Sarah Vahlkamp
Research NotesTrajectoriesPresented Work

Thinking is most interesting before it settles.

Thinking develops through interaction with people, tools, environments, and increasingly, AI. My research examines how those interactions reshape the problems people think they're solving. I'm interested in how understanding evolves through conversation, revision, and sustained engagement before it settles into something that looks like an answer.

Current Lines of Inquiry

How does interaction with AI reshape the problems people thing they're solving?

People often assume that AI influences solutions. I'm interested in the earlier stage: how interaction with AI reshapes goals, constraints, and problem definititions themselves.

How do goals and problem representations evolve over extended human-AI collaboration?

Rather than treating prompts as isolated events, I study how understanding develops across conversations and how AI can influence the trajectory of thought over time.

How should we evaluate human-AI systems rather than AI systems alone?

Many important outcomes emerge from the interaction between people and AI. I'm interested in methods for evaluating those interactions in realistic settings

When does AI support judgment, and when does it subtly redirect it?

As AI becomes embedded in consequential decisions, understanding its effects on attention, confidence, and reasoning becomes increasingly important.

More details...

What is generative drift?

When people interact with AI, the problem they are working on isn't static.
It evolves through the interaction itself.

A single suggestion may seem small. A reframing. An added constraint. A new direction.
But across multiple exchanges, these small shifts can accumulate.

Over time, the problem can become meaningfully different from where it began.

I refer to this process as generative drift.

The gradual reshaping of problem interactions through iterative human-AI interaction.

This matters because these shifts are often subtle. They may not be experienced as change in the moment, even as they reflect on the trajectory of the work.

Understanding generative drift helps us examine how AI shapes the problems we think we are solving.