By Brittne Kakulla, PhD | AARP | Member, Insights Association IDEA Council
The Quiet Part, Said Out Loud
There I was in a meeting, and the quiet part got said out loud. We were working through a research plan when one of the stakeholders cut in. I am paraphrasing, but it was close to this: "We don't have time for a six-month study and then wait for a 30-page report. That way of doing research is not going to work here."
I was a little shocked, mostly because part of me agreed with him. I have never loved writing 30-page reports, and I am not someone who likes to live in the research weeds. But there it was, out loud: research framed as the thing that slows everyone down.
But I hate to say it, he was kind of right. A study can be perfectly designed and perfectly rigorous and still have no impact, because by the time the report is finished, the decision has already been made without it. Rigor that arrives too late is rigor nobody uses. Researchers today are being asked to be careful and fast at the same time, and that pressure is not going away.
Seeing Before Reading
I am a social psychologist by training, and that meant an entire course in cognitive psychology. One lesson has stayed with me: the brain handles images and words differently. Dual coding theory describes two separate channels in the mind, one for language and one for images. Studies have also found that people remember pictures better than words, which psychologists call the picture superiority effect. What people see can matter as much as what they read or hear.
That makes visuals the first place bias can come in. A chart with an axis that does not start at zero, or a comparison against the wrong baseline, can make a small difference look large or hide a real one. Because people remember the picture, they remember the distortion too, even when every number on the chart is correct.
The ability to create and redesign visuals and reports quickly matters. That stakeholder did not want to wait six months or read a 30-page report. In this AI day and age, he does not have to. AI can help with that speed problem. It can take a heavy study and turn it into a faster, lighter, visual top-line read that people actually use.
But speed invites a shortcut, and that shortcut is the second place bias comes in. When you compress months of work into a handful of slides, the easiest things to drop are the caveats and qualifiers, the “this was only true for one group” and the “we should be careful reading too much into this.” The headline survives because it is simple. The details that complicate it get smoothed away.
The researcher in an AI world has to understand the need for speed. The job is to go fast without cutting the caveats that keep the work honest, and to make sure our visuals are not biased in the push to keep everything top-line. That means recognizing our own pull toward heuristic shortcuts and using AI tools as sparring partners that help us fight that and pivot quickly for different audiences.
On LinkedIn, I have been exploring prompting patterns through a researcher's lens. I bring that series here each quarter, pairing one AI tool with prompt patterns that help researchers get value from AI while understanding its limits. This is the third in the series.
Tool Callout: Gamma
What it does: Gamma is an AI tool that turns raw material into a finished presentation. You give it a topic, an outline, notes, a document, or a URL, and it drafts the slides, chooses the images, builds the layouts, and applies a consistent design. It comes with ready-made templates and themes, and you can create your own with your colors, fonts, and logo. You can even pull those brand elements from an existing PowerPoint or Google Slides file. Set your custom theme as the default, and every new deck starts formatted the way you want. When the deck is ready, you can export it to PowerPoint, PDF, or Google Slides. The free plan is enough to get a feel for the tool, but custom brand colors, logos, and fonts require a paid plan.
Research use case: Gamma is built for the last mile of a project, when you have to turn a finished study into something a stakeholder will read. You can upload a brief or a report, let the tool pull out what it judges to be the essentials, and shape that into a short, top-line deck. Work that used to take an afternoon can take a few minutes.
The benefit: Speed. In minutes you can have a polished deck of ten to fifteen slides without spending hours on design, and with a saved theme, it already matches your organization's look. For a researcher asked to move at the pace of the decision, that can be the difference between being in the room and arriving after the decision is made.
The limitation: Something has to decide what counts as essential, and a tool built to produce a clean, confident deck rarely treats a caveat as essential. Lines like "this held for only one group" and "we should be careful here" are the first to be smoothed away. When the tool builds charts, it also chooses the scale and the comparison, and those choices can make a difference look bigger or smaller than it is. Because the output looks so finished, the polish hides what was trimmed. The tool gives you speed. Protecting the caveats and checking the charts is your job.
Prompt Pattern: Persona
Two prompt patterns can help researchers use that speed well. The first, called Persona, lets you reshape your research for very different audiences without starting over each time.
What it does: A persona prompt tells the AI whose point of view to take. You are not changing the findings. You are asking the tool to present the same findings the way they would need to land for a specific group.
Research use case: Picture one study headed to three rooms in the same week. The executive team wants the bottom line and the business implication. The frontline team wants to know what changes in their day-to-day work. The board wants risk and direction. Instead of building three decks from scratch, you give Gamma the same source and ask it to shape the deck for each audience in turn.
Why this prompt pattern works: A persona gives the tool a filter for what to emphasize and what it can assume the reader already knows. That focus makes the output feel tailored instead of generic, and it is a large part of the speed you gain.
Limitation: Emphasis is a form of selection. Tailoring to an audience is where "what this audience wants to hear" can slide into "what we choose to show." A persona prompt may drop the caveat that complicates the story for a particular room, or pick the chart that makes the point most dramatic, because a simpler story is what you asked for. The pattern makes you fast. It does not keep you honest. That is the job of the second pattern.
Prompt Pattern: Flipped Interaction
The second pattern, called Flipped Interaction, is the sparring partner.
What it does: Most prompting runs in one direction. You ask, and the AI answers. A flipped interaction reverses that. You tell the AI to interview you first and ask the questions it needs answered before it builds anything. It draws out the details a rushed researcher tends to skip.
Research use case: Before a single slide gets built, the AI asks the questions a sharp colleague would ask. Who is this deck for? Which findings have to stay in, no matter how short it gets? Which caveat cannot be dropped? Which comparisons need to be shown at their true size? Where is the result that complicates the headline? You answer in plain language, and only then does the deck get built, anchored to what you said must stay in.
Why this prompt pattern works: The risk comes from the mental shortcuts we take when we are rushed, like rounding off the inconvenient parts. A flipped interaction adds a pause at the moment speed would skip it. You have to name the caveats before they can be forgotten.
Limitation: The pattern can only ask. It cannot answer for you, and it cannot ask about a blind spot you do not know you have. A researcher in a hurry can still give thin answers and get a thin result, and the AI has no way of knowing that a group missing from your data ever existed. The questions bring out what you know but have not said. They cannot recover what was never collected. As with every pattern in this series, this is a check on your thinking, not a replacement for it.
Practice Prompt: Persona (with a complete report)
"Act as a research communications lead preparing this study for [insert audience] who will use the study [insert context]. Build a top-line deck from the attached report that emphasizes what this audience [actions to take/ needs to decide]. Do not change or soften any finding. Start every bar chart at zero and label the base size on each chart. Include one slide titled 'What to Read Carefully' that lists the key caveats, base sizes, and any group whose results differed from the headline."
Practice Prompt: Flipped Interaction
"I need to turn the attached report into a short deck for [insert audience] who will use the study [insert context]. Before you build anything, ask me no more than [choose a number/ 3-5] questions one at a time until you know which findings must stay in no matter how short the deck gets, which caveats cannot be dropped, which groups had results that differ from the main story, which comparisons need to be shown at their true size, and what evidence complicates the headline. Then summarize my answers back to me, and build the deck so that every item I named is included."
Closing Reflection
The stakeholder in that meeting was right about one thing. Research that arrives after the decision is research nobody uses. AI tools like Gamma help us keep pace, and that matters. The pace is also changing who tells the story. More research now moves through marketing, product, and strategy teams, and more decks will be built by people who were never close to the data.
As that happens, the researcher's value shifts from building every slide to setting the standard every slide must meet: the findings that must stay in, the caveats that cannot be cut, the charts that show differences at their true size, and the groups whose results did not fit the headline.
Speed and rigor can work together, but only if we check our work as we go. The shortest version of the truth is rarely the whole truth, and the parts we cut first are usually the voices that were already hardest to hear.
________________________________________
ABOUT THE AUTHOR
Brittne Kakulla, PhD, is a researcher at AARP focused on technology adoption among adults 50-plus. She serves on the Insights Association IDEA Council and writes about AI literacy for research professionals.
Prompt patterns come from:
White, J., Fu, Q., Hays, S., Sandborn, M., Olea, C., Gilbert, H., Elnashar, A., Spencer-Smith, J., & Schmidt, D. C. (2023). A prompt pattern catalog to enhance prompt engineering with ChatGPT. arXiv:2302.11382. https://doi.org/10.48550/arXiv.2302.11382
Paivio, A., & Csapo, K. (1973). Picture superiority in free recall: Imagery or dual coding? Cognitive Psychology, 5(2), 176–206. https://doi.org/10.1016/0010-0285(73)90032-7