> - have stats about the meeting (who was actively participating, maybe via gaze tracking or seeing whether the tab is in view). We're they speaking aggressively, friendly, etc.

This is one's dangerous. I'm actually finding it hard to think of a use case for this where it isn't abusive and oppressive.

Here's one: If you see high disengagement, you know you have to improve as an organizer or that the person should not have been invited.

The context for my point is that we already have authoritarian countries (i.e China, Russia) already using the internet for mass surveillance (and selling the technology to other countries), we have Trump (who has to be considered at least a quasi-authoritarian at this point) purging civil servant who are insufficiently loyal to him.

And these are not power that will have intrinsic knowledge or care about things we may understand intuitively: the error rate of a prediction model, false positives/negatives in classification, the difficulty of producing a reliable model of human emotion, how easy it is to build bias into a ML model, etc.

Agreed, the feature could be used for bad purposes. Does that mean we shouldn't build it? That's a tough question.

You can also ask participants for their feedback.

Yes, but the point of all my suggestions is to make things automatic.

Besides the feedback to the organizer part, there could be all kinds of stats about the meeting, compare to the stats of e.g. a sports game. If you have a pool of sales reps, and you want to help them improve, you can compare the stats: on average you ask 2 questions per call, while the average rep asks 10, etc..

Don't get me wrong, there are a lot of nefarious use cases, but there are definitely positive ones as well.

I can see the use cases, but the nefarious ones kind of obscure this prospect.

Maybe the negatives could be mitigated by doing the automated analysis on the client and asking the users if they want to send the feedback, similarly to how crash reports for software are often done.

Yeah, that might be a good mitigation! It will never be perfect, but that's life ;)

You are absolutely right. I don't want my teachers to know that I am not actively attending the class and doing something else.

Also, even if this feature gets implemented, someone will defeat it easily. ( Example: Teachers ask you to keep your camera feed on during the class? OBS + Virtual Camera )

Many of those things already exist and are popular in virtual worlds and in social games. In software in general there's a lack of knowledge of what's already been tried.

This looks like a great start though.

Definitely agree! League of Legends and TF2/CS:GO stand out as games that have a huge community that builds a lot of analytics tools for in-game activity data. Most times, the community tools are way more feature rich and comprehensive than game dev tools. Definitely a lot to learn from games

I've seen some demos of VR based virtual worlds and it's really really cool. Downside is not being able to see facial expressions, and no one I ever have meetings with has a VR headset.

I already commented below, but that's pretty much what we are aiming at with https://laptopsinspace.de It's not VR (on purpose), but instead it has facial expressions and some gaming feel to it

The biggest challenge is background noise cancellation... it can severely degrade the quality of a group call. Most of the others can already be done... and in this case, embed of q&a/polls, document.. can be defaults when setting up a new call.

We run a pretty strict "mute when not speaking" policy and it's working very well. I think giving users automated feedback about the quality of their sound is a much simpler problem to solve. Not as sexy as AI powered noise cancellation, but I think it would be much more effective in the short term.