AI project aims to improve negotiation skills


Doctoral candidate Emmanuel Johnson is collaborating with two professors to create a virtual human that can help low-income populations with learning negotiation skills. (Deanie Chen | Daily Trojan)

Doctoral candidate and Fulbright Scholar Emmanuel Johnson is collaborating with Jonathan Gratch, the director of virtual human research at the Institute for Creative Technologies, and Peter Kim, a professor at the Marshall School of Business, to create virtual humans that can mimic the art of negotiation.

The three are examining methods to develop a better understanding of the human psyche. The group hopes to publish a working model that can be used to teach low-income individuals negotiation tactics, according to Kim.

“One of the goals [of] the [National Science Foundation] grant that we received is to help underrepresented minorities who wish to succeed in STEM-related fields,” Kim said. “And so one of the goals points to sorting organizations you want to join. You need some basic negotiation skills and to be able to negotiate jobs effectively as you enter into these organizations.”

Johnson said that his motivation to conduct this research was based off of personal encounters with the negative effect of inexperienced negotiation in America and his home country of Liberia.

“You think about a negotiation as something that we do every day. And so I saw it as a very critical skill to have, to teach, because many things have been negatively impacted by bad negotiations,” Johnson said. “We look at a lot of urban cities in America, even in Liberia, and so my thinking was, if you can teach people how to negotiate better, maybe I could help them not get taken advantage of.”

The team has developed the virtual human using game theory principles as a foundation, while adding on self-learning techniques to improve the program’s response width. Johnson said the concept of game theory provides a strong starting point for reasoning about multi-agent decision-making.

“Once we have a model based on game theory ideas, we can build a system using various AI techniques that responds in a very natural way, as well as make offers that are appropriate,” Johnson said.

Although the research has many academic implications, according to Gratch, the team is optimistic about the virtual human’s applications in other practical areas.

“We want to explore the idea that AI could negotiate a car for you, or your salary,” Gratch said. “So part of what we’re looking at is how the system negotiates on our behalf, and also the concept of negotiation as a new paradigm where machine and people can interact rather than a person versus machine.”  

While conducting research with his colleagues, Johnson recalled all the valuable lessons he’s learned for his academics and his future career path.

“I got guidance as well as the validity that the system that I’m building is actually teaching that something that can be taught in the classroom,” Johnson said. “As Professor Gratch and I think about how to design these systems, Professor Kim helps us make sure that is grounded in actual practice.”

The virtual human is still undergoing simulation and tests and will require some time before it is ready for public appearance. However, the three are continuing to work on the model and hope to release the negotiating virtual human in the near future.