BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:LLM Social Simulation is a Promising Research Method - Tiancheng H
 u (University of Cambridge)
DTSTART:20250917T110000Z
DTEND:20250917T120000Z
UID:TALK236104@talks.cam.ac.uk
CONTACT:Luning Sun
DESCRIPTION:The emergence of large language models as in-silico subjects f
 or social science poses a central question: can they genuinely simulate di
 verse human behavior\, or do they merely produce plausible\, homogenized a
 rtifacts? This talk demonstrates that LLMs are powerful but imperfect simu
 lators by presenting three core contributions.\nFirst\, we establish the "
 Persona Effect\," showing that persona-prompting a 70B model captures 81% 
 of explainable variance in subjective tasks\, creating a strong baseline f
 or individual-level simulation. Second\, to address data scarcity\, we int
 roduce iNews\, a large-scale dataset of personalized affective responses t
 o news\, enriched with persona information. Finally\, we introduce SimBenc
 h\, the first large-scale benchmark for group-level simulation\, which rev
 eals the strengths and critical weaknesses of current models.\nI conclude 
 by arguing for the specialized datasets and training required to advance t
 he frontier of high-fidelity human simulation.
LOCATION:S3.04\, Simon Sainsbury Centre\, Cambridge Judge Business School
END:VEVENT
END:VCALENDAR