← All research demos
Interactive paper · Golos et al. · PLOS ONE, 2023
What do people need
from science communication?
Dear Pandemic invited readers to send questions to a team of scientists. By combining AI topic modeling with researchers’ interpretation, this study traced how their information needs changed over a year.
The central lesson: readers wanted science they could use in everyday decisions—and their questions changed with the news, vaccine rollout, and family life.
Read the paper · August 24, 2020–August 24, 2021
3,839
question-box submissions
90%
of submissions from the US
Questions followed events
Interest in viral variants peaked in January and July 2021, alongside news attention to Alpha and Delta.
Vaccine questions evolved
The focus moved from development to safety, then dosing, and finally how to behave after vaccination.
Everyday life mattered
Questions about family and children rose around winter holidays and summer, when people were making plans.
Explore the findings · Figure 2
Follow a question through the year
Start with a finding below, or choose any of the study’s 25 topics. Move the date slider to inspect the trend.
Average daily topic probability (%)
Smoothed trend95% confidence bandDaily observations
Smoothed trend
At the selected date
Each question can include several topics. The chart shows the model’s average topic probability, not the percentage of people asking about it. The vertical scale adjusts to the selected topic.
Interactive reconstruction of Figure 2 using the released topic values and the study’s model specification. View the published figure.
How to read this study
The researchers used a topic model to identify patterns in English-language questions, then interpreted those patterns as 25 topics within six themes. Figure 2 tracks daily average topic probabilities with a generalized additive model, a smooth curve through changing values. The shaded band shows uncertainty around this fitted trend; it does not capture uncertainty in the topic model itself. Each observed day contributes equally to the curve.
These are questions from people who chose to engage with Dear Pandemic, not a representative survey of the public. Timing suggests how information needs related to events; it does not establish that an event caused a change.
The topic counts and associated words shown here correspond to Table 1. The interactive curves are reconstructed from the study’s public supplements; they are not the original figure image.