Skip to main content



Can we predict diffusion and cascades on TikTok?

Do the concepts of diffusion of innovations need to be altered because of TikTok?  I would argue yes.  TikTok is a video sharing app which has recently exploded in popularity.  TikTok is different, and personally I think more interesting, because of its “For You Page” which is a feed of videos selected by an algorithm that is not specifically based on your friends, who you follow, or your area, but those characteristics may play a role.  As the New York Times explains it, the For You Page is “an algorithmic feed based on videos you’ve interacted with, or even just watched. It never runs out of material. It is not, unless you train it to be, full of people you know, or things you’ve explicitly told it you want to see. It’s full of things that you seem to have demonstrated you want to watch, no matter what you actually say you want to watch” (New York Times). While there is still a “following” section on TikTok, I would argue that most users spend their time on the For You Page finding new content.  I hadn’t considered how much of an impact the minor difference of having the main stage of a social media app be a feed of not the people you know or follow, but an ever changing mix of people, until something strange occurred to me. Two weeks ago I decided to order a water bottle because a random influencer on my For You Page (FYP) recommended it and it seemed like a good product.  The next day I received a snap chat in a group chat from a friend of mine who lives in a different state, goes to a different college and whom I had not spoken to in a few months, and she happened to be holding the exact same cup I had just ordered.  I immediately asked if she had bought the water bottle because of TikTok and she had, however, no one else in this group chat of 10 had seen or heard of the water bottle on TikTok and were confused as to why we had both ordered it without communicating about this decision. I thought it was funny that we had both been quickly swayed by a few random Tik Tok videos to purchase a water bottle, but as I thought about it further, I wondered what this meant in relation to the concept of diffusion we had learned in  class. 

Diffusion is based on the idea that technology spreads to someone when some fraction (the threshold value) of one’s neighbors adopt the technology.  Establishing a node or humans neighbors (friends, acquaintances, peers etc) is critical to understanding if a technology will spread.  In the case of mine and my friend’s TikTok purchase, it appears that this concept of diffusion can not be applied.  None of my friends (neighbors) had bought this product (technology) when I decided to purchase it.  I could not even tell you the names of the users which influenced me to buy it on my FYP. My friend also experienced the same phenomenon: without the influence of actual friends or neighbors she decided to adopt a new product. This new type of diffusion is not specific just to me, in fact, I asked 20 of my friends and 15 out of the 20 admitted to buying something solely because they saw it on TikTok.  An example greater than just me or my friends, is that of the TikTok leggings.  A random girl, with no prior TikTok fame or followers posted a video in a new pair of leggings and the video went viral.  Within weeks this pair of leggings from Aerie was sold out, simply because a “nobody” posted a video sharing her love for the leggings. This girl was not initially anyone’s actual friend or acquaintance, she was just someone on the FYP, yet her video’s place in user’s FYPs led to the rapid diffusion of this product and eventually a cascade as it approached virality. The popularity and growth of TikTok paired with its entirely unprecedented disregard for your network of friends means that the traditional diffusion model can no longer describe how, and if a technology or product will spread (on a TikTok user to user basis). 

 TikTok’s lack of concern for your real life friend network makes it difficult to predict who will influence you. At the basis of its app, “TikTok questions the primacy of individual connections and friend networks” (New York Times).  Furthermore, unlike Instagram where we could alter the diffusion model to somehow increase the value of celebrity or influencer neighbors because obviously those with more followers have greater influence, on TikTok the celebrity followings aren’t as important because of the FYP the. As an example, I haven’t seen Charlie Damelio (arguably TikTok’s most famous creator) on my FYP in the entire time I’ve owned the app. In order to predict cascades in a network of TikTok users we would need to understand the FYP algorithm, TikTok’s pride and joy.  Using the algorithm we could potentially create a diffusion model to predict a cascade by labeling each video in a person’s FYP as a neighbor, despite them not knowing each other in any other facet of life, and finding some threshold value where if a certain number of videos in the FYP mention a product you would adopt the product.  While this sounds simple in concept there are a few aspects which make it difficult. One being the fact that the FYP is infinite, new content will always appear so how do you track the percentage of videos which feature a product? Another aspect is that the algorithm is constantly changing with every video you view and every profile you click on, so to predict what videos someone would see, even in a day, would be difficult.  Finally, many times huge TikTok trends, those greater than just the purchase of products by individual users as I have been discussing, occur once a video reaches a certain number of likes then other people “duet” them or mention that video in other videos so the popularity almost becomes exponential.  At this point a cascade would almost always occur for that product or technology, but to predict which videos will go viral is almost impossible as demonstrated by the accidental fame of the leggings girl.

Finally, I would like to discuss the idea of clusters within a potential TikTok diffusion model. As I have mentioned before, TikTok’s algorithm shows you things you may not have actually expressed interest in.  This feature makes the formation of clusters almost impossible.  Formally put a cluster is “ is a set of nodes such that each node in the set has at least a p fraction of its network neighbors in the set”(Networks). As we have learned in class, clusters can halt the spread of a cascade because a node has too many neighbors within that one cluster, its density p is higher than the threshold value.  In the TikTok realm there are no “neighbors” so there is no possibility of clusters. This means that any technology has the possibility to spread throughout the entire app. While that doesn’t normally happen, the fluidity and lack of definition of TikTok allows for an idea or product to reach many different groups of people much easier than other social media apps.  Thus, although we may not be able to easily create a diffusion model for cascades on TikTok because of the perpetually shifting FYP, the lack of clusters on the app tells us that cascades are very likely to occur once an idea, technology or product gains traction. 

https://www.nytimes.com/2019/03/10/style/what-is-tik-tok.html

 

https://www.purewow.com/fashion/aerie-crossover-high-waist-legging-tiktok

 

https://www.cs.cornell.edu/home/kleinber/networks-book/networks-book-ch19.pdf

 

Comments

Leave a Reply

Blogging Calendar

December 2020
M T W T F S S
 123456
78910111213
14151617181920
21222324252627
28293031  

Archives