New Research Shows Personal AI Agents Help Networked Teams Solve Problems Faster

The Center for Collaboration Science at Carnegie Mellon University (CMU CoLab) and Unanimous AI, a world leader in hyper-communication, released new research showing that networked teams can solve collaborative problems faster when each member communicates exclusively through a personal AI agent instead of traditional group conversation. This research aims to solve communication bottlenecks in high-stakes, time-pressured scenarios like emergency response, where members multitask while sharing and receiving critical information from one another. 

Anyone who has participated in a large video conference or tried to keep up with a busy Slack channel knows the problem. On a group call, only one person can speak at a time, so important information waits its turn. In a group chat, everyone can post at once, but critical details get buried under a wall of text. Research shows both problems get worse as teams grow and compound when individual members have other critical tasks to address while collaborating, like monitoring an urgent situation.

In this study, 15 networked teams, each with 7 or 8 members, were tasked with solving a simulated high-stakes problem framed as identifying a terrorist based on a set of evidence that was split up across team members. The only way to solve each problem was to combine multiple pieces of evidence from each member, while rejecting conflicting evidence and red herrings.

Each team solved the problems in two ways: (A) by traditional group chat, as is used in many settings with tools like Slack and Microsoft Teams, and (B) by using a novel hyper-communication framework called Hypercom AI in which each team member engages only with a dedicated AI agent. These agents, called Conversational Surrogates, are tasked with passing relevant information to other human team members via their dedicated AI agents. The AI agents are there to coordinate communication – they inject no new information into the conversation and generate no deductions of their own. 

What the study found

Teams using Hypercom AI converged on the correct solutions in an average of 11.6 minutes, compared with 14.3 minutes in traditional group chat. The 19 percent reduction was statistically significant and held after controlling for which “mock terror plot” each team solved. Accuracy was essentially identical: 93 percent for hyper-connected teams and 92 percent for traditional ones. The researchers describe these results as the first demonstration that a distributed team can solve a problem requiring every member’s contribution while communicating and collaborating entirely through a network of intervening AI agents. The benefits of this technology are likely to increase with group size, especially among multi-tasking team members.

Where it matters

The researchers see immediate applications wherever people must share information while doing their jobs under pressure, including emergency response, disaster relief, cybersecurity, medical triage, and military command. A first responder in the field, for example, could keep working while her agent shares her observations with the other team members and brings back relevant updates from others in the field. The same approach applies to the critical operations across large organizations.

“Leveraging AI technology to integrate distributed expertise offers significant benefits to many organizations, particularly when decisions need to be made quickly,” adds Professor Anita Williams Woolley, Professor of Organizational Behavior at CMU’s Tepper School of Business, and a co-author on the research. “There could be additional opportunities to facilitate completely asynchronous coordination and dramatically reduce the need for meetings that currently clog everyone’s calendars and slow down progress.”

The research study, “Evaluating an Agentic Hyper-Communication Framework for Optimized Human Deliberation Across Distributed Teams,” is available now as a preprint. Its authors are Louis Rosenberg, Hans Schumann, and Gregg Willcox of Unanimous AI, and Anita Williams Woolley and Ganesh Mani of Carnegie Mellon University. The research was funded by the U.S. Air Force through AFWERX contract FA864926P0228, Hyper-Communication for Optimized Command and Control.

Availability

Hypercom AI is available as a beta feature within Unanimous AI’s the Thinkscape® platform. The views expressed are those of the author and do not necessarily reflect the official policy or position of the Department of the Air Force, the Department of Defense, or the U.S. government.

About Hyperchat AI technology

Hyperchat AI enables networked human teams of potentially unlimited size to hold productive real-time conversations by text, voice, or video. A breakthrough technology, it enables teams with dozens or even hundreds of members to efficiently share information, debate options, brainstorm ideas, evaluate risks, forecast outcomes, and converge on optimized solutions.

About Unanimous AI

Unanimous AI develops technologies that make human teams significantly smarter. Modeled on the dynamics of biological swarms, its patented Swarm AI®, Hyperchat AI™, and AgentSwarm™ technologies turn networked teams into a Collective Superintelligence. Unanimous AI’s award-winning software platforms, Swarm® and Thinkscape®, have been validated by numerous academic studies and are used by large organizations around the world. For more information, visit: unanimous.ai.