Amplifying Group IQ with Conversational Swarms

In recent years, the field of artificial intelligence has made significant strides, particularly in enhancing human collaboration and decision-making. 

A groundbreaking advancement in this domain is the development of Conversational Swarm Intelligence (CSI), a technology that leverages the principles of Swarm Intelligence (SI) and the capabilities of Large Language Models (LLMs) to boost the collective intelligence of networked human groups. 

The paper “Towards Collective Superintelligence: Amplifying Group IQ using Conversational Swarms” by Louis Rosenberg, Gregg Willcox, Hans Schumann, and Ganesh Mani, explores this innovative approach and its potential to achieve Collective Superintelligence.

The Foundation of Swarm Intelligence

Swarm Intelligence is a natural phenomenon where biological groups enhance their collective decision-making capabilities by forming real-time systems. A phenomenon in bird flocks, bee swarms, and fish schools, to name a few.

This collective behaviour allows these groups to make smarter decisions than any individual member could on their own​​. Inspired by this natural process, Artificial Swarm Intelligence (Swarm AI) was developed in 2015 to enable human groups to make better decisions by forming similar real-time systems​​. 

Swarm AI has proven effective in various applications, from financial forecasting to consumer insights, but it has been limited to narrowly defined tasks such as probabilistic forecasting and multiple-choice decision-making.

Introducing Conversational Swarm Intelligence (CSI)

To overcome these limitations, researchers developed Conversational Swarm Intelligence (CSI) in 2023. This method combines Swarm AI with the power of LLMs, allowing large, networked human groups to engage in real-time conversational deliberations on open-ended problems. 

Unlike traditional Swarm AI, which requires questions to be in numerical or multiple-choice formats, CSI enables more natural and dynamic interactions​​ .

The CSI platform, known as Thinkscape, divides large groups into smaller subgroups, each consisting of 4 to 7 members, and inserts an AI agent into each subgroup. 

These AI agents observe the deliberations, distil salient points, and pass critical insights to other subgroups. This process creates an overlapping conversational structure that allows the entire group to function as a unified system, converging on solutions that maximise collective intelligence​​ .

Amplifying Collective Intelligence

The study described in the paper aimed to evaluate the effectiveness of CSI in amplifying group intelligence. Participants were divided into groups of approximately 35 individuals and tasked with answering IQ test questions using the Thinkscape platform. A baseline group took the test individually, achieving an average score of 45.7%. In contrast, the groups using CSI achieved an average accuracy of 80.5%,with an effective IQ increase of 28 points​​ .

This significant improvement suggests that CSI is a powerful method for enabling conversational collective intelligence. The results place CSI groups in the 97th percentile of IQ test-takers, demonstrating a substantial amplification of collective intelligence compared to both individual performance and traditional wisdom of crowds (WoC) methods​​ .

Overcoming Human Limitations

One of the key challenges addressed by CSI is the “cocktail party effect.” This is where humans naturally focus on local conversations and tune out distractions from neighbouring groups.

CSI overcomes this limitation by using AI agents to facilitate real-time overlap among deliberating subgroups. Which allows for the propagation of information and insights across the entire network, enhancing the group’s ability to converge on well-supported decisions​​ .

Real-World Applications and Future Research

The potential applications of CSI are vast, ranging from enterprise collaboration and civic engagement to strategic priority-setting and market insights. The technology’s ability to scale across groups of any size makes it a viable pathway to achieving Collective Superintelligence. Future research aims to evaluate larger groups and more complex, open-ended problems, further exploring the capabilities and benefits of CSI​​ .


The development of Conversational Swarm Intelligence marks a significant advancement in enhancing human collaboration and decision-making.

By leveraging the principles of Swarm Intelligence and the power of LLMs, CSI has been able to amplify collective intelligence. Which also achieves results far beyond traditional methods. As research and development in this field continue, CSI holds the promise of transforming how large groups of people work together to solve complex problems.

Ultimately, paving the way towards Collective Superintelligence.

More in the Blog

Stay informed on all things AI...

< Get the latest AI news >

Join Our Webinar Cloud Migration with a twist

Aug 18, 2022 03:00 PM BST / 04:00 PM SAST