Advice from information technology (Artificial Intelligence) experts can have a huge impact as human experts, according to a Penn State research team. However, both bad news carriers can find that they are losing influence when their negative ideas are against the good crowd.
In the study, researchers found that machines that generate recommendations — or Artificial Intelligence experts — were as influential as human experts when AI experts recommended which image a user should add to their online business profile. However, both Artificial Intelligence and human experts failed to make comments when their response was incorrect and in line with popular opinion among other users, S said. Shyam Sundar, James P. Jimirro Professor of Media at the Donald P. Bellisario College of Communication and director of the Media Effects Research Laboratory.
Sundar, who is also an ambassador for Penn State’s Institute for Computational and Data Science (ICDS), said the findings could indicate that there are times when the crowd’s viewpoint – also called the bandwagon effect – may contradict experts’ views that they are Artificial Intelligence or human. He also pointed out that both the AI-powered experts and the people who evaluate the good in the business profile picture were able to influence the user experience of the image. However, if the experts did not like the picture and the crowd said they could properly examine it, the influence of the experts diminished.
Because people are increasingly using social media to look for feedback, references that suggest expert opinions and the bandwagon effect can be important factors in influencing decisions, according to Jinping Wang, a senior communications technology consultant and first author of the study.
“Nowadays, we often turn to online platforms to get input from other people – such as peers and experts before making a decision,” Wang said. “For example, we can turn to those sources where we want to know what movies to watch, or what photos to upload to social media platforms.”
Artificial Intiligence experts tend to be more expensive than human experts and can work 24 hours a day, that is, Wang suggests, which can make them attractive to businesses online.
Investigators also found that the group’s Artificial Intelligence status – in this case, determined by the country’s origin – did not seem to have any effect on the person’s acceptance of its recommendation. However, for human experts, an expert from the same national survey who gave the negative image evaluation often had the impression that a human expert from an unknown country had provided an equally bad picture.
While findings suggesting group status may not mean that one appreciating the judgment of experts sounds like good news, Sundar suggests that the same cultural assumptions may still apply to Artificial Intiligence experts, but they can be hidden in programs and data training.
“It could be two – it’s bad because it all depends on what the Artificial Intelligence feeds,” Sundar said. “While it is better to have faith in the power of AI than to discriminate against culture, we must remember that if you train AI in images from a single culture, they can give misleading recommendations to images designed for use in other cultural contexts.”
Investigators, who report on their findings on the upcoming computer issue on Human Behavior, also found 353 people with online crowdfunding to participate in the research. Participants are randomly selected to view a screenshot of a website that gives users recommendations for their business profile photos. Participants were also told that the website allows feedback from other users of the platform, in addition to professional experts. Screenshots showed a variety of factors studied by investigators, including whether expert ratings were human or Artificial Intelligence, whether their response was positive or negative and whether the source of the rater came from a national identity of the same, different or unknown.
In the future, researchers plan to investigate the group’s strengths more closely and examine whether the gender of the therapist played a role in influencing the perception.