01 / HUMAN + AI
Collective intelligence
Understanding how human and language-model groups reason, deliberate, and reach decisions.
Human–LLM deliberation · Responsible AI learningExplore this lens
Tengfei Shao 邵 騰飛
Assistant Professor · Waseda University
My research connects AI agent simulation, complex networks, and collective behavior, with an emphasis on empirical validation.
AI AGENT SIMULATION / COMPLEX NETWORKS
Simulating agents.
Understanding
collective behavior.
Nodes: agents · Lines: interactionsConceptual visualization, not research data.
SCROLL TO FOLLOW THE CONNECTIONS
Explore the research narrative in five chapters.
01 AGENTS
Intelligence starts
with individuals.
Different agents bring different attributes, information, and perspectives. Their differences are the starting point for studying a collective.
How do individual differences shape collective outcomes?
Heterogeneity · Individual behavior
Select an agent to explore its connections.
02 INTERACTIONS
A connection
changes the system.
An exchange creates a relation. Following who interacts with whom makes the process of collective reasoning visible.
How do interaction patterns influence collective reasoning?
Information exchange · Relational evidence
Light pulses trace existing connections.
03 NETWORKS
Local structure.
Collective patterns.
Neighborhoods, communities, and recurring local structures offer different views of a complex network. Their interpretation requires careful comparison and analysis.
What can network structure reveal about complex systems?
Network motifs · Communities · Null models
Inspect an agent’s neighborhood or a connected triad.
04 COLLECTIVE BEHAVIOR
See what emerges
between agents.
Coordination, agreement, and differentiation are properties of a collective. I investigate how these patterns relate to the interactions beneath them.
What emerges when many agents interact?
Coordination · Collective reasoning
Color changes are a scripted illustration, not a research result.
05 VALIDATION
A convincing simulation
needs evidence.
Simulations become scientifically useful when their behavior can be tested against observations. I examine outcomes, interaction processes, and network structure to clarify what a model can support.
How faithfully does a simulation represent the system it studies?
01ObserveDefine the empirical reference.
02CompareTest behavior and structure.
03ClarifyEstablish the model’s limits.
Holdouts · Calibration · Reproducibility
Validation connects simulated behavior with empirical observations.
RESEARCH / THREE CONNECTED LENSES
From interaction
to understanding.
AI agents make behavior explorable. Networks make relations visible. Evidence determines what a simulation can explain.
02 / RELATIONAL STRUCTURE
Complex networks
Examining how local interactions and network structures reveal patterns in complex systems.
Network motifs · Communities · Attributed graphsExplore this lens03 / SIMULATION + EVIDENCE
Empirical validation
Testing where agent-based simulations align with observed behavior, and where their limits matter.
Holdouts · Null models · CalibrationExplore this lensCURRENT WORK
Questions
in progress.
My work connects collective reasoning and social simulation with applications in AI ethics education and sustainable markets.
Collective intelligence
Measuring validity gaps between human and language-model groups
Responsible AI
Evaluating how case-based AI ethics education changes reasoning
Social simulation
Testing generative simulators against emergent network structure
Sustainable markets
Tracing trust, value retention, and recovery in resale systems
SELECTED PUBLICATIONS
Ideas become
evidence.
Research across networks, human behavior, and connected markets.
Full publication record2025Dissecting the Second-hand Luxury Market Dynamics: Insights from E-commerce versus Brick-and-MortarJournal of Information Processing 33, 219–230Lead & corresponding author2024Multiple Clusters Discovery Utilizing Network Motifs for Community Improvement: Insights from Tourism and Goods’ TransactionsJournal of Information Processing 32, 308–318Lead & corresponding author2024Time Series Network Analysis for Profit Dynamics in Pre-owned Luxury Goods Market Based on Network MotifsNew Frontiers in Artificial Intelligence, 5–20Lead & corresponding author2024A Study of AI Ethics Education in the Context of Japanese Job-Hunting Based on Case Method Using Network AnalysisACM International Conference Proceeding Series, 92–100Lead & corresponding author2022Value Creation Framework for Tourist Destinations Based on Designable Evaluation NetworkSocial Networks 71, 1–11Co-author2021Discovering Multiple Clusters of Sightseeing Spots to Improve Tourist Satisfaction Using Network MotifsIEICE Transactions on Information and Systems E104.D, 1640–1650Lead & corresponding author
BEHIND THE RESEARCH
Curiosity.
Structure.
Evidence.
I am an Assistant Professor at Waseda University in Tokyo. My research brings together agent-based simulation, network science, and computational social science.
I develop methods for studying how interactions shape collective patterns, with an emphasis on transparent analysis and empirical validation.
researchmapTEACHING & ACADEMIC PRACTICE
Making methods
accessible.
I teach database systems, data analytics, AI ethics, and computational research practice. My academic work connects methodological rigor with accessible education and reusable research workflows.
START A CONVERSATION
What could we
understand together?
Research collaboration, academic exchange,
and questions worth exploring.