Week 1: Gen AI for Sociology

Course Overview

  • goals, schedule, grading

Discussion

  • introductions
  • how do you use AI (LLMs)?
  • what are your interests?
  • is it stopping your from learning?
  • do you ever stop to question yourself?
  • how does it feel to be a grad. student in this moment?

AI “Policy”

  • Here’s my thoughts…

AI is booming

AI is booming (part 2)

AI is booming (part 3)

Where do we place all of this stuff??

Situating the field

  • Text as data
  • Machine learning
  • Algorithmic bias
  • Simulation

Text as data

  • Using text as a source of data for social science research
  • Coding according to rules and textual features

Some early examples

Some early examples

Some early examples

Some modern-day renewals

Machine learning

Some recent examples

Some recent examples

Some recent examples

Algorithmic bias

Some recent examples

Some recent examples

Some recent examples

Simulation

Some early examples

Some early examples

Some early examples

Some more recent examples

So how do LLMs bring any of this together

  • LLMs both tool and object of inquiry (similar to ML)
  • LLMs can do (all? most?) of the above tasks

What am I talking about?

LLMs come from text as data

  • trained on massive corpora of text
  • learn patterns in text
  • generate text
  • can follow codebooks

LLMs are also machine learning

  • trained using machine learning techniques
  • can be fine-tuned for specific tasks
  • can be used as components in larger ML systems

LLMs can also be used in simulations

  • agents can be powered by LLMs
  • can simulate human-like behavior
  • can generate realistic scenarios

LLMs can be objects of study

  • how do they reflect biases in training data?
  • how do they impact society?
  • what are the ethical implications?

This creates massive opportunities

  • We can annotate text at scale
  • We can build complex models of social phenomena
  • We can simulate complex social systems
  • We can study the impact of AI on society

But also massive challenges

  • Ethical considerations
  • Bias and fairness
  • Interpretability
  • Reproducibility