How Herd Mentality Influences Our Decisions Online
Have you ever held back from posting on a Facebook group or an online forum to skim through other people’s comments? Doing so means you’re probably trying to get a sense of what the group as a whole is thinking says a new research paper published in the journal Science Advances.
As humans, we evolved to make decisions in groups. However, for most of human history, these groups were offline. The majority of these groups were also relatively small, meaning that you knew or could get to know most people in them. Today, thanks to social media, we communicate directly or indirectly with thousands of anonymous, or unfamiliar, individuals regularly.
So, when you are trying to get your opinion heard or figure out what to do next, how does your brain deal with millions of virtual strangers voicing their thoughts?
Researchers based in the University of Washington combined AI with psychology to try and understand how people make sense of online environments. The researchers also wanted to see if they could make an AI algorithm that could better understand how humans make group-based decisions.
To do that, the researchers recruited 29 students and introduced them to a digital game based on the “volunteer’s dilemma.”
In the game, players start with a set sum of money. As the game goes on, the players can either contribute money to a communal pot or “free ride” (i.e., keep the money to themselves). The latter makes sense if all or nearly all other players choose to donate their earnings. If the total amount contributed by all players exceeds a certain amount, every player (as part of the group) gets a reward on top of their initial endowment.
During the experiment, the students were placed inside MRI machines. They were also told that they were playing with dozens of other anonymous players located elsewhere.
Throughout the course of up to 50 rounds of the game, researchers gave the players information on how the group was doing overall, i.e., how many people chose to contribute to the communal pot and how many chose to freeride instead. Unbeknownst to the students, the group they were playing with was actually a computer program.
Studying how people played the game with the computer, the researchers noticed that people often made decisions based on what they thought the “group” was doing rather than what made the most sense for them as individuals.
For example, in the last round of the game, when all the other players were said to have contributed to the communal pot, the most sensible thing for someone to do was to “freeride.” However, during the experiment, nearly all people decided to contribute instead.
In other words, most players chose to donate their money when they were given information that other people in the group were likely to have contributed too.
Speaking to the website Science Daily, lead author of the research paper Koosha Khalvati explained the participants’ surprisingly altruistic behavior. He said, “they are aware that even though they are anonymous to others, their selfish behavior would decrease collaboration in the group in future interactions and possibly bring undesired outcomes.”
After most of the participants had played the game under the same conditions, the researchers plotted a mathematical model of their reactions. They then used this model to create an AI algorithm designed to predict how people would react in the future.
Generating an AI algorithm proved surprisingly easy, at least when compared to modeling individual human behavior. The AI developed by the group was able to predict the likely behavior of the group better than any other method.
So what does this mean for online groups? According to the senior author on the study, Rajesh Rao, “In online forums and social media groups, the combined actions of anonymous group members can influence your next step, and conversely, your own action can change the future behavior of the entire groups.”
This research could also have enormous implications for developing AI that interacts better with large numbers of people online.
Online groups have a huge impact on people’s behavior. Unfortunately, studies have shown that this impact is often negative. In the future, AI might be able to help change this. For example, an anti-hate speech AI might be able to deter groups from inciting violence by sensing what a group is thinking and promoting peaceful comments. Until then, we should try our best to go against the herd— and think for ourselves.
Imagine a world where soldiers will have cyborg exoskeleton armor that will give them increased physical and mental performance. The Warrior Web program, funded by the