(A shorter version on Facebook, here. There are some good points in the comments.)
Why are there so many in-person-first EA/academic presentations, but so few online-first presentations or videos?
Presentations are a big deal in academia, but from what I can tell, they almost always seem to be mainly done in person. A significant number of presentations are repeated multiple times to multiple groups of people, in person.
Sometimes these presentations are recorded and posted online, but they typically are optimized more for in-person audiences, with online audiences as a bit of an afterthought.
EAs sometimes spend a lot of time on presentations at EAG. These get posted online, but I don't think these videos get much attention. I basically don’t ever remember hearing someone discuss an EAG video they watched in conversation, or referencing it in a blog post.
If the online audience were the main target, I think presenters could just make videos directly and upload those. That way they'd have much more flexibility (can do multiple takes, add effects, and polish).
Online webinars are kind of awkward. First, streaming has a lot of disadvantages compared to recorded video. Worse sound quality, more equipment needs, having one single take. Second, these webinars are often barely interactive. If they're not interactive, there's really no first-principles need to have the first 40 minutes made up on the fly and presented at 1x speed, as opposed to being sent to people ahead of time.
Or take tech presentations. Several companies just did in-person keynote addresses at CES. In comparison, Apple has gotten really good at recording (much of) their work ahead of time, and presenting that in highly polished videos[1]. Maybe it’s a flex to present your work in real-time in one take and not mess up, or to physically be in the audience when that one take happens.
There seems to be some dramatic pull that gets people to converge on presenting and watching things live, especially when they are in person. It's hard to explain this pull with any simple cost-benefit analysis.
Some quick hypotheses (in part, from comments):
- The EA Funders and leaders don't seem to watch videos much, so it wouldn't help many orgs.
- It's a lot of work to do this well. The setup cost is significant.
- The standard for "well-made Youtube videos" is absurdly high, and people are worried that content not at that level will look bad.
- Video is associated with stupider people (in the eyes of many)
- The required skills are pretty different than writing. Lots of strong EA researchers are not great presenters.
- A lot of people in our community don't like watching video or listening to audio. They often really prefer text.
- It's assumed that if you make video, it's not for EAs, but for "the generic Youtube audience".
(Of course, this parallels the question of why flipped classrooms haven't done incredibly well. Having teachers give long monologues on repeat seems stupid prima facie, but it still seems to work.)
[1] I remember their segments during Covid mostly seemed prerecorded. Even the current ones seem highly prerecorded to me, but I’m not very sure where the boundaries are. The fact that I can’t tell, and there didn’t seem to be a lot of controversy here, is telling.
Because text is better than video?
This raises the question: why do academics bother with in-person presentations, rather than just sending each other (even more) papers? The answer seems to be some mix of:
Posting online videos loses all the value of the latter two points, and also a fair chunk of the first, insofar as you're unlikely to have any professional obligation to be in the general space of watching the video. (Though some might find watching the video to be less effort than reading a text version?)
In the context of scientific academia, the reason we do webinars and conference presentations is that they're way easier to put together than a slick youtube video. Our target audience is not the general public, it's usually other scientists in the same narrow field as us. These are the people who will cite our papers, use our methods, collaborate with us, and potentially score us a job opportunity in the future (or sign up to us as a postdoc/student). Academic conferences are mostly excuses to network with these aims in mind.
Putting a ton of effort into a video doesn't make sense in this context. The actual target audience can be as few as a couple dozen people, and if they want to know more they can just read the accompanying journal article when it comes out, which is where it makes more sense to put all our effort in order to get past peer review.
I don't know how much of this is applicable to EA conferences as I have not been to any. Certainly it makes sense if presenting work that it highly technical and of little interest to the general public.
Thanks for the comment!
I'd note that I think there's a very wide range between "webinar" and "highly polished youtube clip".
If you don't do it in real-time, there can be less pressure to get every part right.
Even 5-30 minutes of editing could go a long way. There are tools out there that use AI to do some of the cleanup for you automatically (I'm using Descript). And it's easier to get high-quality audio and video for recording instead of streaming. (Live zoom/streaming setups typically have more compression than processed video)
While I respect this perspective, I think it's likely post-hoc.
When it comes to spending hundreds of hours on a paper, especially some research which requires signifiant lab costs, what is a few hundred towards paying a freelance video editor? Would this lead to significant increase in citation factor - quite possibly - there's been some studies showing papers are more likely to be cited after being covered in a major newspaper.
Which is lovely for the type of paper that could get covered in a major newspaper, but the majority of papers are not. If you look at the typical physics paper, you're gonna have a hard time convincing someone to drop hundreds of dollars promoting "Degeneracy between even- and odd-parity superconductivity in the quasi-one-dimensional Hubbard model and implications for SR2RuO4undefined". Also, a freelance video editor is not going to understand how superconductivity relates to the hubbard model. What are they going to do, put flashy animations on top of a few intro slides?
A large part of science is about hyperspecialists developing knowledge very slowly, in collaboration with other hyperspecialists. Eventually a lot of cool and useful stuff comes out of it, but most of the steps along the way are mainly of interests to other specialists.
Having said that, I'm not sure how much of this applies to EA, apart from the highly technical stuff. A lot of it is intended to guide a general audience, so i see the appeal in aiming towards said audience.
l think a lot of it comes down to goals and function.
In my experience a significant portion and function of academic conferences are to get feedback on works in progress: speakers present the project they're working on, get feedback from a specialized and expert audience, and the audience get to keep up with the current state of things. It's a part of a collaborative process, and I think that this does actually benefit from there being no recording or written published artifact, because it allows for an unpolished and much rougher version of the work to exist without there being much risk of errors in it being publicized or perpetuated through citation or sharing - kind of like a lower pressure and less risky preprint. This obviously isn't the case for all conferences or for all presentations, but it is something I've personally majorly benefitted from in my research. I've also attended some conferences where speakers were open about being more able to safely speak about politically contentious issues because they weren't worried about there being an easily accessible document for the government to use against them, but that's obviously specific to a context where academic freedom is limited by political pressure and threat of persecution.
Also, for academia, in person presentations like job talks function as a kind of dry run for hiring decisions. Academics teach, and presenting or teaching in an in person context is a different dynamic and skill set than presenting something in a prerecorded format.
In person conferences also serve major functions for both networking and building a community of researchers that have the resources and ability to interact with each other in ways that are less formal than the back and forth of academic publishing.
There's also a lot of institutional factors that come into play. Alongside the networking aspect, academics use conferences for CV building. Conferences are usually peer reviewed and having your abstract selected for presentation signals a level of quality and rigor that isn't necessarily present for producing and posting a video.