Sessions
What’s Coming
Assistant Professor, International Relations and Korean Studies, Leiden University
From Pixels to Patterns: Vision-Language OCR and LLM-Based Text Analysis
Drawing on ongoing research, Steven will walk through a practical pipeline for turning images of text into analyzable data — first using multi-modal and vision-language models to perform OCR, then applying LLM-based reading and classification to extract structure and meaning. The session will cover when vision-language OCR outperforms traditional approaches, how to design prompts and validation steps for downstream text analysis, and what these methods open up for researchers working with archival, multilingual, or otherwise difficult corpora.
Register via ZoomAssociate Professor, Department of Economics and Department of International Affairs, University of Georgia
AI Is Not an Algorithm: What That Means for Research Practice
Generative AI is not an algorithm. Identical inputs produce slightly different outputs across sessions, which makes AI supervision in research qualitatively different from (and harder than) code review. It also reframes what gets called “AI slop” as miscalibration on a consistency–creativity dial: AI was either too creative where faithful protocol execution was needed, or not creative enough to add anything beyond the obvious. The real research skill is deciding, task by task, where on that dial you want the tool to operate and configuring it accordingly. Laura will present this framework through four concrete research experiences: a 19-volume biographical OCR pipeline where AI would not consistently self-apply its own quality protocols; a two-hour methods discussion with a PhD student reprocessed into a working memo; a dictation-plus-vault-grounding-plus-Claude-Cowork loop in Obsidian that neither LLM Wiki patterns nor standard agentic workflows quite capture; and an R standard-error episode where an unusual package choice produced the wrong statistical conclusion. The talk complements the opening DAAF session: where DAAF formalizes the algorithmic end of the dial for data analysis, Laura argues for the fuller taxonomy research work can also benefit from.
Register via ZoomUniversity of Southern California
Tracing AI Assistance and AI Agents in Survey Research
Identifying traces of AI in survey research, and using AI agents to test vulnerabilities of surveys before fielding experiment.
Register via ZoomToronto Metropolitan University
Ethical, Effective, and Transparent Workflows: How to Set Clear Guidelines Surrounding AI Usage for Academic Research
AI poses existential questions for professionals in all fields to a large extent because clear rules and guidelines surrounding ethical AI have not been created yet. How do we set such rules and guidelines? Concerns and controversies about AI usage are rooted in ambiguities about the relationship between human principals and AI agents, and the key is to encourage transparent usage that clarifies this principal-agent relationship. Like data-manipulation computational tools such as R or Stata, AI agents are research tools that require training to use ethically and effectively by human researchers who direct the research process. By proactively establishing this principal-agent relationship through best practices devised through transparent experimentation, academic institutions can effectively mitigate the risks as well as seize opportunities offered by AI tools.
Register via ZoomDina Pisareva
Nazarbayev University
AI as a Cognitive Partner in Research and Teaching
What changes when AI is treated not as a shortcut or a tool, but as a genuine cognitive partner — something scholars and students think with rather than merely through? Dina is building a practice around AI-augmented social science and developing some of the first AI-integrated methods courses in the field. In this session, she will share what that partnership looks like across research workflows and the classroom: how it reshapes the questions we ask, what it demands of graduate training, and why she sees the terrain as wide open, with most of the interesting questions still unasked.
Register via ZoomStatistical Horizons
Vibe Coding Then and Now: Building Statistical Packages with Humans, Claude Cowork, and Codex
Abstract coming soon.
Register via Zoom