[Paper Review] AI-enhanced Collective Intelligence

Hao Cui, Taha Yasseri|arXiv (Cornell University)|Mar 15, 2024

Artificial Intelligence in Healthcare and Education4 citations

TL;DR

This paper proposes a multilayer network framework for AI-enhanced collective intelligence, integrating cognition, physical, and information layers to model synergistic human-AI collaboration. It demonstrates that diverse human and AI agents, when strategically coordinated, can achieve superior collective intelligence beyond individual or isolated AI capabilities, with real-world applications in knowledge creation and decision-making.

ABSTRACT

Current societal challenges exceed the capacity of humans operating either alone or collectively. As AI evolves, its role within human collectives will vary from an assistive tool to a participatory member. Humans and AI possess complementary capabilities that, together, can surpass the collective intelligence of either humans or AI in isolation. However, the interactions in human-AI systems are inherently complex, involving intricate processes and interdependencies. This review incorporates perspectives from complex network science to conceptualize a multilayer representation of human-AI collective intelligence, comprising cognition, physical, and information layers. Within this multilayer network, humans and AI agents exhibit varying characteristics; humans differ in diversity from surface-level to deep-level attributes, while AI agents range in degrees of functionality and anthropomorphism. We explore how agents' diversity and interactions influence the system's collective intelligence and analyze real-world instances of AI-enhanced collective intelligence. We conclude by considering potential challenges and future developments in this field.

Motivation & Objective

• To conceptualize a multilayer network model of human-AI collective intelligence integrating cognition, physical, and information layers.
• To analyze how diversity in human attributes (surface-level to deep-level) and AI functionality/anthropomorphism shapes system dynamics.
• To investigate the impact of temporal mismatches and feedback delays on human-AI coordination efficiency.
• To identify ethical challenges in AI integration, including transparency, bias, and liability in human-AI teams.
• To advocate for interdisciplinary collaboration in building Social AI systems that address societal challenges through enhanced collective intelligence.

Proposed method

• Develops a multilayer network representation with distinct cognition, physical, and information layers to model human-AI interactions.
• Classifies human agents along dimensions of surface-level (e.g., demographics) and deep-level (e.g., cognitive styles, values) diversity.
• Categorizes AI agents by functional capability (e.g., NLP, computer vision) and anthropomorphic traits (e.g., perceived agency).
• Analyzes temporal dynamics by aligning human circadian rhythms with AI response latency to optimize coordination.
• Applies network science and complex systems theory to model interdependencies and emergent intelligence in hybrid systems.
• Uses real-world case studies (e.g., Wikipedia, citizen science, collaborative software) to validate the framework’s applicability.

Experimental results

Research questions

• RQ1 How can a multilayer network model effectively represent the interplay between human and AI agents in collective intelligence systems?
• RQ2 What role does agent diversity—both in humans and AI—play in shaping the emergent intelligence of hybrid collectives?
• RQ3 How do temporal mismatches between human behavioral rhythms and AI response delays affect coordination and system performance?
• RQ4 What ethical risks emerge when AI is integrated into human-AI teams, particularly regarding transparency, bias, and liability?
• RQ5 In what ways can AI not only assist but also co-evolve with human collectives to solve complex societal challenges?

Key findings

• The multilayer network framework successfully captures the structural and dynamic complexity of human-AI collective systems, enabling analysis of interdependencies across cognition, physical, and information layers.
• Collective intelligence emerges not from individual intelligence alone but from synergistic interactions, with the 'c-factor' predicting group performance independently of average individual IQ.
• Feedback delays between human input and AI response can significantly impair coordination, especially in time-sensitive collaborative tasks.
• Ethical challenges such as reduced trust and increased defection in human-AI teams arise when transparency is lacking, particularly in strategic interactions.
• AI integration can amplify existing biases in decision-making if training data or system design is not carefully audited, highlighting the need for bias mitigation strategies.
• Real-world applications like Wikipedia and citizen science demonstrate that AI-enhanced collective intelligence outperforms isolated human or AI efforts, especially in large-scale knowledge creation and data analysis.