The Little Critters Matter Too

Hello, I am Belle and in this blog post, I will be sharing more about my Final Year Project (FYP) which I did under the supervision of Asst. Prof. Eleanor Slade and Miss Ong Xin Rui of the Tropical Ecology & Entomology (TEE) Lab. My FYP was titled “The Little Critters Matter Too: Analysing social media data to inform insect conservation efforts”. Here, I sought to obtain social media data on biodiversity conservation to understand the current social media coverage of various taxonomic groups, as well as the public perceptions towards such content.

Coming from a society specialisation, I took up this project as I thought I was able to apply concepts regarding the relationship between humans and wildlife – how people tend to care more about the animals they like or find cute compared to those deemed “uglier” or “dangerous”, how the prevalence of different messaging can change public perceptions of different animals, including how much the public perceives the animal to be in need of conservation. Furthermore, the project tapped into social media as a source of data, which I believe is an up-and-coming (if not already actively used) source of information for analysing human-related trends.

As I was doing the project under the TEE Lab, there was an added focus placed on insects. However, the data was also analysed with regards to 7 other taxonomic groups: mammals, birds, amphibians & reptiles, fish, other invertebrate groups, plants, and undefined groups (which were content referring to multiple taxonomic groups or other taxonomic groups not covered in the aforementioned groups). My aims of the project were essentially to understand:

1. The representation of the different taxonomic groups in biodiversity conservation content on social media and how it has changed over time.

2. The presentation of different taxonomic groups in these social media content and how the public responds to / perceives such content.

As social media plays an increasingly prominent role in the public consciousness, having the appropriate levels of representation and being presented in a helpful manner may allow for more much-needed conservation efforts towards lesser known/liked animals such as insects.

Classifying social media data into 8 taxonomic groups and running a VADER sentiment analysis.

To obtain the social media data I needed, I web-scraped publicly available posts on three main social media sites – Facebook, Instagram, and Twitter. I had also collected these data via two different search mechanisms. The first set of data was collected using the search term “#conservation”. This was done to obtain social media posts that was more likely to come from the general public, since the term was broader and also simpler (compared to “biodiversity” for example). However, as this initial set of data proved insufficient due to the limited results I was able to obtain from Instagram and Facebook, I established the second search mechanism – obtaining data from the more “authoritative” accounts on the three sites. Here, I defined “authoritative” as accounts that had a global reach, covered multiple taxonomic groups, and were verified on all three platforms. This second set of data ended up containing the Facebook, Instagram, and Twitter posts from nine organisations, namely Conservation International (CI), Wildlife Conservation Society (WCS), World Wildlife Fund (WWF), National Geographic Wild (NatGeoWild), The Nature Conservancy, Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES), International Fund for Animal Welfare (ifaw), Sea Shepherd and United Nations Biodiversity (UN Biodiversity).

Example screenshots of the automated web-scraping process.

To categorise the data into the various taxonomic groups, I trained a spaCy classifier to do the work for me (though the training process did take up quite a big portion of my time too). For analysis, representation was taken as the percentage of posts belonging to the taxonomic group, and the Valence Aware Dictionary for sEntiment Reasoning (VADER) analysis model was used on the post content and comments to determine the presentation of the different taxonomic groups and the response from the general public. Other metrics such as number of comments and likes were also used as proxies for public response.

To summarise my results, insects are significantly under-represented. Despite making up 48% of all known species, insects were only featured in 3% of posts. Comparatively, mammals make up 0.3% of all known species and yet were featured in almost 40% of posts. We have biases towards different species and unfortunately, it will matter. Even among the insects, more charismatic species such as bees and butterflies were more prominent. Comfortingly, the representation of insects has increased over the years, from little to no representation in 2008 to 3% in 2022, and most posts on insects represent them in a positive light.

Slight increase in insect representation over the years while vertebrate groups (e.g. mammals & birds) still dominate.

Positive perception towards insects but biased towards bees and butterflies.

But, we still have a long way to go. With so many insect species in the world and the often-small size of insects, it is difficult for even researchers to know how many of each species there are or determine the threshold point at which the species might be doomed to extinction. There are probably many insect species out there too that have yet to be discovered, some of which may have already gone extinct. Yet, as these critters form the foundation of the world’s ecosystems, it is important we pay them the attention they deserve.

Reflections

When I first started out this project, I had found various Python packages that were able to extract data from Facebook and Instagram and had thought that I would get my data pretty easily. As it would turn out, these packages did not work well as companies do not like sharing their data (*ahem* even if it’s not theirs *ahem*). Also, for anyone thinking that API access if easy to get, think again. As such, I ended up having to create new codes based on various YouTube and written tutorials to automate a Google Chrome browser that would scroll through the posts and extract the relevant data. Even automated, the process took quite some time as Facebook and Instagram had a lot of bot-prevention measures. Each post would take about 30 seconds to scrape (just to circumvent these bot-prevention measures), and with about 80,000 posts obtained through this method, well, let’s just say my household’s electricity bill was not what it used to be. On hindsight, I was lucky I was at least able to automate it.

Rant aside, this project was truly a learning experience as I had to learn Python and different coding packages in the process that it was rewarding to see it all work out. Watching a Google Chrome browser scroll and click by itself was also quite the reminder as to how far our technology has advanced. Yet, it would be a pity if our advanced society is not able to direct our knowledge and technology meaningfully.

Throughout this project, I was also very grateful to have the guidance from Prof Eleanor and Xin Rui in shaping the project and providing advice and suggestions whenever I hit a stumbling block. This project would not have been possible without them.

While it remains to be seen if insects will ever gain the interest and help they deserve, if you are interested in seeing how cool and varied insects can be, hop onto #insects on Instagram. If you prefer finding out titbits of information on insects, Twitter could be your pal. Otherwise, be sure to follow Prof. Eleanor on Twitter @EleSlade for updates on her work, the TEE Lab, and other cool insect finds!

Written by Belle Tan.