You get absolutely zero feedback and discussion generated from a talk. Maybe a few questions, but no actual discussion. If you're really, really, really lucky, someone might track you down or e-mail you later. For the most part, though, you don't interact with people during a talk.
In contrast, during a poster session, you talk yourself hoarse over three to twelve hours. You wind up in intense arguments and deep discussions. I've never seen that happen during talks.
Personally, I find a 12.5 minute talk is a lot more limiting than a poster, where you'll be talking to people about your work for several hours. Figures are the main focus of any scientific presentation, and a poster gives a great format to walk people through the concepts using figures.
That having been said, at the tech conferences I've been to, posters have always been an afterthought. It's a real shame.
I used to hate poster sessions until I realized this. Something poster sessions provide, that other formats like podium talks really fail at, is pretty fine-grained control over the level of engagement. You can spend five seconds glancing at a poster, or a minute or two skimming the whole thing; you can spend a few minutes listening in while the presenter talks a bit about the work to someone else; you can skim and then ask one or two clarifying questions; or you can engage the presenter for as long as you're both interested in discussing.
Posters are usually used in high-volume settings with very light review. "Everybody" attending the conference gets to present a poster if they want to.
> a video... an interactive computer screen
Now you need 500 monitors at the conference venue, power cables, etc. The costs add up big time; hotels make a killing on this type of thing.
> a talk
You want to limit the number of talks so that the accepted talks are actually well-attended! So, not everyone gets to give a talk. Also, most people who give talks also present a poster because maybe there was a parallel session that stole some of your audience (who are interested enough to find you at the poster session).
At least in CS, no one is choosing a poster over a talk! talks are reserved for the work that passed the highest bar of peer review this year. (That said, I always recommend people present posters for some of the reasons described below. Even though talks are "higher status", IMO posters are often more effective once people are engaged.)
> a demonstration
These are sometimes rolled into poster sessions, but require more space and planning on the part of the organizers. Space at conferences is often hard to find.
> an interactive exhibit, a diorama, a play
This is kinda funny coming from a CS/math background. I guess in some fields this could make sense but would have the same drawbacks as videos or talks, respectively.
A very efficient way to do this is the poster session, where you have enough of your work on a poster that you can get the general ideas across in a few min (or even less). Some people do bring laptops with videos and things. My experience though is that spoken words and a few graphs are sufficient to get the story across and spark deeper conversations.
- they parallelise well: you can fit a lot of posters in a room, and have a person standing beside each one talking concurrently
- it's easy to quickly glance at each poster in a room and move on if they're not of interest to you; it is harder to leave a talk that turns out to be uninteresting, and you can't be physically present for more than one talk at once
- they allow the presentation of a contribution that is currently too small to form the basis of a complete talk, which lets people get early feedback and advice (and gives students at an early stage in their PhD a chance to present something)
In some venues people have interactive demos beside their poster, running on a laptop/iPad/monitor.
At some conferences there is a session where a short (e.g. 30s) video for each poster is played as an advertisement.
Posters can give rise to deeper conversations, but they can also be more or less completely ignored (whereas a talk would have at least a few people in the audience).
