and actions in-the-moment, interpreting what is occurring with professional knowledge, and deciding how to respond. This is also something we work on in our teacher ed program. We do see an association between how well someone engages in noticing and their physical actions when viewing 360 video. We do not have longitudinal data with eye-tracking though. 

The key association we see across our work with either field of view (FOV) or eye-tracking is that teachers who discuss students' mathematical reasoning will focus on students more (for longer durations) than teachers who discuss students' procedures (but not their conceptual reasoning). Those who talk about more generic pedagogy are often the messiest group in terms of how they go about viewing a 360 video (at least in terms of FOV - we are still analyzing eye-tracking data). 

With eye-tracking studies, there are typically very small samples because of the difficulty in collecting the data (and expense of the equipment). We have been using the Neo Pico Eye headsets which, although expensive, are MUCH cheaper than standard eye-tracking equipment. Our platform, Gaze XR, allows for multiple headsets to be used at once (our data collection sessions in April resulted in sessions of 4-7 at a time). Our next challenge is getting this logistically implemented into a methods course. 

Hi Kelly! Great question. Currently we are focusing on 360 video cases only (and not self-recorded videos). This is primarily because we are being conservative on server and storage capacity. Eventually we hope to support including recordings of one's own practice. With that said, I know of several folks that have used 360 video in this way. Kaltura, as an example, supports 360 video (single-perspective) and ThingLink does as well (the latter having better annotation tools for 360). Of course, neither of these support multi-perspective 360 nor do they report data on "where" someone looked. 

So to get at the heart of your last question - eventually we hope this tool can be used for both 360 video cases and 360 videos of one's own teaching. We would like the current lineup of tools to be combined with others to support such discussions. However, some of this will take time (primarily for capacity). We are definitely hoping to work on this in a large-scale version of this project though!