Real-time Internet Control of Situated Human Agents
We present an online platform, called BeeMe, designed to test the current boundaries of Internet collective action and problem solving. BeeMe allows a scalable internet crowd of online users to collectively control the actions of a human surrogate acting in physical space. BeeMe demonstrates how intelligent goal-oriented decision-making can emerge from large crowds in quasi real-time.We analyzed data collected from a global BeeMe live performance that involved thousands of individuals, collectively solving a sci-fi Internet mystery. We study simple heuristic algorithms that read in users' chat messages and output human actionable commands representing majority preferences, and compare their performance to the behavior of a human operator solving the same task. Results show that simple heuristics can achieve near-human performance in interpreting the democratic consensus. When human and machine's output differ, the discrepancy is often due to human bias favoring non-representative views. We discuss our results in light of previous work and the contemporary debate on democratic digital systems.