The Main Flaw in Drinking Games and Feedback Loops
The story began when I saw this episode of a Russian late-night show called “Evening Urgant”. The episode featured a drinking game that involved pronouncing tongue twisters. The rules were simple: the person, who failed the challenge had to drink, then the challenge repeated. The game I just described has an inherent flaw of games that consider drinking a punishment. These rules create a system with positive feedback, and this isn’t a good thing when it comes to drinking games. Activities of this kind are played to build up a company, which implies getting drunk together, but a positive feedback loop makes those who failed once drink further but those who don’t fail tend to stay sober, which effectively divides the company into two parts.
Party modeling
Let’s simulate a party, that plays a drinking game with a positive feedback loop. The party involves 100 people playing 100 rounds of a game. The Drunkness of each player will be characterized by a number from 0 to 1, where 0 — completely drunk and 1 — absolutely sober. The state of the party is characterized by a python list containing conditions of players. Each round a player has a chance to win which is proportional to their sobriety level. If the player loses he drinks and gets 10% more drunk. It’s that easy.
Now we can see drunkness dynamics. As expected, the plot shows that the party quickly divides into two groups: while some people stay sober, others slide into drunkness after making the first mistake.
The animated histogram illustrates my point vividly. Two narrow peaks indicate that the majority of players are either drunk or sober, with almost no one in between.
The solution
The situation can be improved by introducing negative feedback to the game. Say, after each round those who completed the task are to drink, the players who failed skip. This makes those players who are better at the game drink more, which reduces their skills, so those who are outliers on a drunkenness scale tend to either drink towards the majority or skip until the majority reaches their level.
Let’s look at the model, the only difference here is that I changed “>” to “<” in the 6th line, turning positive feedback into negative one :
The state of the party now is much more homogeneous. Obviously, now people get drunk at the same rate.
The animated histogram is now an individual peak slowly approaching 0. The dynamics indicate a perfect party, with all people in the same state.
Conclusion
Many popular drinking games including beer-pong consider drinking as a punishment for losing. That creates a positive feedback loop, that negatively affects the cohesion of the party. Instead, rules should consider drinking as a reward. Models show that this approach creates negative feedback instead, which makes the collective more homogeneous in terms of drunkness level.
