The following article was originally published by AI at Wharton.
A curious thing happened at Stack Overflow following the public release of ChatGPT in late 2022. The world’s largest website for computer programmers to ask each other technical questions lost about 1 million daily visits, resulting in a 15% decline in traffic in about four months.
Meanwhile, nothing changed at Reddit, the popular social site where people chat about topics as broad and unlimited as their imaginations. Gordon Burtch, information systems professor at Boston University’s Questrom School of Business, believes the disparity reveals an important social effect that happens when people rely on generative artificial intelligence, like the kind employed by ChatGPT. Instead of turning to their peers for help, Stack Overflow users are prompting the chatbot for a fast, automated answer. But Reddit users are sticking around for human interaction.
Burtch shared this tidbit — an interesting finding from his much deeper co-authored research — during the first episode of a new webinar series offered by AI at Wharton titled “AI Horizons.” Streamed live on Dec. 15, the episode focused on “AI and Wellbeing.”
Wharton marketing professor and AI at Wharton co-director Stefano Puntoni hosted the hourlong webinar featuring Burtch, Julian De Freitas, a business administration professor and director of the Ethical Intelligence Lab at Harvard Business School, and Weiguang Wang, a computer and information systems professor at the University of Rochester’s Simon Business School. (Watch video below.)
The webinar series was inspired by Wharton’s Business and Generative AI Conference held last fall in San Francisco. The conference sparked so much excitement that organizers organized the series as a way to share the latest information on a larger scale. Topics for upcoming webinars include AI’s effect on the workforce, machine creativity, and innovation.
“Interest in AI has exploded, and this is an opportunity for us to bring out cutting-edge research around specific topic areas,” said Mary Purk, executive director of AI at Wharton.
The Power and Potential of AI
The presentations piqued interest in topics becoming more relevant as AI infiltrates every aspect of work and play. Burtch said the Stack Overflow results can be seen more generally.
“People are not engaging as much with human peers,” he said. “You might be worried about a new employee, for example, not connecting with peers in the organization as a result of these new tools.”
De Freitas shared his research into chatbots and mental health. As loneliness increases, more people are turning to chatbots for interaction. That’s an upside to the technology, De Freitas said, but it could also be dangerous for people in a crisis.
“It will be up to humans to put guardrails around it.”— Stefano Puntoni
He and his colleagues, which include Puntoni, analyzed about 3,000 app-based conversations and found about 5% of them were about mental health. About 37% of the 5% involved crisis messages, such as a mention of suicide. Yet the apps recognized the crisis only 60% of the time.
“A few apps provide appropriate responses to explicit messages, yet all apps have room to improve,” De Freitas said. “I think this research raises important questions for how these apps should be regulated, and also what the managers of these apps should do.”
Wang observed the propensity for people to treat ChatGPT and other new bots as if they were human, even before interacting with them. It’s a stark contrast from previous models that were seen as task-oriented calculators.
“Imagine treating a fridge in a similar manner? It would be quite funny,” he noted. The personification of technology is why Wang examines the socialization of chatbots, especially in health care. For one experiment, he launched a socialized fitness bot to interact with Twitter users through personalized messages. For example, when an overweight user complained to the bot that people make fun of her at the gym, the bot replied by writing that others were just jealous of how much she was lifting. In contrast, the same query fed into ChatGPT resulted in a short, academic essay on societal judgments about obesity.
“I see the power of combining these predictive models, plus medical knowledge, plus social skills,” Wang said. “This is a great time for business to take the chance to combine this strong power of generative AIs in their specific domains.”
Puntoni agreed, saying the challenge is striking the right balance and tone. “I can imagine in health care there are people who don’t want chatty, friendly banter,” he said.
The panelists view the future of AI with excitement coupled with concerns about bias, safety, privacy, standards, and regulations. Those issues should be addressed as the technology develops, they said.
Puntoni sees the technology as almost limitless. “It will be up to humans to put guardrails around it.”
“The webinar series is about democratizing knowledge,” said Purk. “Everyone can have a voice in how we shape the future, but you have to lean in and learn and join the conversation. Then you can be part of the solution.”
Register for upcoming “AI Horizons” webinars here.