clear_talk_tips.md

Judy's CogSci-oriented supplement to Kayvon's Clear Talk Tips

I suggest first checking out Kayvon's Clear Talk Tips (KCTT). This document is meant to provide some extra tips that may be especially useful when giving talks to a Cognitive Science (i.e., Psychology, AI, Neuroscience) audience, that aren't explicitly addressed in KCTT. Another nice slide deck on this subject is Ranjit Jhala's "On Preparing Talks".

Audience

Think about your audience. Is this for lab meeting? For an internal seminar? For an external seminar at another institution? For a major conference? A job talk? Each of these talk venues vary slightly in terms of what your audience is aiming to get out of it, and what you are trying to accomplish. By and large, this document will focus on tips for formal research talks (i.e., internal/external seminars, conference talks, and job talks). However, I believe that it's generally a good idea to take all, even informal, talk opportunities seriously: one, because it is good practice for giving formal talks; two, because a good talk often fosters higher quality discussion; and three, because your reputation as a communicator is one of your most precious assets as an academic and there is a chance someone in the audience will be in a position to recommend you for promotion one day. ("Every talk is a job talk.")

Preparation

I understand the third point above may be daunting. That is why we prepare. Set aside ample time to prepare your slides, what you plan to say, and how you will answer common questions that arise. Before your FYP, before major conference talks, and before job talks, plan to schedule one or more practice talks with the lab. Bring a physical notepad to these practice talks so you can jot down notes and questions that people ask while your slides are still up. If possible, also ask a labmate to take notes so you have a backup in case you miss anything. Be gracious when receiving constructive criticism: the discourse norms in our field are such that the burden is primarily on the speaker to hold the attention of the listener -- if the listener did not understand or follow something you said, that means there is an opportunity to communicate that point more clearly next time!

Full practice talks in the lab will typically follow this procedure:

1. full run-through
2. science questions
3. high-level feedback
4. detailed slide-by-slide feedback

Talk structure

A talk is typically organized in a way that mirrors the structure of an academic paper: Introduction, Methods, Results, Discussion. However I do not suggest that you explicitly label the sections of your talk in this way! Most members of your audience will already be using this schema to understand your talk, so what you should focus on is communicating the unique content of your talk in each section. This means using informative slide titles like "Visual complexity of objects increases visual search time" rather than generic slide titles like "Outline" or "Results."

Introduction

An Intro is shaped like a funnel. You begin with a big picture question, and you end by motivating your specific study.

Begin with a compelling & concrete example that shows rather than tells the audience about a phenomenon that is pervasive and relatable. Unlike computer graphics (focus of KCTT), CogSci is a scientific and not an engineering field, so we do not typically present a "problem to solve," but rather a "puzzle to explain."

Your goal in the Intro is to pose a worthwhile question and "situate" it theoretically and empirically so that it becomes obvious why the research approach you took was the sensible next step.

Situating your question means connecting to either/both empirical gaps in knowledge and theoretical alternatives at stake. It means communicating to your audience:

• how to think about the question you are interested in + why it matters, and
• enough about what we currently know to communicate why you're doing the work you're doing.

Including a slide with a laundry list of Author1, Author2 (Year) citations and some keywords you associate with each citation is NOT an effective way to situate your question. When you are referencing key prior work, use the slide to communicate what role that prior work plays in your argument. This is not the place for a literature review. Citations to relevant work should be placed in the bottom corner of the slide.

Outline

An outline slide can be a useful organizing tool for your audience to help them appropriately "chunk" each section of your talk. Use it to pose a series of related questions that will correspond to the major results you will show them. Formulate these in a way that will be meaningful even without having heard your talk. Do not write "Outline" or Research Questions" at the top of the slide.

Methods

Motivate your approach. In other words, don't just tell the audience what you did, but why you did what you did.

Spend time developing graphics that illustrate the key components of your methodology. When you deploy them in slide form, spend time animating each build of the graphic so that your audience only has to focus on one chunk of information at a time — the chunk of information you are currently telling them about.

Resist the temptation to include a generic bullet-point slide with "N=64," "2x2 factorial design." These aren't the details that matter enough to devote a whole slide to them. The bottom corner is the place to tuck sample-size information.

Results

Pick only the results that matter the most and explain them well. It may be tempting to tell them about all of the complicated analyses you've done. Don't do this. :)

Motivate each result by situating it in the broader narrative you are telling. What is the question that your result is answering? Remember to state your hypotheses, and translate them into specific predictions that can be evaluated by inspecting your plot. In fact, tell them what your plot would look like if each candidate hypothesis were true. That will help contextualize and lend greater impact to the result you actually found (b/c it could have been otherwise!)

When you show a key results slide, slow down. Do not assume that your audience already knows how to read your graph ("As you can see, ...").

• Start with a blank plot and orient your audience to what your axes mean. Tell your audience what a single dot or line or curve represents.
• If you can, animate each portion of the plot separately that corresponds to a specific observation you want to make.
• Make sure any text labels are large enough for the person in the back of the room to be able to read it. - After you tell your audience what is shown on the plot, tell them what it means with respect to your hypotheses.
• It may also be a good idea to include a "take-home" message in large typeface somewhere on the results slide after animating in all of your data. This helps to reinforce what you are saying and proactively helps the people who zoned out in the middle of your slide to still come away with the main point.

If you have an outline slide with specific questions, there should be a one-to-one correspondence between the specific questions you pose on your outline slide and your major results. After you state your major results, revisit that outline slide and briefly summarize how your result addresses that question.

Summary

If you've made good use of your outline and results slides, you will have effectively been building up a summary slide throughout your entire talk. Recap what you've shown them / argued.

Discussion

These sections of talks are rarely done well. The point of it is to connect your work back to the broader question you posed at the beginning of the talk. Beyond just the specific findings you shared, what are the implications of the perspective / approach you've taken? What are the larger theoretical stakes that your work speaks to?

Slide Design

• Use a consistent and semantically sensible color scheme throughout. If you are going to use a specific color to correspond to a condition, use that color consistently in every slide, graphic, and plot in your talk.

• Choose a typeface that is easy to read and use font weights and styles consistently throughout your talk. If you are going to make something bold, think about why you are doing that and what it means, then apply that same principle throughout all of your slides.

• The key here is consistency. Use the same phrase to refer to the same things, rather than cycling through several quasi-synonyms.

• Use animation judiciously and for a specific purpose -- namely, to direct your audience's gaze to a specific data point or phrase by making it appear just in time. Resist the temptation to include gratuitous animation, just because you think it looks cool -- these usually end up being distractions.

• Avoid unnecessary acronyms. If you must use acronyms, explain what they are the first time you use them. Use informative labels that reduce the working memory demands of following your talk, so your audience can focus on asking incisive questions and making useful suggestions.

Managing questions

This is an advanced talk skill. TODO.