Benefits of a Humble Intelligence
It's not uncommon for people today to interact with artificial intelligence. With easy access to internet chat bots, video games, and smart phones; it's likely that you've already spent a lot of time talking to, or otherwise interacting with, machines. With their growing abundance in mind, researchers at Yale University sought to learn more about how these machines influence human interaction with other humans. (Source) Margaret L. Traeger, Ph.D. candidate in sociology at the Yale Institute for Network Science (YINS) and lead author of the study, explains that "[b]ecause social robots are becoming increasingly prevalent in human society, people are encountering them in stores, hospitals and other everyday places. This makes understanding how they shape human behavior important...We know that robots can influence the behavior of humans they interact with directly, but how robots affect the way humans engage with each other is less well understood,"
In this experiment, 153 volunteers were divided into 51 groups consisting of three people and a robot. Their task: Work together to design efficient rail systems over 30 rounds of game play. The only variable throughout the experiment was the AI assigned to each group. At the end of each round, an AI would either remain silent, say a generic comment related toward the game, or say a more "vulnerable" message. (joke, short story, or acknowledgment of a mistake if one was made)
Sad Robot, by Benjamin Eberhard
After collecting data from all 51 groups, the researchers found that people teamed up with a vulnerable AI spent twice as long talking to each other when compared to other groups. These players also reported higher levels of enjoyment compared to players grouped with a silent or generic AI. These results show that influence from a more humble AI can augment the quality of interaction between multiple people.
This experiment reminds me of the Turing Test: An experiment where a person must determine whether they are talking to a human or a computer. However this experiment run at Yale is not focused on the interaction between human and computer, but between human and human. For this reason, I was a bit skeptical when I first read the article. I didn't expect the AI to impact the interaction between humans as much as it did. Additionally it seems that, like the Turing Test, more "human like" communication will gave an AI an advantage over those with more generic messages. (I'm sure it'd also perform much better than a silent AI, for obvious reasons...)
Quick and Easy Image, by Jason Palmer
I never thought of the influence of an AI on human interaction before today. I'm glad that there are people dedicated to learning more about how it applies to how we treat each other. Especially since AI are becoming more common, and they made a clear difference in how participants in the study interacted with each other.
I'd never thought about how AI could influence how humans interact with each other. It makes sense that an AI telling jokes and stories would lighten the mood because people probably found it amusing and unique. I wonder if this effect will decrease as AI becomes more common and the novelty wears off or if it will stay the same?ReplyDelete
We should keep in mind that the reason that it's an attractive quality in humans is because it is not present in everyone, and is in fact less common than other kinds of responses. I think that if there were A.I. with different kinds of personalities like humans, then the effect would remain.Delete
I think it is interesting to think about this study and to compare the results to group projects that I have participated in, robots aside. Typically when working with group mates who are willing to admit mistakes or who make jokes throughout the assignment is far more enjoyable than groups with little to no engagement. It is interesting to see that there are such strong similarities even though the teammate in this study was AI.ReplyDelete
I think that this is something that people would not think about initially, but after looking at the results, it is not surprising. This is an interesting concept that produces other questions relating to the impact of artificial human qualities. When we simulate human qualities in AI (or the AI has human qualities), does this cause the same effects as humans portraying these qualities such as being humble? What are the differences in results?ReplyDelete
Humans are psychologically predisposed to attribute human traits to non-human entities. Like Olivia said, its easier to work with someone who is willing to make jokes and accept blame, so the AI that simulates this would have the same effect as an actual human doing the same.ReplyDelete
It's very interesting point, many people ignore the fact of how robots influence us. I would have never taught that it will affect a person. This reminded me of the Eliza chat-boot that could also be used to spark conversation and help people interact with one another.ReplyDelete
Cool blog, Jason. While reading I found myself thinking about the times where conversational robots would be nice to have and other times where it would be a nuisance to have a robot making a conversation with you. For example, a chat-bot or a host at a restaurant may be a good place to incorporate those conversational robots, however I think that if I were talking to a robot taking my order at a fast food place, I wouldn't want to engage in a conversation which may be unnecessary. I think it's interesting to think about ways that these types of AI would be used in society today because I don't think we're all that far off from seeing them in our every day lives.ReplyDelete
Maybe technology like this could be combined with that of automated driving. Uber rides can be some of the most awkward spots for conversation. Would a conversational Uber driver help through the awkward and typical routine of, "So... how long have you been driving for Uber?"Delete
I wonder if people were more responsive to the humble robot because it was the most emotional of the options. Humility is good quality for team work but I wonder if that was the key variable in the result as opposed to humanity.ReplyDelete
I feel like Dr. Stonedahl and other professors doing online lectures might appreciate an AI to break the awkward silence before each class starts to get the students engaged and talking to one another... It will be interesting to see how this technology progresses as it becomes more available. Great article by the way Jason!ReplyDelete
Totally agree with you. I found that virtual classes are pretty one-sided. In other classes, I see that professor is the only person talking, very few students are contributing. And it is really difficult for professors to keep an interesting virtual class like that because it feels like you are talking to a machine,not interacting with real people.Delete
I think that's same for all human. Joke and short story brings out something to talk about and allow people associate them to other funny things, yet keep silence or just saying general comment everybody know won't really help for communication.ReplyDelete
I agree. I watched this show called "100 Humans" on Netflix and they have this experiment about human interaction and their efficiency doing tasks. There are two separated groups of people, both of which strive to build the tallest marshmallow-spaghetti tower. One group is motivated by a money prize, the other was building it for fun. The result was that the 'building for fun' group outright beats the other group. One major difference between those two group is how they approach the problem. The prize money group is much more intense than the other group, and that kind of communication really worsen the result by a large margin.Delete
I believe this brings light into the fear of AI that some people might have in today's society. I can only imagine how these robots that tell jokes and tell stories would be in for example in nursing homes. Like maybe robots can be pets? Just some thing silly to think about.ReplyDelete