About Me
Brief Bio
I am a PhD student in Information Science at the University of Michigan, advised by Ceren Budak and Eric Gilbert. Before starting my PhD, I did data science in advertising/ad-tech and grew up in the NY area.
Research Interests
I am broadly interested in human-AI interactions at the collective level, with a focus on alignment challenges that emerge at population scales. Concretely, this has two research directions—one about new systems and one about current systems.
System building → future AI systems. I build multi-agent systems that help users engage with diverse viewpoints through AI. These systems have two components: (1) representing different perspectives through steerable AI agents, and (2) rendering these perspectives as practical decision aids. I am particularly interested in cases where simulated AI viewpoints can improve complex decision-making, like in direct democracy. My Plurals ("A system for guiding LLMs via simulated social ensembles") paper/library is a good example of one engine, where I am now doing follow-up RCTs with it. I am currently evaluating a second engine, "The As-If Machine", for increasing action towards long-term risks. My focus is on designing scalable systems and conducting empirical evaluations to ensure they're useful.
Impact surfacing → current AI systems. Designing experiments to make visible the non-obvious effects of human-AI interactions beyond immediate (single human, single AI) interactions. My creativity paper that surfaced how AI ideas affect the evolution of human ideas is a good example. I am also interested in designing experiments that measure alignment-relevant capabilities at collective scales—for example, whether large language models learn deep values or shallow preferences of humans when given choice data.
These two directions have tight connections with:
Pluralistic alignment: If a system can align to diverse viewpoints, it can provision those as helpful interventions.
Computational social science: I draw on social science theories for Direction 1 and I draw on social science methods for Direction 2.
Collective intelligence: The motivation for both directions is strongly rooted in CI.
“Evolution of a Typical Internet Subculture”
Non-Research Interests
Outside of research, I enjoy making mathematical art. Mathematical art (different from generative art in the AI sense) involves writing programs that produce art. In my case, I like to combine different statistical distributions (different randomness) with color theory. Here is my art portfolio. If anyone would like to collaborate on creative coding projects, feel free to reach out!
I write and read haikus. I wrote a blog post about why I like haikus. I have been playing guitar since I was 13. My favorite genres are shoegaze and jazz. My first job was a guitar teacher. My favorite book genres are (A) those that are a history of a single idea (e.g., information or novelty) and (B) short stories.
Contact
email: jashkina [at] umich [dot] edu