Seoul Halloween Stampede and Information Cascades
This past weekend, many people in Seoul, South Korea set out to celebrate Halloween, the first large-scale celebration of this holiday following the start of the pandemic. Due to the incredibly crowded volume of people on the streets of the Itaewon area, over 150 individuals passed away when a stampede started, leaving even more injured. Over 130,000 people were estimated to have gathered in this popular nightlife district that evening, and this number is significantly higher than in past years, most likely due to a particularly heightened sense of excitement from a lack of full-blown COVID-19 restrictions. This fatal crowd surge was catalyzed by the intense overcrowding of an already packed, narrow alleyway. Even more horrifically, many partygoers were unaware of the severe injuries occuring and kept trying to push their way through the crowd, worsening the stampede and causing a ripple effect throughout the mass of humans.
A major cause of this chaotic surge could be most likely due to the effects of information cascades. As discussed in Chapter 16 of the Networks, Crowds, and Markets textbook, this phenomenon of “following the crowd” happens due to the sequential decision-making process of individuals, as people’s earlier actions influence others’ later decisions. So, while people visiting Itaewon might have had previously-held private information about their specific intended party venue of the night, masses of crowds increasingly developed in concentrated areas due to information cascades. A tightly-packed alleyway could signal to individuals that something exciting and popular is occurring in that location, so perhaps despite private information that people hold, the size of this crowd entices them to similarly check out what is going on in that area and mimic their behavior. The information cascade in this instance would be considered to have caused informational effects, rather than direct-benefit effects, because the actions of early decision-makers indirectly affected later actors by changing the information they hold due to the suggestion that they have some sort of knowledge about their present location that is unknown to the later actors.
Consequently, following this severe crowding due to herding, many people were left to be trampled. And, again, in a repetitively cyclical manner, because of information cascades, when one person starts to run or move in fearful or frantic way, this signals to others that they should engage in a similar action, despite their own private information, and this only further worsens the stampede. Bayes’ Rule is a way in which to mathematically model such information cascades; due to the incredibly large scale of this tragic incident — as opposed to the herding experiment example with urns from the textbook — it would be difficult to use this rule of conditional probability and pinpoint the precise actions of specific individuals, but with further investigation of this incident, it would be interesting to identify smaller-scale catalytic actions that triggered larger-scale crowd behavior.
In summary, the enormous scale of this horrific indicent truly epitomizes the immense power of information cascades over human behavior. Law enforcement officials have been criticized for not planning ahead enough to monitor the effects of such massive crowds in Seoul this Halloween, so perhaps it would be beneficial for them to take lessons from the effects of information cascades in relation to crowd stampeding behavior to mitigate the risks of future tragedies.
Sources:
https://www.nytimes.com/2022/10/31/world/asia/seoul-halloween-crowd-accountability.html
https://www.cs.cornell.edu/home/kleinber/networks-book/networks-book-ch16.pdf