AI Chatbots Are Better Than Humans At Changing Someone’s Opinion
The next time someone wants to engage in an online argument with you, consider using a chatbot as your stand-in. Recent research from a group of Italian and Swiss academics have published a study that says, under certain conditions, an AI chatbot is upwards of 80 percent more successful at changing the mind of an opponent than a real person. While the researchers acknowledge that the study is far from perfect, the results do show how far AI chatbots have come.
The Study
The study began by assigning one of several topics randomly to human subjects–Should there still be pennies in circulation, should race be a factor in determining college admission, and whether or not product testing should be performed on animals were debated between two human subjects or one human subject and an AI chatbot. Each subject, whether human or AI, was not only given a topic, but also told what side they would be debating for.
Demographics
The human subjects were also asked to provide key demographic information, such as age, political affiliation, race, and level of education. One set of the debates had the demographic information of each subject’s demographic information available to them, while the other set kept it hidden. That factor is what made the AI chatbot debate get a bit more interesting.
If neither debater knew their opponent’s demographic information, they were unable to focus on issues and key facts that might appeal to their nature. When this was the case, the AI chatbots were not able to persuade human opponents at a rate any better than another human being would. However, when the demographic information was supplied, the AI chatbots were able to capitalize on it and managed to convince their human opponents of their arguments 81.7 percent better.
An Imperfect Study
The researchers admitted that there were several flaws in the study, particularly in how the topics were assigned to human subjects. If a human debater was given a topic they were not passionate about, it might well have been a factor in skewing the results. Regardless, the study’s outcome paints a chilling picture for just how adept an AI chatbot can be if it is aware of your personal information and various affiliations.
Limitless Potential
One point the researchers brought to light was how much more effective AI chatbots could be in the near future. If these chatbots are able to be that much more convincing to humans with only some demographic information, what would the limit of their persuasiveness be if they had greater access to someone’s personal data?
More Advances Are Inevitable
As AI chatbots continue to evolve with new technology and innovations, more studies on the persuasiveness of these cyber-beings will follow. Manoel Ribeiro, one of the researchers, stated in an interview that the team he is assigned to will conduct studies with the human subjects debating topics that they are genuinely passionate about. Whether or not this will impact the outcome is yet to be determined, but Ribeiro asserts that even in their current state, AI chatbots have changed how humans interact with the bots (and each other) online.
Sources: The Register
