This interesting insight about the organization of content in a science talk comes from Adam Waksman, a PhD student in Computer Science at Columbia University:
“One challenge that is always present in computer science talks is the need to constantly remind the audience what is happening. Computer scientists expect to be told what they are going to hear before they hear it.
While I do not necessarily think that results need to be a mystery, I also think the constant re-telling gets tedious. For example, the audience might be told something on an initial overview slide, then again during motivation, then reminded of it on an outline slide and then told it in detail later in the talk. For me, that is too much.
One example of where that is an issue for me in my Berlin talk [presented at CCS 2013 – http://www.sigsac.org/ccs/CCS2013 in November] is the issue of multiple scores. At some point in the talk, the audience needs to understand that there are two different scores being computed and why they are both relevant. Saying it too early gets redundant and derails the story. Saying it too late makes it seem like it came out of the blue and also somewhat derails the story. I have moved the relevant slide a couple of times in recent revisions, but I am not sure I am happy at the moment with the placement.
In my experience, these are often tricky cases — where there is a detail to the story that is important because of scientific reasons but not inherently obvious. One cannot withhold these details in a long talk, but when introduced at the wrong time they make the story seem less simple and perfect, which is often what a computer science audience wants to hear. This is something I expect to improve upon as I do more practice talks.”
Thank you Adam!