GenAI Safety
We stress-test LLM alignment by identifying failure modes in safety filters and developing robust defense mechanisms for real-world deployment.
// research/
Research
Making sense, building together, teaching machines.
khyang-lab:~$ initializing mi-me-unit...
> loading: social_science.mod ✓
> loading: artificial_intelligence.mod ✓
> loading: mascots (mimiku, memeron).mod ✓
khyang-lab:~$ ready.
khyang-lab:~$ ls research/
> where social science meets the command line_
// general focus
// Collective Intelligence
Building Together
How do human-AI collectives synthesize information and influence social norms across digital and physical environments?
We study the emergent dynamics of collaborative platforms — from open-source communities to social media — to understand how collective intelligence forms, evolves, and sometimes breaks down.
// Media Intelligence
Making Sense
How does information shape public understanding of complex issues?
We apply natural language processing to track how media narratives shape public discourse on polarized issues, ranging from climate change to public health.
// Machine Intelligence
Teaching Machines
How can AI and humans learn from each other?
We integrate social science theory into machine learning to develop AI systems that move beyond pattern matching toward nuanced social reasoning.
// ongoing research
Social Science for AI
AI for pets/cats, LLM bias evaluation, and social simulation — bridging social inquiry and machine intelligence.
Climate & Environmental Communication
Heatwave journalism, risk perception, and attention fatigue — how media shapes public understanding of environmental crises.
Online Knowledge Production
Wikipedia dynamics, collaborative platforms, and the formation of collective opinions in digital spaces.
Narrative Economics & Media Analysis
Social media and market data pipelines — tracing how narratives propagate and shape economic behavior.
Cognitive Science for AI-Human Interaction
We bridge cognitive psychology and machine learning to map the behavioral gaps between human reasoning and LLM architectures, optimizing the synergy of human-AI collectives.
AI for Education
We build platforms that scaffold academic literacy, using LLMs to translate complex research into structured learning paths grounded in pedagogical frameworks like Bloom's Taxonomy.
// all publications
26 publications
Geeks are different: Exploring the relationship between engineering knowledge and categorization patterns
From Inner Diversity to Outer Crowds: How Instructions Reshape Error Structure Before Aggregation
Measuring Attention Fatigue in News and Audience Responses to Recurrent Extreme Weather Events Through Episode-wise Substitutive Process
The Multilingual Safety Divide: Cross-Lingual Disparities in LLM Climate Misinformation Detection
Temporal Diversity as a Design Principle for Within-Person Collective Intelligence
An Ensemble Method Utilising Multiple Thinking Styles That Boosts the Wisdom-of-Inner-Crowd Effect
Bridging Conformity and Deviance: A Minimal Model of Popularity, Identity, and Collective Intelligence
Development and Validity of the Japanese Version of the Divergent Association Task
Heat in Real Time, Climate in Retrospect: Mapping Seasonal Gap in Japanese Heatwave Journalism
Media Framing of Heatwaves in Japan: Misalignment between Temperature and News Coverage
Impact of the Quality and Diversity of Reference Products on Creative Activities in Online Communities
Emotion Detection on Fraudulent Accounts Using BERT and Kleinberg's Burst Algorithms in Indonesian Text
Exploring Divergence in Collective Perception to Noto Earthquake Through News Comment Data
Polarized Collaboration Benefits Knowledge Production: Empirical Analyses of the Mediating Effect of Co-production Pattern in Wikipedia Articles on Climate Change
Unsupervised Learning Methods Reveal the Complexity of User Segregation in Political Sphere in Online Communities
Introducing an 'Invisible Enemy': A Case Study of Knowledge Construction Regarding Microplastics in Japanese Wikipedia
Constitutive and Material: An Empirical Analysis of the Two Dimensions of the Communication on Microplastics in Japanese Journals
The Variant of Efforts Avoiding Strain: Successful Correction of a Scientific Discourse Related to COVID-19
Crowdsourcing Knowledge Production of COVID-19 Information on Japanese Wikipedia in the Face of Uncertainty: Empirical Analysis
An Ensemble Method Utilising Multiple Thinking Styles That Boosts the Wisdom of the Inner Crowd Effect
On an Effective and Efficient Method for Exploiting the Wisdom of the Inner Crowd
Social Influence Makes Outlier Opinions in Online Reviews Offer More Helpful Information
Media Communication in the COVID-19 Pandemic and its Issues
More Knowledge Causes a Focused Attention Deployment Pattern Leading to Lower Creative Performances
Cooperation Patterns of Members in Networks During Co-creation
Interplay of Network Structure and Neighbour Performance in User Innovation
//
$ cat quote.txt
> We do this NOT because it is easy...> ...> But because we THOUGHT it would be easy.