How Human-Machine Networks Shape Collective Behavior in a Hybrid Society

Table of Links

Abstract and Introduction

Human-machine interactions

Collective outcomes

Box 1: Competition in high-frequency trading markets

Box 2: Contagion on Twitter

Box 3: Cooperation and coordination on Wikipedia

Box 4: Cooperation and contagion on Reddit

Discussion

Implications for research

Implications for design

Implications for policy

Conclusion, Acknowledgments, References, and Competing interests

Abstract

From fake social media accounts and generative-AI chatbots to financial trading algorithms and self-driving vehicles, robots, bots, and algorithms are proliferating and permeating our communication channels, social interactions, economic transactions, and transportation arteries. Networks of multiple interdependent and interacting humans and autonomous machines constitute complex social systems where the collective outcomes cannot be deduced from either human or machine behavior alone.

Under this paradigm, we review recent research from across a range of disciplines and identify general dynamics and patterns in situations of competition, coordination, cooperation, contagion, and collective decision-making, with context-rich examples from high-frequency trading markets, a social media platform, an open-collaboration community, and a discussion forum. To ensure more robust and resilient human-machine communities, researchers should study them using complex-system methods, engineers should explicitly design AI for human-machine and machine-machine interactions, and regulators should govern the ecological diversity and social co-evolution of humans and machines.

Introduction

Robotic trains and cars drive us around, auction bots outbid us for purchases, ChatGPT answers our questions, while social media bots feed us with dubious facts and news. Modern society is a complex human-machine social system in which machines are becoming more numerous, human interactions with machines – more frequent, and machine-machine interactions – more consequential. With recent advances in generative AI models, the existential threat of unexplainable and uncontrollable general AI is looming large again[1,2,3].

However, when they are numerous and interdependent, even simple unintelligent artificial agents can produce unintended and potentially undesirable outcomes. If we want to prevent financial crashes, improve road safety, preserve market competition, increase auction market efficiency, and reduce misinformation, it is no longer sufficient to understand humans – we need to consider machines, understand how humans and machines interact, and how the collective behavior of systems of humans and machines can be predicted. We require a “new sociology of humans and machines.”

This narrative review synthesizes research and ideas related to social systems composed of multiple autonomous yet interacting and interdependent humans and machines such as algorithms, bots, and robots (Fig. 1). Similar to the conceptualizations of socio-technical systems [4], actor-network theory 5,6, cyber-physical social systems [7,8] social machines [9,10,11] and human-machine networks12,13, our principal assumption is that humans and machines form a single social system. In contrast, we do not approach machines as a single medium or entity – “technology” – but emphasize their multiplicity, independence, and heterogeneity and their interactions [14]. Thus, we target machines capable of social behavior, namely, a behavior that influences and is influenced by the behavior of other actors (human or machine) in the system.

Further, we approach the aggregates as complex social systems where network effects and nonlinear dynamic processes produce collective outcomes that cannot be necessarily deduced from individual preferences and behavior alone [15]. Our conceptualization extends and complements the ecological approach to studying machine behavior[16], the “hybrid collective intelligence perspective” [17], and the budding field of Social AI [18]. We aim to offer a conceptual overview of the topic. Hence, we do not utilize a systematic literature search strategy and instead present selected examples to showcase the new conceptualization[19,20,21].