diff --git a/_stamina_talks/2026-10-27-baronchelli.md b/_stamina_talks/2026-10-27-baronchelli.md new file mode 100644 index 000000000..a77af7bae --- /dev/null +++ b/_stamina_talks/2026-10-27-baronchelli.md @@ -0,0 +1,10 @@ +--- +date: 2026-10-27 +title: Social Coordination in Human and AI Populations +presenters: +- name: Andrea Baronchelli + url: https://www.andreabaronchelli.com/ +affiliation: City St George's, University of London +bio: Andrea Baronchelli is Professor of Complexity Science at City St George's, University of London. He investigates how interacting agents, human and artificial, develop and inhabit shared realities. These are the cognitive, social, and economic structures through which groups coordinate, assign meaning, and create value. His approach combines large-scale data analysis, mathematical modelling, and controlled experiments with human participants and AI agents. Andrea's work has been published in journals including Nature, Science, PNAS, and Nature Human Behaviour, and has been widely covered in the press, informing public debate and helping shape policy. +--- +Social conventions are the foundation of social coordination, shaping how individuals come together to form a society. In this talk, I will present theoretical and experimental findings that demonstrate the spontaneous emergence of social norms in human groups, as well as the existence of tipping points in social convention. I will then focus on the case of populations of large language models (LLMs). As AI agents increasingly communicate using natural language, understanding how they develop conventions is crucial for interpreting and managing their collective behaviour. I will show that LLM populations can establish social conventions and highlight how collective biases can emerge even when individual agents appear unbiased. I will conclude by showing how group size affects the dynamics in a non-linear way, revealing model-dependent dynamical regimes. These findings establish group size as a key driver of multi-agent dynamics and highlight the need to consider population-level effects when deploying LLM-based systems at scale. diff --git a/_stamina_talks/2026-11-10-li.md b/_stamina_talks/2026-11-10-li.md new file mode 100644 index 000000000..40cea6fe0 --- /dev/null +++ b/_stamina_talks/2026-11-10-li.md @@ -0,0 +1,10 @@ +--- +date: 2026-11-10 +title: Emergent Misaligned Communication in Long-Horizon Multi-Agent LLM Commerce +presenters: +- name: Zeyuan Li + url: https://mitsloan.mit.edu/programs/phd/zeyuan-li +affiliation: MIT +bio: Zeyuan Li is a PhD student in Information Technology at MIT Sloan. +--- +Frontier LLM agents are increasingly being deployed to transact on behalf of separate principals, often using natural language rather than structured APIs. Much of the safety literature studies misaligned LLM behavior through adversarial-elicitation evaluations on single agents or stylized tasks. Its prevalence and structure in settings that combine long horizons, separate principals, real operational state, and inter-agent natural-language exchange remain insufficiently measured. We study 2,583 inter-agent emails from 20 one-year simulation runs of Vending-Bench Arena, a competitive vending environment spanning 13 frontier LLMs. We operationalize speech-act misalignment as emails containing false factual claims, manipulation, collusion, or threats, combining message content with ground-truth simulator state and logged reasoning traces to classify and validate such behavior. Under our primary classifier, 12.6% of emails are labeled misaligned; misalignment appears in all 20 runs and in 74.7% of individual agent-runs. Both the magnitude and the composition of this misalignment are preserved under repeated classification at different sampling temperatures and under full-pipeline replication with judges from two other frontier-model families. However, available reasoning summaries rarely confirm explicit misaligned intent. Misalignment also shows reciprocal and stress-conditioned patterns: receiving a misaligned email from a counterparty raises the odds of a misaligned reply by 1.65×, and low-inventory conditions raise them by 1.58×. Across tests of capability-asymmetric exploitation, we find no evidence that higher-capability models differentially exploit weaker counterparties, and model performance rank does not predict misalignment rates. Together, these results indicate that measurable, state-dependent misalignment can arise in competitive multi-agent environments without engineered elicitation, in patterns associated with operational scarcity and counterparty behavior rather than model capability alone. diff --git a/_stamina_talks/2026-11-24-jha.md b/_stamina_talks/2026-11-24-jha.md new file mode 100644 index 000000000..fc0b76592 --- /dev/null +++ b/_stamina_talks/2026-11-24-jha.md @@ -0,0 +1,10 @@ +--- +date: 2026-11-24 +title: 'From Solo to Social Learning: Characterizing Recursive Social Improvement in LLMs' +presenters: +- name: Kunal Jha + url: https://kjha02.github.io +affiliation: University of Washington +bio: Kunal Jha is a third-year Computer Science and Engineering Ph.D. student at the University of Washington, advised by Natasha Jaques and Max Kleiman-Weiner. Supported by the CSERF Fellowship, ARCS Scholarship, Sony Focused Research Award, and Cooperative AI Early Career Grant, his research focuses on multi-agent systems, social learning, and emergence, drawing from cognitive science, economics, and physics. He holds a B.A. in Computer Science and Philosophy from Dartmouth College and previously spent two years as a visiting researcher at MIT. Outside of research, he enjoys mixed martial arts, basketball, and drumming. +--- +Large language models (LLMs) can now improve themselves by revising the instructions they follow, and LLM agents are increasingly orchestrated to work together on complex problems. However, self-improvement methods typically optimize one system at a time, and multi-agent frameworks often have every model work toward a shared goal. We ask a different question. When each agent pursues its own reward, can self-improving LLMs learn from one another well enough to improve the whole population? We call this capability recursive social improvement. We study populations that revise skill files and choose whether, when, and whom to copy from. Independent search, learning from peers, and acting all share one token budget. In controlled environments, established social-learning algorithms benefit from peers, but three LLMs do not. They earn less reward per token than solo learners, and explore too narrowly or run out of tokens before acting. We then let the models write and revise their own skills. Observing peers changes how they improve, helping one model find useful skills sooner and another spend less on private search. Neither, however, outperforms independent learners at the same cost. Skills are copied, revised, and passed on, so one discovery can seed further search. Yet these exchanges concentrate the population around fewer independent discoveries. Together, these results show that LLMs can make learning more efficient by copying from peers, but not yet more effective. diff --git a/_stamina_talks/2026-12-08-michailidis.md b/_stamina_talks/2026-12-08-michailidis.md new file mode 100644 index 000000000..a5aa3ef06 --- /dev/null +++ b/_stamina_talks/2026-12-08-michailidis.md @@ -0,0 +1,10 @@ +--- +date: 2026-12-08 +title: 'From Caricature to Collective Artifact: How Persona Construction Drives Polarization in Generative Social Simulations' +presenters: +- name: Dimitris Michailidis + url: https://dimichai.github.io/ +affiliation: University of Amsterdam +bio: Dimitris Michailidis is a postdoctoral researcher at the University of Amsterdam, studying the effects of link-recommendation algorithms on social networks through simulations and online experiments, and working on generative social simulation with LLM-based agents. +--- +Generative Social Simulation (GSS) leverages Large Language Models (LLMs) to create sophisticated personas that define agents with demographics, emotions, and political opinions. While prior work shows that personas compress outputs into stereotypical caricatures, there is no comprehensive study of how these caricatures evolve into collective artifacts when they interact. Using a Twitter-like GSS platform with personas derived from the American National Elections Study (ANES), we ablate three layers of partisan identity: affective polarization markers, party identification, and vote choice, while retaining demographics, ideological positioning and important issues. We find that collective outcomes depend heavily on how survey responses are translated into personas. Removing a few explicitly partisan statements moves the network from near-complete segregation to cross-party ties approaching those of real Twitter networks. We argue that matching one outcome is not evidence for behavioral validity and point to salience as a possible explanation. Surveys ask for a position on every question regardless of its importance to the respondent, and persona construction turns each answer into a direct behavioral instruction. diff --git a/stamina/css/plain-style.css b/stamina/css/plain-style.css index 9a863c9d5..933778d50 100644 --- a/stamina/css/plain-style.css +++ b/stamina/css/plain-style.css @@ -225,3 +225,113 @@ font-size: 18pt; stroke: none; } +/* ============================================================ + Light/dark mode + is set by the script in index.html's ; + the navbar button (#theme-toggle) flips it. + ============================================================ */ + +.theme-toggle { + display: block; + padding: 15px; + line-height: 20px; + background: none; + border: 0; + color: #777; + cursor: pointer; +} + +.theme-toggle:hover, +.theme-toggle:focus { + color: #333; +} + +.theme-toggle svg { + display: block; + width: 20px; + height: 20px; +} + +.theme-toggle .theme-toggle-sun { display: none; } + +@media (max-width: 767px) { + .theme-toggle { padding: 10px 15px; } +} + +html[data-theme="dark"] { color-scheme: dark; } + +html[data-theme="dark"] .theme-toggle { color: #aab4bf; } +html[data-theme="dark"] .theme-toggle:hover, +html[data-theme="dark"] .theme-toggle:focus { color: #fff; } +html[data-theme="dark"] .theme-toggle-moon { display: none; } +html[data-theme="dark"] .theme-toggle-sun { display: block; } + +html[data-theme="dark"] body { + background-color: #15191e; + color: #d5dbe2; +} + +html[data-theme="dark"] a { color: #7fb8ea; } +html[data-theme="dark"] a:hover, +html[data-theme="dark"] a:focus { color: #a9d0f3; } + +html[data-theme="dark"] .btn-primary, +html[data-theme="dark"] .btn-info, +html[data-theme="dark"] .btn-primary:hover, +html[data-theme="dark"] .btn-info:hover, +html[data-theme="dark"] .btn-primary:focus, +html[data-theme="dark"] .btn-info:focus { color: #fff; } + +html[data-theme="dark"] hr { border-top-color: #2c343d; } + +html[data-theme="dark"] .navbar-default { + background-color: #1c2229; + border-color: #2c343d; +} + +html[data-theme="dark"] .navbar-default .navbar-nav > li > a { color: #aab4bf; } +html[data-theme="dark"] .navbar-default .navbar-nav > li > a:hover, +html[data-theme="dark"] .navbar-default .navbar-nav > li > a:focus { color: #fff; } + +html[data-theme="dark"] .navbar-default .navbar-nav > .active > a, +html[data-theme="dark"] .navbar-default .navbar-nav > .active > a:hover, +html[data-theme="dark"] .navbar-default .navbar-nav > .active > a:focus { + background-color: #2a323b; + color: #fff; +} + +html[data-theme="dark"] .navbar-default .navbar-toggle { border-color: #3a444f; } +html[data-theme="dark"] .navbar-default .navbar-toggle:hover, +html[data-theme="dark"] .navbar-default .navbar-toggle:focus { background-color: #2a323b; } +html[data-theme="dark"] .navbar-default .navbar-toggle .icon-bar { background-color: #aab4bf; } +html[data-theme="dark"] .navbar-default .navbar-collapse { border-color: #2c343d; } + +html[data-theme="dark"] .box { + background-color: #1c2229; + border-color: #3a444f; +} + +/* supporter logos are drawn for a white page */ +html[data-theme="dark"] #partners img { + background: #fff; + padding: 6px; +} + +html[data-theme="dark"] .bibbase_group, +html[data-theme="dark"] span.bibbase_paper_titleauthoryear, +html[data-theme="dark"] span.bibbase_paper_title, +html[data-theme="dark"] span.bibbase_paper_author, +html[data-theme="dark"] span.bibbase_paper_year { color: inherit; } + +html[data-theme="dark"] .stam-diagram { + --stam-ink: #d5dbe2; + --stam-muted: #9aa7b4; + --stam-line: #36414c; + --stam-tech: #6fb0d8; + --stam-sim: #4fc4b5; + --stam-ai: #d6a25f; +} + +html[data-theme="dark"] .stam-chip { background: #1f262e; } +html[data-theme="dark"] .stam-venn-lobes circle { mix-blend-mode: screen; } +html[data-theme="dark"] .stam-venn-core-shape { fill-opacity: 0.12; } diff --git a/stamina/index.html b/stamina/index.html index 8b06dbb24..b4daec7fb 100644 --- a/stamina/index.html +++ b/stamina/index.html @@ -12,6 +12,19 @@ STAMINA Working Group + + + @@ -52,8 +65,20 @@
  • About
  • Upcoming Talks
  • Past Talks
    • +
  • Community Events
  • Organizers
  • ComplexDataLab
    • +
  • + +
    • @@ -233,6 +258,33 @@

    {{ talk_term }}

    + +
    +
    +
    +
    +
    +
    + +

    Community Events

    +

    Workshops in the Social Simulation with LLMs series:

    + + +
    + +
    +
    +
    +
    +
    + +