Graduate Research Presentation Expectations

I have sat through enough lab meetings and qualifying exams now to notice a pattern, and I want to write it down before the next round of presentations, so that everyone in the group can read it beforehand instead of hearing it, again, after the fact.

The pattern is this: students spend their effort in the wrong place. They spend hours choosing a color palette, hunting for a clean icon set, animating a transition, and then they spend twenty minutes, the night before, checking whether the claim on slide fourteen is actually true. It should be the other way around. A slide deck is not the product of your research. It is a vehicle for a story, and the only job of the vehicle is to get the story across clearly. Nobody remembers a beautiful slide with a weak claim on it. Everybody remembers a shaky slide that clearly explained why the problem mattered.

So let me walk through what I actually look for, roughly in the order I notice it.

Content before appearance

The first thing I ask, silently, about every slide is whether I believe it. Is the statement accurate. Does the claim have evidence behind it, a citation, a number, something. Does the figure actually support the point being made, or is it just sitting there because a slide “needs” a picture. If a slide cannot pass that test, no amount of polish saves it. I would much rather see a plain, slightly ugly slide that says something true and specific than a gorgeous one that says something vague. Get the content right first. The formatting is a half-day of work at the end, not the main event.

Make it yours

Please do not reuse my slide layouts, my color choices, my diagram structures. I do not mean this as a matter of pride. I mean it because copying a layout without understanding why it exists is a tell — it tells me you have not yet decided how you want to organize your own thinking. A good exercise, before you commit to a layout, is to ask yourself why you chose it. If the honest answer is “because that’s what the professor’s slides looked like,” redo it. Your presentation should look like the inside of your head, not like a template. This is also, frankly, how you develop a style of your own over the next few years — nobody arrives with one, it comes from making these choices badly a few times first.

One slide, one job

Every slide should be answering exactly one question. Why does this problem matter. What exactly is the problem. What is still unsolved. What do we already know. What theory grounds this. What are we investigating. What did we do. What did we find. What does it mean. What comes next. Each of those is its own slide, its own job.

The mistake I see most often is a single slide trying to carry motivation, gap, theoretical framing, and contribution all at once, usually because it felt inefficient to “waste” a slide on just one idea. It is not a waste. A slide that tries to do six things does none of them well, and the audience leaves not quite sure what any of the six were.

The story has to move forward

A research talk should read like a chain, each link pulling you to the next: motivation, into the problem statement, into the gap, into what the literature already tells us, into the theory grounding the work, into the research question, into the method, into the results, into the discussion, into what comes next. At any point in the talk, your audience should be able to answer three things without effort — where are we right now, why are we here, and what is coming next. If they cannot, the chain has a broken link somewhere before that slide, not on it.

Motivation is about the problem, not the technology

A motivation slide exists to answer one question: why should anyone in this room care. That is a very different question from “what can this technology do.” I keep seeing motivation slides that open like this:

VR can create fully immersive learning environments. LLMs can generate adaptive dialogue in real time.

That is background, not motivation, and it is a common enough trap that it is worth naming directly. Real motivation follows the shape of a genuine problem:

Students need effective, ongoing assessment of oral performance. Current approaches — human raters, scheduled check-ins — do not scale. Existing digital tools that do scale tend to break immersion the moment they interrupt the learner. Therefore, a new approach is needed.

Notice that a technology never has to appear in that second version at all. It shows up later, as your answer, not as your opening line — something important is hard, current approaches fall short of it in some specific way, that shortfall has a real consequence, and only then does the solution enter.

State the problem plainly

Once you’ve earned the audience’s attention, tell them exactly what the problem is, in a sentence or two, no more. Something like: current immersive learning environments lack a scalable way to assess oral performance continuously without breaking the learner’s sense of immersion. That is a problem statement. It is not the place to start reviewing literature or describing your method — those come next, and folding them in early just muddies the one thing this slide needs to do.

A literature review synthesizes, it does not summarize

This is probably the single most common failure I see, so let me be blunt about it. This is not a literature review, it’s a list:

Paper A found that VR improves engagement. Paper B found that LLMs support adaptive interaction. Paper C found that pedagogical agents improve social presence.

A literature review should instead sound like this:

Research shows VR improves engagement, LLMs support adaptive interaction, and pedagogical agents strengthen social presence. However, these strands are rarely integrated, and adaptive oral assessment in particular remains largely unexplored.

Same three papers, but the second version has done work — it has drawn out what the field agrees on and pointed at where it stops agreeing. The whole point of the exercise is to build a bridge from what we collectively know to the gap you are about to name. If your review does not end somewhere near that gap, it has not done its job yet.

Say the gap out loud

Do not assume the audience will spot the gap on their own, even if it feels obvious to you after months of reading. Say it, in this shape:

Existing research has addressed X. However, Y remains unaddressed. Therefore, we investigate Z.

By the time that slide is done, everyone in the room should understand, in their own words, why your project needs to exist.

Figures should earn their place

A good figure — an architecture diagram, a process flow, a framework, a results plot, a comparison table — carries information the audience could not get as fast from text alone. A decorative AI-generated illustration, a generic stock photo of a robot or a brain, an image that just re-draws what the bullet point beside it already said — none of that earns its place. Here is the test I actually use: if I removed this image, would the audience lose something. If the honest answer is no, the image was decoration, and decoration is exactly what we agreed not to spend time on.

Build your own frameworks

This one I feel strongly about. Do not hand a conceptual framework over to AI and put whatever comes back on a slide. Every box in a framework needs a definition you can state from memory. Every category needs a reason it belongs. Every arrow needs to represent a relationship you can defend, ideally one grounded in theory or evidence, not one that merely looked plausible. If there is a single element in your framework you cannot explain unprompted, take it out. A framework is supposed to be a picture of your understanding. If it is instead a picture of what a language model guessed you meant, it is worth less than an empty slide.

Stay the author, even when AI drafts for you

AI is useful, I use it too, and I am not asking anyone to avoid it. But you remain the author. That means the accuracy, the logic, the theoretical consistency, and the interpretation are still yours to answer for, in the room, when I ask about them. Before anything AI-generated goes on a slide, run it through a short checklist: is it correct, is it actually supported by the literature, does it fit the story you are telling, and — the one that matters most — can you explain it yourself, right now, without looking at the slide. If any answer is no, revise it before it goes in the deck. “The AI wrote it” is not a defense I will accept, because it was never really an answer to the question I asked.

Say each thing once

Every slide should add something new. Watch for the quiet ways redundancy creeps in: an image followed by a bullet that just narrates the image, motivation content resurfacing in the problem statement, the same limitation restated three slides in a row, or four bullets that are really one idea wearing different clothes. Each slide is a chance to move the story forward — spend it on something new, or don’t spend it.

Design in service of clarity

Good design, in a research talk, is almost invisible — consistent formatting, a clear visual hierarchy, restrained text, enough white space to breathe, fonts you can read from the back of the room, content that stays focused. What you want to avoid is the opposite instinct: too many colors, too much animation, decorative backgrounds, layouts crowded past the point of legibility, three fonts where one would do. When in doubt, simplify. Simple has never once cost you a point in a talk I’ve sat through; cluttered has cost plenty.

Know your one takeaway

Before you consider a slide finished, ask yourself what you want the audience to remember if they remember only one thing from it. If you cannot answer that in a single sentence, the slide is probably still trying to do too much, or hasn’t found its point yet. A good slide has one main message, and everything on it is there to support that message.

Rehearse, and prepare to be questioned

Slides are half the job. The other half is what happens once you start talking, and this is the part people rehearse least.

Practice out loud, not silently in your head — reading silently hides exactly the timing problems and awkward phrasing that a spoken run-through exposes. Time yourself against the actual slot; if you have fifteen minutes, rehearse until you land there consistently, not once. And know the sentence that carries you from each slide to the next — if you cannot say why slide fourteen leads to slide fifteen, that seam in the story isn’t ready yet, no matter how clean either slide looks on its own.

Then prepare for the room to push back. Before you present, sit down and write out the five hardest questions a genuinely critical committee member could ask you: why this method and not the obvious alternative, how this differs from the closest prior work, what the real limitation of your results is, how you know your measurement is actually valid, what breaks if your key assumption turns out to be wrong. If you cannot answer one of these about your own work, that is a sign to go fix your understanding, or the slide behind it — not to hope nobody asks. I have watched a confident, well-rehearsed delivery survive a bad slide. I have never watched it survive an unanswerable question about the student’s own claim.

What weak slides are usually telling you

In my experience, a run of weak slides is rarely a slide problem. It is usually the visible symptom of something further upstream — a gap in understanding, a literature review that never quite converged, a research question that is still fuzzy, reasoning with a hole in it somewhere. So when a talk isn’t working, I would rather you fix the understanding first: the problem, the literature, the theory, your own contribution. The slide design comes after that, almost as an afterthought, because by then there is finally something clear to design around.

What I am actually optimizing for

I am not grading you on how professional the deck looks. I am watching for whether you understand the problem, whether you understand the literature around it, whether you understand your own methodology, whether you understand why your contribution matters, and whether you can walk someone else through all of that clearly and in order. A talk that does those five things with plain, honest slides will always outscore one with beautiful graphics and a borrowed structure sitting on top of thin thinking. Original thinking, argued clearly, in your own words — that is worth more than any template I could hand you.