Lessons from YouTube on COVID-19, hand washing and its algorithm.
- Written by Marija Indriūnaitė, Anna Kopf, Beatrice Silaite, Meira Gilbert -
7th April, 2020 – Amsterdam, NL
“Wash your hands!” – something that now not only your mother tells you to do, but every single government in the world. Why? Because currently, it is the strongest weapon in the artillery that we have for the fight against COVID-19. According to the World Health Organization “COVID-19 is the infectious disease caused by the most recently discovered coronavirus. This new virus and disease were unknown before the outbreak began in Wuhan, China, in December 2019.” If you are currently not living under a rock, this disease has influenced your life in one way or the other immensely over the past few months, whether it is your travel plans that got cancelled, you are forced to move back in with your parents, or there have been some other more serious consequences.
Fig. 1: Total confirmed COVID-19 cases (December 31st, 2019 – March 15th, 2020)
As you probably already know, the number of COVID-19 infections have skyrocketed since the beginning of the year, where the virus was first detected. Fig. 1 shows the worldwide, exponential development of the COVID-19 infections – leading one to believe that the infection-rate will not flatten-down drastically any time soon, if there are no extreme measures implemented! But no matter how strict the quarantine rules are in your country, social distancing by itself will not protect you from the virus if you do not wash your hands.
Let’s be honest here for a minute: remember all those memes, showing us how to apply song lyrics while washing your hands for 20 seconds so the process would be more fun? Or seeing instructions on how to properly wash the hands hanging in the public restrooms or other public spaces (good times, right)? Did you not for once at least find yourself thinking “why on Earth would I need instructions for this”? Well, while for some people it might seem like common sense to take proper care of their hand hygiene, for the rest, unfortunately, it is just not “a thing”. And it does not necessarily apply only to underdeveloped or highly polluted countries – yes, we are looking at you, the Netherlands.
In 2015, WIN/Gallup International conducted a worldwide survey in which more than 62 000 participants were asked whether they wash their hands with soap and water after a visit to the bathroom. The findings revealed that only 50% of the Dutch find it as a necessary ritual, making this nation the least hygienic within the borders of Europe. While the percentages were much lower in countries such as Mexico or China, where clean water and sanitation are known for being of poor quality, the Netherlands do not experience this problem, which proves that educating people on handwashing is something that should not be taken naively and is not related to regional resources.
But the question at hand is who is in charge of educating the public about the best measures to “flatten the curve” and create a safer environment for everyone? Some might argue it is solely the task of the government. But as you might know from your own experience, when uncertain about a specific topic, seeking information, or just want to learn the newest trends – we love to turn to our trusty social media platforms. Research has proven that the human brain processes images 60,000 times faster than text – meaning that people are mostly seeking their information from the probably most visual social media platform out there: YouTube. Furthermore, According to Alexa Ranking, this video-sharing platform is the second most visited website in the world, only giving place to its parental corporation Google. Since it’s early days, YouTube has evolved into a multi-functional social media space, where both entertainment and education peacefully co-exist. Regarding the latter aspect, YouTube is a great tool to enlighten as many people as possible since it accommodates around 2 billion monthly users.
Now, that the general perimeters of the topic are covered, let’s get into the nitty-gritty of this article. What follows is an investigative research on the recommendation system of Youtube, before and during the COVID-19 pandemic. More specifically speaking, all this chit-chat about ‘hand-washing’ was not without reason, as it will function as a representative sample of this study.
You might ask yourself, why should I have to care about this? Well, if you are an avid social media user, as most of us are nowadays, this paper can give you insight on a larger discussion at hand that affects you on a day-to-day basis! In times like these, when our safety and health are on the line, can YouTube be another weapon in our artillery during the fight against the global pandemic? Can the platform be seen as an objective and sufficient actor for educating the public or must we discover that YouTube is strongly leveraging this increased media coverage for its own advantage? And what does a public health and data expert think about YouTube’s role in all this? Keep reading to find out!
Let’s Talk Numbers
In order to answer our questions, we (a group of Data Journalism students at the University of Amsterdam) turned to the Youtube Data Tools software. There are currently five tools, or “modules,” available – each focuses on a different part of the YouTube platform, such as “Channel Info” and “Video Network.” For our research we decided to use the “Video List” module. This module creates a list of videos and their associated statistics from one of four sources: videos uploaded to a specific channel, a playlist, videos specified by a list of ids, or videos retrieved from a particular search query. Bingo.
Using the search query setting, we were able to enter “hand washing” and retrieve a list of 50 videos associated with that query for a specific time frame. Because we were tracking the “hand washing” associated videos with the rise of coronavirus, we extracted 5 datasets: From November 1st – December 1st (as a control, before COVID-19), December 1st – January 1st (first official cases reported), January 1st – February 1st (major outbreak in China), February 1st – March 1st (outbreak in Italy, Iran, and spread throughout the globe), and March 1st – March 15th (state of widespread pandemic). These datasets were each 50 videos long, with each video possessing an additional 22 characteristics such as “channelTitle” (name of the title which published the video, “video Title” (name of the video), and so on.
First, we wanted to analyze how many of the videos the “hand washing” query retrieved were related to coronavirus. To do this, we manually looked at each of the 5 datasets. For each month, we counted how many of the videos either explicitly mentioned “corona” “coronavirus” “COVID-19” or another related term in their video title or description. After doing this, we found that each month, the number of “hand washing” videos greatly increased (Fig. 2), which demonstrates how the YouTube algorithm increasingly relates “hand washing” with corona. Furthermore, this suggests the “corona” hand washing videos are taking the spots of other hand washing videos, ultimately suggesting that the algorithm as a whole has favored them compared to other “hand washing” content.
Fig. 2: Number of videos mentioning Coronavirus (November 2019 – March 2020)
The next characteristic we analyzed was “VideoCategoryLabel,” which classified each video into a larger category. We found that the breakdown of the categories from November, before the coronavirus outbreak, to March, in the midst of it, was very interesting.
In November, the videos related to the “hand-washing” query were primarily from 2 major categories. The largest, at 34% of the videos, was “Education.” This category primarily included videos published by medical or educational institutions, and demonstrated how to properly hand-wash to prepare for the typical flu-season. The next biggest category was “People & Blogs,” where there were more informal videos published by individual users about proper hand hygiene. Notably, in November only 9 categories were represented in the 50 videos. A full breakdown of the categories in November 2019 can be viewed in the visualization below (Fig. 3).
Fig.3: Category Labels in November 2019 for search query “hand-washing”
We then compared the categories from the November dataset to those from March, in the midst of the COVID-19 pandemic. In March, the most striking difference is the number of categories represented. Here, 11 categories are represented (Fig. 4), and most only with a small percentage. This may represent how by March, “hand washing” not only was related to educational “how-to” videos about hygiene, but many other aspects of society. We see in the graph below that surprising fields such as “Music,” “Comedy,” and “Entertainment” are represented alongside “Education,” “News & Politics,” and so on. Many of these videos are coronavirus-related viral dance challenges about hand washing, or coronavirus themed bits from talk shows such as The Late Show with Jimmy Kimmel, rather than videos purely related to “hand washing.” Overall, by March “hand washing” had permeated additional aspects of society through an affiliation with the coronavirus.
Fig. 4: Category Labels in March 2020 for search query “hand-washing”
To summarize our findings of the categories, we made one last additional data visualization on the subject. This graph shows the percentage of each category represented per month. We can see how the “People & Blogs” category, where many doctors demonstrate proper hand hygiene, had one of the most drastic downfalls. By March, many categories which were not previously significantly represented, such as “Entertainment” or “Nonprofits,” were gaining representation because the videos related hand washing to the coronavirus. Overall, the change from November to March demonstrates how the algorithm chooses many videos which are only related to “hand washing” through a coronavirus affiliation.
Fig. 5: Video categories of “hand-washing” content over time (November 2019 – March 2020)
Next, we looked at the change in viewing activity over time (Fig. 6). We found that the median number of views per month has been steadily increasing, while the average number of views (Fig. 7) have been decreasing since January. While this may originally seem counter-intuitive, this is probably because many of the videos related to “hand washing” have been newly published and related to corona. Examples of such videos might include news reports or PSAs. Meanwhile, the increasing median may show that starting in January, people started to watch more hand washing videos in general. Therefore, this viewing number would be more equally distributed among videos. Overall, these two graphs paint a picture of how the corona virus has led to both new “hand washing” videos being published, and more “hand washing” videos being watched in general.
Fig. 6: Median of views per month of ‘hand-washing’ videos (November 2019 – March 2020)
Fig. 7: Average numbers of views per month for ‘hand-washing’ videos
Lastly, we wanted to further investigate how viewer engagement may have changed with the increase of corona-related hand washing videos (Fig. 8). We separated the videos into two categories: “corona” and “non-corona,” by looking whether the title and description used any key words related to COVID-19 (e.g. Corona, Coronavirus) and compared how the two groups differed. We found that over time, the engagement, in the form of likes, was noticeable higher with videos related to the COVID-19 outbreak. The like average of these videos peaked during the month of February, where the magnitude of the disease became apparent. As for the videos not explicitly mentioning the disease, the like average is under 5000 during all months, with a light rise in March.
Fig.8: Averages number of likes of videos including and excluding key-words related to COVID-19 in the title and description of ‘hand-washing’ videos
So far, we have learnt two main things from this data: that the YouTube algorithm could be seen to favor or push coronavirus related videos, and the change in engagement with ‘hand washing’ videos reflects the panic around COVID-19. At first glance this may seem as quite an obvious finding, but if considered more in depth, it does raise questions of how reliable and trustworthy is YouTube as an information source and how helpful it is in combating real life issues such as COVID-19 pandemic. Is YouTube’s algorithm taking advantage of current situation to gain more engagement, metrics and further fueling the panic, or is it educating people about hand washing and helping to fight the pandemic?
We interviewed Leighann Kimble, a public health expert and educator, to gain more insight in our questions. Kimble is pursuing her PhD in Public Health at the Universidad de Chile, and is currently a Public Health Educator in Amsterdam through the Council on International Educational Exchange (CIEE). In addition to her background in Public Health, Kimble also has a passion for data, which makes her a perfect person to speak to about our data-driven investigation.
According to Kimble, YouTube does have educational value, and she says “YouTube is a great platform for developing educational materials that can be disseminated to the public for health promotion and prevention but has also been used and continues to be used to develop technical material.” Kimble stresses that YouTube has already been used for public health purposes in the past and that it is effective for distributing necessary information both to the public and professionals. Considering the current coronavirus outbreak, educational ‘hand washing’ videos are also an outbreak prevention measure. “They are reliable in that they are able to teach the basics of handwashing” states Kimble, “and handwashing was first identified as one of the most effective means of preventing transmission.”
When asked about the role of visualisations and entertaining content in public health, Kimble also saw the educational potential, as it “helps to keep learner attention and also may promote the desired behaviors by allowing the viewer to associate the practice with something positive/entertaining”. Thus, even the viral handwashing videos including dances, songs or comedic bits can be effective, if the informative part is correct.
However, there are things to look out for. Kimble clarifies that even though hand washing videos provide necessary primary information, they leave out “the context for why handwashing is important and why these practices should be maintained in addition to other practices (such as staying at home and practicing social distancing).” So as much as these videos are reliable and educating, they do not provide all crucial information and may lead to misconceptions. People should evaluate available information provided to them, as according to Kimble misinformation is common in public health and “YouTube is possibly a source of misinformation and misconception.”
In fact, the role of YouTube in spreading misinformation has been explored before. In 2018, The Guardian released an article comparing YouTube’s algorithms to “misinformation engines”. The article claims that YouTube does not care about the educational value of their videos, only how to best attract and keep viewers. Specifically, an Ex-employee of YouTube featured in the article spoke out about the sketchy tweaking of algorithms happening in order to keep the viewers glued to the screen.
More recently, YouTube has come under fire for not doing enough to combat misinformation. In April 2019, an article from Bloomberg laid out how YouTube executives purposefully let toxic and conspiracy videos run rampant in order to maintain high engagement and profits. In a follow-up article from the Columbia Journalistic Review, reporter Mathew Ingram makes a strong case for how YouTube has acted “too little, too late” to combat the rampant misinformation published on their website and pushed by their recommendation algorithm. In short, these articles expose how YouTube has known about the problems with their recommendation algorithm, but didn’t act. Therefore, there is no way of knowing where YouTube algorithm, combined with misinformative content, may lead.
Nevertheless, Kimble does believe ‘hand washing videos’ found on YouTube do have educational value, but they might just represent ‘one side of a coin’. We can not be sure of the primary goals of these videos – be it to gain more engagement, metrics or to educate. It should also not be seen as a revolution because YouTube is only one of the tools used to educate about coronavirus. In short, according to Kimble, YouTube is not powerful enough by itself:
“In terms of the use of YouTube to promote/teach handwashing, I do not believe YouTube on its own has been effective. I believe the promotion of materials on YouTube and the urging of the public to practice proper handwashing techniques as a measure in COVID-19 interventions are what has made the promotion of handwashing more effective than it may have been with simply using YouTube videos. It is my opinion that the increase in handwashing as linked to these videos is specific to the concerns surrounding COVID-19 and the emphasis placed on handwashing as a means of preventing transmission.”
It is the reaction of the public and professionals to the virus that creates the discourse around it, and the YouTube algorithm simply builds around this pre-existing discourse. The intentions behind the way YouTube algorithm works are not clear cut either and we have no means to tell if it aims to educate at all. Thus, rather than be seen as educating, YouTube should be seen as a tool that can be used to public health officials to educate.
So YouTube: Good Guy or Bad Guy?
So, we now know that even silly YouTube videos can be used as educational tools (maybe the Netherlands should try this tactic). However, our research and interview clarify how this content must be consumed critically. Ultimately, we found how YouTube is “good” in the sense that it allows for widespread information dissemination and reflects current trends and concerns: for example, it knows that in March 2020, someone searching “hand washing” is more likely to want information about the coronavirus than a vlog on how African women wash their clothes (a video featured in our November 2019 ‘control’ dataset). However, YouTube is also “bad” because the information it gives may lack the context or quality necessary to be effective, and lead to misinformation. So basically YouTube is like a high school jock in a romantic teen comedy – a popular guy with a relatively good heart deep down and a few bad qualities outside, but no matter how “douchey” the latter might be, oh boy, we cannot take our eyes off him.
Overall, this article highlights both aspects of the algorithm, and raises questions one should keep in mind when using it. And of course, we also learned that a simple YouTube query is never as simple as it seems!
P.S. Do not forget to wash your hands!