A Computational Analysis on Creating Engaging Scientific Outreach and Educational Content for Social Media

Introduction:: Throughout the height of the COVID-19 pandemic the scientific community faced a public relation problem with the broader public as misinformation was rampant and distrust grew. The day-to-day integration of social media in the lives of the general public facilitated this as it came with the advantage for the user to be able to curate and control the type of content that they prefer to consume. Scientists are quickly faced with a more overarching question of how to effectively communicate with the public and appeal to the next generation of scientists, but lack guidelines on specifically how to do so effectively in the new age of social media. To study the gap between these the scientific and general public communities, we created the TikTok account @Scieneers for social outreach and to investigate effective methods of science communication. To this end, as a biomedical engineering laboratory, we leveraged computational and machine learning methods to identify what types of “interest signals” (likes, comments, shares, bookmarks) TikTok preferred, and resulted in high performing ( >50,000 views) lighthearted science-based content. Based off of this analysis, we have created a baseline guide for scientists to use to get started with creating their own effective science communication pages, and achieve our mission statement of humanizing scientists and making academia more accessible.

Materials and Methods:: The Tiktok page (@Scieneers) was created and 125 videos were released over the course of 17 months (fig a). During this time the “Analytics” feature available to all content creators on Tiktok was used to track key engagement markers for each of our released videos such as video views, likes, comments, bookmarks, shares, average watch time, and percent of audience that watched the full video. Other metrics from our account that were recorded were followers over time, amount of time that it took to produce the Tiktok, categorized data of types of content produced, and growth of follower count. Computational and machine learning methods were then used to calculate descriptive statistics, clustering, classification, linear regression, and correlation analyses.

Results, Conclusions, and Discussions:: Throughout the growth of our TikTok account(fig b), there were key increases in followers that occurred from videos that were labelled as “high performing”, meaning that they received higher than 50,000 views. Low performing videos were classified as videos that got less than 10,000 videos, and “medium performing” videos were between 10,000 and 50,000 views. Throughout this time, we experimented with the amount of time that it took to produce high performing videos(fig c). We saw that there was no correlation between videos that took the lowest amount of time to create versus videos that took a significant amount of time to produce. A comparison between percent of the full video that was watched versus the video length(fig d), showed that the most engaging type of content is generally below 25 seconds with 80% of few high performing videos being below this threshold. The correlation analysis for our account’s engagement metrics(fig e) showed with no surprise that “likes” were most highly correlated(0.92) with video views with 1 being the most ideal correlation. The next best predictor of video success was bookmarks that had a correlation of 0.86. This emphasizes the need to create content that stays with the viewer long after the first time view. Further data and commentary on different types performance metrics, and a comparison between different TikTok science accounts will be included in presentation.

Social media is a powerful tool that can be used to engage with audiences of all ages and backgrounds as videos can quickly become viral and gain upwards of billions of views. Through proper utilization scientists can relate and provide meaningful connections with the broad public to educate and inspire the next generation of scientists. Here we present a thorough computational analysis on how to grow a science-based TikTok account, and analyze the characteristics that resulted in high performing videos. Moving into the age of a dominant social media presence, the image of science and the scientist must, also, adjust to this paradigm shift of the consumption of news and entertainment to maintain engagement with the broader public.


Acknowledgements (Optional): :

References (Optional): :