Towards Collective Superintelligence, a Pilot Study

Abstract— Conversational Swarm Intelligence (CSI) is a new technology that enables human groups of potentially any size to hold real-time deliberative conversations online. Modeled on the dynamics of biological swarms, CSI aims to optimize group insights and amplify group intelligence. It uses Large Language Models (LLMs) in a novel framework to structure large-scale conversations, combining the benefits of small-group deliberative reasoning and large-group collective intelligence. In this study, a group of 241 real-time participants were asked to estimate the number of gumballs in a jar by looking at a photo. In one test case, individual participants entered their estimation in a standard survey. In another test case, participants converged on groupwise estimates collaboratively using a prototype CSI text-chat platform called Thinkscape. The results show that when using CSI, the group of 241 participants estimated within 12% of the correct answer, which was significantly more accurate (p<0.001) than the average individual (mean error of 55%) and the survey-based Wisdom of Crowd (error of 25%). The group using CSI was also more accurate than an estimate generated by GPT 4 (error of 42%).

Introduction. I. INTRODUCTION. Collective Intelligence (CI) refers to the field of study that investigates how human groups can solve problems and make decisions that exceed the capabilities of the individual members. Often referred to as “wisdom of crowds,” the most common CI methods involve capturing asynchronous input from large groups through polls, surveys, or prediction markets, and then aggregating the data statistically. This often produces more accurate results than the median individual produces on their own, thereby exhibiting measurable intelligence amplification. Such statistical techniques are commonly used in groupwise estimation as well as probabilistic forecasting and other simple decision-making tasks. Applicability is generally limited to tasks that can be abstracted down to discrete questions in which participants are asked to provide numerical input or make multiple-choice selections or forced-choice comparisons.

Discussion / Conclusion. VI. CONCLUSIONS We conducted a collaborative estimation task using a new technology called Conversational Swarm Intelligence. It combines the methods inspired by the biological principle of Swarm Intelligence with a novel use of LLMs that enables large, distributed groups to hold coherent conversational deliberation and reach thoughtful and unified solutions. We performed a study with a large 241-member group of randomly selected users and tested their ability to estimate the number of gumballs in a jar. This task was chosen as it replicates a very common test used to illustrate the power of collective intelligence. The results showed that “conversational collective intelligence” is a viable method in which human groups deliberate through natural language and reach solutions of amplified accuracy. In this case, the use of CSI technology reduced the mean absolute error across individual participants by 77% (p<0.001). This was more effective than the traditional collective intelligence method of aggregating asynchronous survey results.