Why most technical talks are bad

In one sentence: The typical talk is inefficient because the presenter fails to motivate their subject matter or present their talking points as a relatable story.

To clarify, I’m talking primarily about conference talks, symposia, lab meetings, and other presentations given by researchers in technical fields.

I’m not the only one who feels this way

• I used to think it was my problem; that I was uniquely incapable of absorbing information from technical talks. And to some extent, that was probably true: I was a novice mathematician/scientist.
• I no longer think it’s totally my fault, though. I think most talks just aren’t very easy to digest.

Fifty percent of people
understand fifty percent
of fifty percent of the talks in a conference.

–Moshe Vardi¹

• If the brilliant Moshe Vardi feels this way, then I seem to be in good company.
• I suspect scientists don’t talk about this more often because
  1. It’s difficult to deliver a good talk. So, you know, “glass houses”.
  2. If you admit that you didn’t understand a talk then you risk “looking dumb,” which is kind of scary to do as a researcher in a STEM field.

Common weaknesses

• lack of motivation
  • Why should anyone care about the information you’re presenting?
    • If the audience doesn’t care about your topic, then they won’t pay attention.
    • Most of the time, you’ll have to do some work in order to help your audience care about your topic. A very small number of people work in exactly the same area as you; the rest will need some help to see why your work matters.
  • Motivation sharpens our communication. Abraham Lincoln quote. A variant: give me 20 minutes to present, and I will spend about half of it making sure the audience pays attention.

Give me six hours to chop down a tree
and I will spend the first four sharpening the axe.

–Abraham Lincoln

• Lack of plot
  • humans absorb information best in the form of stories.
  • From before the days of Homer, humans have passed wisdom and values to each other via recited stories.
  • A story presents information as a network of related observations. It usually has a sequential structure.
  • The alternative is to present a set of disconnected facts. Humans have difficulty remembering or making sense of information in this form.
    • See “chunking” from cognitive psychology.
    • Or think of it from a probabilistic viewpoint, where factual propositions are random variables. If we treat them as totally independent, then information about one proposition gives no information about any other proposition. But if we introduce dependencies between them, then we can make inferences and recover any forgotten facts. The whole set of facts is stored more robustly.
  • It’s possible (but not necessarily true) that “story-izing” your talk may reduce its technical rigor. However, rigor is arguably not the point of a talk. The point of a talk is to convey high-level ideas to a diverse audience with limited time and attention. If people want rigor, they should read the damn paper.
• Bad slides
  • Too much content on each slide
    • The outdated idea that slides should be a form of “documentation”
      • Cite the bad slides that (arguably) caused the Columbia shuttle disaster
    • Too much text
    • Busy figures

An extremely useful exercise: one-sentence summaries

• Can you capture the thrust of your talk in one sentence?
• The point of the exercise is that it forces you to (i) identify the most important pieces of information, and (ii) embed them in a simple narrative structure: e.g., subject, predicate, object.
• The resulting sentence is lossy, or perhaps even technically incorrect. But it conveys the bulk of your message in a highly transferable form and enables the listener to begin thinking about the subject.
• You can do this exercise at any level of detail. You can distill your entire talk, each slide, each bullet point, each figure.

Extensions to other scholarly media

• Lectures
  • I’ve had similar difficulties absorbing information from lectures.
  • Over time, I’ve come to believe that it’s largely a matter of the lecturer’s ability, and not just my own comprehension abilities.
    • This comes from taking dozens of courses in math, physics, computer science, and genetics.
    • It also comes from my limited teaching experience – it isn’t easy to prepare or deliver a lecture.
  • Ultimately I think similar insights apply to lectures, especially in technical subjects. A lecture is not the place to focus on technical details; students will get those from (a) reading and (b) doing their assignments. The lecturer’s time is better spent motivating the subject, identifying the most important ideas, and showing how they relate to each other.
  • Alternatively, I’ve had very positive experiences in “flipped classrooms” where in-class time is spent working on problems and out-of-class time is spent watching recorded lectures, reading, and doing homework. I don’t think this negates my hypothesis, though. It just means that the lecture happens in a different environment.
• Papers
  • I suspect that many technical papers are hard to digest because they’re poorly written.
  • However, the goal of a paper is to provide full technical detail. So my criticisms of papers are different from those of talks or lectures.
  • This deserves an entire post of its own.

Caveats and speculation

• It may be that the qualities that make a good researcher are exactly the wrong qualities for making a good communicator.
  • An analogy from data structures: we can store the same information in two different data structures. But we prefer one data structure over another based on the task at hand. The very qualities that make a data structure optimal for one task can make it extremely inefficient for another task.
  • There are of course prominent counterexamples like Richard Feynman. So I’m not really convinced of this. It’s more likely that communication is a difficult skill that requires conscious effort.

\( \blacksquare\)