If a chatbot can Slack convincingly in the boss’s voice, will employees follow orders once they realize the CEO is actually a machine?
A novel two-part study finds that an artificial intelligence (AI) chatbot trained to write like a technology company’s CEO responded to questions so believably that many employees thought the answers came from the boss himself. But there’s a caveat: When employees perceived that a response came from AI—even if it didn’t—they rated those responses as “less helpful” than those they thought came from the CEO, demonstrating a classic case of “algorithm aversion.”
It’s an AI-age twist on the classic Turing Test, developed by British computer scientist Alan Turing in 1950 to judge whether machines could exhibit “intelligence.” Called the “Wade Test,” after the CEO of the company the researchers studied, the analysis is among the first to showcase AI’s ability to replicate the unique characteristics of a specific person’s writing style, says Prithwiraj Choudhury, the Lumry Family Associate Professor of Business Administration at Harvard Business School.
“We trained an algorithm to write using the same words and phrases and punctuation meter and grammar and mistakes and abbreviations the CEO uses,” Choudhury says. “What that tells us, generally, is that at least technologically, we can create a writing bot for any one of us.”
Generative AI stands to make workplaces more efficient by automating busy leaders’ routine tasks—such as the electronic communication that takes 24 percent of a CEO’s time, studies show. In theory, this would allow the executive to devote more time to strategic planning, for example. Yet the new research indicates that there’s still a long way to go before humans cede the craft of writing to machines—if that ever happens in an organizational context.
Overcoming aversion “is the billion-dollar question in front of the AI industry,” says Choudhury, who teamed on the paper with Bart S. Vanneste, associate professor at University College London’s School of Management, and Amirhossein Zohrehvand, assistant professor at Leiden University.
“My prediction is that every single employee one day will have their own communication bot, just like today we all have email,” Choudhury says. “But the question is, how do we get this to work, so it’s credible and widely accepted?”
Featuring Prithwiraj Choudhury By Ben Rand / Courtesy of Harvard Business Review