How can I tell if I have simplified my talk too much?

Question

My experience listening to talks (of all kinds - seminars, conference talks, even Masters/PhD defenses) is that they tend to be very difficult to understand. This happens even if they're within my own field or even subfield. Occasionally there are conceptually simple talks (projects that use machine learning are an example) that can be easily grasped, but most of the time the talks are too technical, or are the "blink and you will never understand the rest" kind. I estimate I rarely get more than 30% of what was said, even when I'm in the same field.

I certainly don't feel like I'm the only one who don't understand the talks, since I notice other listeners often get distracted by their laptops & mobile phones.

I personally think this is ridiculous, so when I give talks I try to simplify them to the point where someone with an undergraduate degree can understand it. For example one talk went:

• We're dealing with [technical term], which is analogous to [much-less technical term]
• This is what we want to measure, why we want to measure it, and what we expect to see
• After some very complicated mathematics (I literally used this phrase and skipped the details), this is what we get

I especially liked this talk since my family were able to grasp the gist of the work. However, the danger is that the talk might become too simplified. An actual expert would likely gain more by reading the paper during the talk instead of listening to it. Also, it seems some people think complicated language is the "industry standard".

How can I tell if a talk has been simplified too much?

Answer 1

I definitely hear where you are coming from.

In my modest opinion, a good conference talk is not necessarily the one that lets me understand the technical contribution (I can read the paper for that), but the one that lets me understand

• The importance and/or hardness of the problem studied
• What were the key new ideas
• Why they worked and when can I expect them to work

If I got this from the talk, I may be able to recall the result or the technique in the future when I have a need for it.

Of course I am not saying there should be no technical detail, the key idea may be the way a theorem was proved, but I believe any technical content should not be gratuitous.

Answer 2

It's important to keep in mind that different talks have different audiences and goals. How much you should simplify depends heavily on what kind of talk it is:

• For a job talk, part of the goal is to convince the audience that your research is advanced and hard. Oversimplifying is very dangerous in this context - frankly, it's better if parts of the audience does not understand your talk than if everybody understands it and some people think it's very easy. Simplify the problem and the conclusions (this is the part that everybody should get - why do you do this and why does it matter?), but don't simplify the actual technical content.
• For a public outreach talk, i.e., any talk directed at a general audience, it's important that everybody can follow the talk, while they won't care in the slightest if your methods are really complicated. Simplify as much as you have to for your audience to follow (how much this is clearly depends on who exactly the audience is - at a developer conference not much simplification may be necessary, when you are talking at a "Science for Kids" event you will need to break it down considerably).
• For a conference talk, opinions vary. I have come to the same conclusion as Pronte (that a conference talk is ultimately more an advertisement for reading the paper), but not everybody shares this sentiment. Decide what you want to do with your talk, and simplify enough to achieve this goal.

Another angle of this discussion is also which parts you should simplify for which audiences. For instance, I do a lot of empirical work nowadays (e.g., interview or survey studies). Oftentimes, the subject of my studies are technically deep topics (so the subject itself is rather complex), and the methods are, well, not exactly complex, but they are often lengthy to explain properly and somewhat intricate. How much I go into the details of either of these two aspects strongly depends on who I give the talk to. A scientific audience may appreciate me talking for 5 full minutes about how subjects were sampled and how results were coded etc, but an industrial audience will not care - "we did a study and interviewed X developers" is all the detail they need to hear. Conversely, the scientific audience is often not particularly interested in the technical complexities of the study subject, but an industrial audience almost certainly is.

Answer 3

I think that in almost all cases, people go to talks for insight not for details. If there is a reference/citation of a paper (or several) that provide the details, then the details needn't be in the talk itself. If you are at a conference with 300 listeners for your talk, then only half a dozen or so are likely to be interested in the pedantic details. But all are looking for the essential insights.

Some are also looking for ideas that they can expand themselves, of course. Students looking for important trends and interesting research possibilities. Even they don't need a complete, finished, product to make an informed decision. Insight triumphs.

But even in a formal lecture to students, which is expected to be quite detailed, you still expect the students to either have read the material before the lecture or to follow it up with readings and exercises afterwards.

The talk should say something about why the issues are important and about what the key, insight expanding, results were. It doesn't need to recapitulate the entire research process to do that.

If everyone understands it at some level it is a good thing. If a few want to follow up and go deeper it is also a good thing. Exhaust the audience with insight, not details.

As simple as possible, but not simpler.

Answer 4

The other answers are excellent, especially Buffy's.

One aspect that hasn't really been mentioned that I'd like to explore is the difference between how simplified something is, and how much detail is given. These are different concepts.

One of the common problems from many scientific presenters is not one of presenting ideas that are too difficult or complex, but rather an inability to reduce the level of detail. Don't think about "what would a reviewer ask on my paper", but rather "what is needed to tell the story that my audience cares about?". Unlike writing a journal article, you do not have to justify every single thing. As others have said, if people want more detail they will ask.

Answer 5

How can I tell if a talk has been simplified too much?

When everyone understands everything!

Ultimately, to engage everyone, a talk (typically) needs to be too complicated for everyone to follow (otherwise some will understand everything), whilst being simple enough for everyone to learn something.

You need to know your audience and you need to make a judgement call on how much each will learn. E.g., everyone should understand the problem, most should understand why it is important, many should understand what you achieve, some should understand how you achieve it, and perhaps several should be able to replicate your results.

You need to ensure experts don't get distracted (with email, their thoughts, ...), so make sure they are listening when you are telling them something new, perhaps even explain this upfront. E.g., I see the room is half full of experts in [technical term]. Parts of my talk will be uninteresting to them, but the other half need some background. So, I'll tell you when to disengage and I'll clap my hands when I need your attention.

Answer 6

Lots of good stuff in the responses here. As a (former) mathematician, I would add that a (colloquium | conference | job) talk can have a focus/understandability trajectory over its duration. Some of the best talks I have heard:

1. Started with an engaging but oversimplified example - of the question being posed and/or the answer, not necessarily just the latter.
2. Then they gave a framing of the relevance of the problem and its linkage to adjacent fields
3. Followed by an overview of the approach taken and insights gleaned
4. and finished with a few of the more gnarly but crucial details

Without trying to be overprecise about it, a (say) 40 min talk with 10 mins in each of these buckets should hit the sweet spot of everyone in the audience at one time or other, as well as then go (at least a bit) over their heads for the "knock their socks off" factor. You'll gradually lose everyone in the room except 2-3 people for the final 10 mins, but that's OK.

Don't be afraid to be understandable, as long as you also give a hint of depth. So few talks take the time to be accessible as well as grounded in the broader context that most audience members will be thankful you took the time rather than assume that your work is shallow!

Answer 7

I can definitely feel your point of view, I had the same concerns while I was preparing my bachelor and master thesis presentations.

My main approach was to keep simple the aspects that could be interesting to a more broad audience, like the explanation of the problem and qualities of the solution ( it is manageable, it is cheap, etc.).

Other aspects, like technologies involved, mathematical aspects of the solution and such, cannot really be simplified much but it should not matter because the information they convey is usually interesting only for the people able to understand it.

Also, if you have an experimental part (actual experiments, simulations etc), it is more useful to show them than to provide only cold numbers: show video of the experiments, create an animated version of the simulations and so on.

Answer 8

Your talk needs to be pitched at the demographic makeup of your audience. A talk needs a completely different based on who will be attending.

• A general-audience talk for the public must explain everything, try to use everyday-life analogies, assume no mathematical background, avoid equations and gloss over technical details. Try to use images but avoid graphs unless they're very simple. It usually must spend a lot of time motivating why the subject is interesting at all.

• A student talk can usually assume the audience cares about the subject and knows some basic maths. You still need to explain all the jargon, use analogies to very well known problems and gloss over difficult details. Graphs and images are good for this audience. You probably still want to motivate why your specific topic is interesting.

• A broader-topic expert talk (say, a selection of professionals from an entire department) will still need explanations of jargon from your specific subfield. They will need to be told the context of your contribution, but will probably understand the need for your subfield generally. Now is the time to add the difficult technical details, which will be understood only by a few people, so expect to spend time on them. Now you can add in the equations and difficult graphs.

• A subfield expert talk (say, for a conference on your subfield) should omit almost all the background or just include enough to establish which notation you're using. Emphasize things like why your method is superior to previously understood methods and be prepared to field hostile questions from competitors.

You should always know the demographics of your audience before you write the talk. If you are unsure, ask the organizers. A talk written for the wrong audience is going to bomb regardless of whether it would have been a good talk for a different audience.

Don't worry too much about how bad your colleagues' talks are. Academic communication is really difficult, and a lot of academics are just terrible at it.

Also don't worry about who's using their phone or laptop. Allocation of attention is a personal problem, and you're not their mother. (If you are their mother, simply end the talk early and send them to bed.)