D01-06PM: Artificial Intelligence: Theory and Practice

The course offers an intensive exploration of AI technologies and their applications in social science research. This cutting-edge course bridges theoretical foundations with practical implementation, enabling researchers to leverage AI tools effectively in their scholarly work. Participants will gain a comprehensive understanding of AI architectures and master the technical fundamentals of transformer models. The course provides hands-on experience with leading AI models (including GPT-4, Claude, and open-weight alternatives), emphasizing their comparative advantages for different research tasks. Through practical sessions, participants will learn API implementation, prompt engineering, and ethical considerations while applying these tools to real research scenarios, including interview analysis, qualitative coding, and large-scale text analysis. The course culminates in a project workshop where participants develop AI-enhanced research designs relevant to their disciplines. Ideal for researchers seeking to incorporate AI methodologies into their research workflow while maintaining methodological rigor and ethical standards.

The “AI: Theory and Practice” summer course offers an intensive exploration of artificial intelligence technologies and their applications in social science research. Combining foundational theory with practical implementation, this course equips participants with the tools to integrate AI into their research workflows effectively.

Day 1

Begins by exploring the distinctions between AI, Generative AI (Gen AI), and Artificial General Intelligence (AGI), followed by an introduction to Large Language Models (LLMs). Participants will learn about transformers, self-attention mechanisms, word embeddings, and tokenization, as well as how these models are trained to perform reasoning tasks, setting the stage for deeper engagement throughout the week.

1 – AI vs Gen AI (and AGI)

• What’s the difference between AI and Generative AI? What about AGI?

2 – What are LLMs?

• What are “transformers”?
• How does the “self-attention” mechanism work? What is “word embedding”? How do tokens work? Etc. 
• Teaching an LLM to reason: what are “reasoning models”? 

Day 2

Shifts focus to access, transparency, and ethics in AI. Discussions will center around open-source versus proprietary models and the ethical implications of AI, including issues such as data privacy, the “right to be forgotten,” and the importance of guardrails and guidelines to ensure responsible use. Participants will also compare leading AI models—such as GPT, Claude, Gemini, Llama, and others—gaining insights into their strengths, limitations, and practical applications for various research tasks. Hands-on sessions will introduce API functionalities, preparing participants to use these tools effectively in their own projects.

3 – Access and Transparency:

• Open-source/Open-weight models vs. Proprietary/Black-box models
• Ethics and Transparency:
  • The “right to be forgotten”
  • LLMs and sensitive data 
  • Guardrails and Guidelines: Implications and “Red teaming”

4 – Using LLMs:

• GPT (4o, 4o mini, o1), Claude (3.5 Sonnet Preview), Gemini (NotebookLM) vs. Llama, Mistral (Large and Small), Qwen & Deepseek, Falcon
• How to access and use these models?
• How do they work, what are the differences between them, when to use them, etc. 
• API: What is it? How does it work? 

Days 3 and 4

Are dedicated to hands-on practice with LLMs, focusing on prompt engineering and practical applications. On Day 3, participants will learn how to craft effective prompts, employing techniques like zero-shot and few-shot prompting, chain-of-thought prompting, and API querying using R. Day 4 emphasizes real-world applications, such as analyzing interviews, coding qualitative data, speech analysis, and working with synthetic data generation through tools like Silicon Samples. These sessions are designed to empower researchers with practical skills to address their specific research needs.

5 – Hands-on 1:

• GPT (4o and o1), Gemini (NotebookLM), Mistral (Large and Small), Deepseek (reasoning/”DeepThink”).
• Prompts: How to write a “good” prompt? How do prompts work? Etc.
  • Zero-shot and Few-shot prompting, Chain-of-thought (CoT) prompting, etc. 
• Querying with APIs (Using R). 

6 – Hands-on 2: 

• Using LLMs to work with: 
  • Interviews; 
  • Coding; 
  • Speeches; 
  • Silicon Samples

Day 5

The course culminates with a project workshop where participants present research designs that integrate AI methodologies. This collaborative session allows attendees to showcase their understanding, receive constructive feedback, and refine their approaches. By the end of the course, participants will be equipped to leverage AI technologies in their research, maintaining methodological rigor and ethical standards while embracing cutting-edge advancements in the field.

7 – Project Workshop: 

• Individual presentation of research projects you could/will use AI and how, etc. 

Suggested readings: 

1 – Argyle LP, Busby EC, Fulda N, Gubler JR, Rytting C, Wingate D. Out of One, Many: Using Language Models to Simulate Human Samples. Political Analysis. 2023;31(3):337-351. doi:10.1017/pan.2023.2

2 – Bisbee J, Clinton JD, Dorff C, Kenkel B, Larson JM. Synthetic Replacements for Human Survey Data? The Perils of Large Language Models. Political Analysis. 2024;32(4):401-416. doi:10.1017/pan.2024.5

3 –  Ashokkumar, A., Hewitt, L., Ghezae, I., & Willer, R. (2024). Predicting results of social science experiments using large language models. Work. Pap., New York Univ., New York, NY.

This description is subject to change at the discretion of the Instructor

2 Credits

For completion of all work before and during the course, as outlined by the Instructor, and 90% participation and attendance of the course.

2 Additional Credits

Course specific extra assignments. These can include submitting assignments before the course, daily assignments, and/or a final assignment to be completed after the course as decided by the Instructor.