BEGIN:VCALENDAR VERSION:2.0 X-WR-CALNAME:EventsCalendar PRODID:-//hacksw/handcal//NONSGML v1.0//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York LAST-MODIFIED:20240422T053451Z TZURL:https://www.tzurl.org/zoneinfo-outlook/America/New_York X-LIC-LOCATION:America/New_York BEGIN:DAYLIGHT TZNAME:EDT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZNAME:EST TZOFFSETFROM:-0400 TZOFFSETTO:-0500 DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT CATEGORIES:College of Engineering,Thesis/Dissertations DESCRIPTION:Thesis Advisor: Dr. Lance Fiondella - Electrical and Computer E ngineering Committee Members: Dr. Gokhan Kul - Computer & Information Sci enceDr. Long Jiao - Computer & Information Science Abstract: Past studies indicate that nonviolent resistance has achieved higher success rates in g lobal political transformations than armed conflict. Agent-based modeling has been employed to simulate these movements, but prior models rely on ri gid numerical thresholds and oversimplify agent interactions. This limits the capacity of simulations to analyze the strategic reasoning underlying human leadership. To address these limitations, this thesis presents a hyb rid generative agent-based model that integrates large language models to guide the decision-making of activist agents while keeping other agent cla sses rule-based for computational feasibility.  The proposed model introd uces three methodological enhancements. First, LLM-guided decision-making replaces rule-based activist movement. To prevent behavioral drift, a four -component structured system prompt anchors agent identity through role, p ersona, domain-specific world knowledge, and scenario context. Second, eig ht interdependent institutional support pillars replace a single abstract pillar type. Third, a linguistic transformation layer translates continuou s numeric states of agents into semantic social descriptors, allowing acti vists to perform reasoning tasks, such as identifying and mobilizing highl y aggrieved civilians. We drive activist agents with three open-weight Lar ge Language Models (Llama-3.1-8B-Instruct, Ministral-8B-Instruct, Qwen3-8B ) under Zero-Shot and Chain-of-Thought (CoT) prompting, yielding six exper imental configurations. Macro-level validation against the Nonviolent and Violent Campaigns and Outcomes (NAVCO) 1.2 dataset suggests all configurat ions approximate historical campaign success rates at low participation le vels. Simulations indicate that high activist coordination and low fatalit y rates predict campaign success. However, micro-level analysis shows that CoT reasoning is required for behavioral realism. While Zero-Shot agents default to generic protests, CoT-enabled agents semantically evaluate inst itutional vulnerabilities and execute tactics aligned with pillar suscepti bility: protest and persuasion, noncooperation, and intervention. These ag ents prioritize institutional and civilian outreach, actively avoiding con frontation with security forces. Micro-level validation against the Global Nonviolent Action Database (GNAD) suggests these CoT configurations appro ximate the historical prevalence of diverse tactical behaviors. While diff erent foundational LLMs capture distinct aspects of real-world resistance, this study constitutes a simulated instance of Gandhian strategic princip les. Ultimately, this thesis identifies CoT-enabled agents as a promising method for analyzing how specific leadership paradigms operate within dive rse societies. For further information, please contact Dr. Lance Fiondella at lfiondella@umassd.edu. \nEvent page: https://www.umassd.edu/events/cm s/8-25-26-a-hybrid-generative-agent-based-model-of-nonviolent-resistance.p hp\nEvent link: https://teams.microsoft.com/meet/252984977868302?p=H2LcD3U FCOp6BL31w6 X-ALT-DESC;FMTTYPE=text/html:

Thesis Advisor: Dr. Lance Fiond ella - Electrical and Computer Engineering

\n

Committee Members: 
Dr. Gokhan Kul - Computer & Information Science
Dr. Long Jiao - Co mputer & Information Science

\n

Abstract:

\n

Past studies indica te that nonviolent resistance has achieved higher success rates in global political transformations than armed conflict. Agent-based modeling has be en employed to simulate these movements\, but prior models rely on rigid n umerical thresholds and oversimplify agent interactions. This limits the c apacity of simulations to analyze the strategic reasoning underlying human leadership. To address these limitations\, this thesis presents a hybrid generative agent-based model that integrates large language models to guid e the decision-making of activist agents while keeping other agent classes rule-based for computational feasibility.  
The proposed model intr oduces three methodological enhancements. First\, LLM-guided decision-maki ng replaces rule-based activist movement. To prevent behavioral drift\, a four-component structured system prompt anchors agent identity through rol e\, persona\, domain-specific world knowledge\, and scenario context. Seco nd\, eight interdependent institutional support pillars replace a single a bstract pillar type. Third\, a linguistic transformation layer translates continuous numeric states of agents into semantic social descriptors\, all owing activists to perform reasoning tasks\, such as identifying and mobil izing highly aggrieved civilians. We drive activist agents with three open -weight Large Language Models (Llama-3.1-8B-Instruct\, Ministral-8B-Instru ct\, Qwen3-8B) under Zero-Shot and Chain-of-Thought (CoT) prompting\, yiel ding six experimental configurations.

\n

Macro-level validation again st the Nonviolent and Violent Campaigns and Outcomes (NAVCO) 1.2 dataset s uggests all configurations approximate historical campaign success rates a t low participation levels. Simulations indicate that high activist coordi nation and low fatality rates predict campaign success. However\, micro-le vel analysis shows that CoT reasoning is required for behavioral realism. While Zero-Shot agents default to generic protests\, CoT-enabled agents se mantically evaluate institutional vulnerabilities and execute tactics alig ned with pillar susceptibility: protest and persuasion\, noncooperation\, and intervention. These agents prioritize institutional and civilian outre ach\, actively avoiding confrontation with security forces. Micro-level va lidation against the Global Nonviolent Action Database (GNAD) suggests the se CoT configurations approximate the historical prevalence of diverse tac tical behaviors. While different foundational LLMs capture distinct aspect s of real-world resistance\, this study constitutes a simulated instance o f Gandhian strategic principles. Ultimately\, this thesis identifies CoT-e nabled agents as a promising method for analyzing how specific leadership paradigms operate within diverse societies.

\n

For further informatio n\, please contact Dr. Lance Fiondella at lfiondella@umassd.edu. 

E vent page: https://www.umassd .edu/events/cms/8-25-26-a-hybrid-generative-agent-based-model-of-nonviolen t-resistance.php
Event link: https://teams.microsoft.com/meet /252984977868302?p=H2LcD3UFCOp6BL31w6

DTSTAMP:20260731T155843 DTSTART;TZID=America/New_York:20260825T120000 DTEND;TZID=America/New_York:20260825T130000 LOCATION:Microsoft Teams SUMMARY;LANGUAGE=en-us:A Hybrid Generative Agent-Based Model of Nonviolent Resistance with Large Language Model-Enabled Activist UID:4ed8091a29710358974fc104a9a8f3c4@www.umassd.edu END:VEVENT END:VCALENDAR