About me:

I’m an assistant professor in the School of Operations Research and Information Engineering at Cornell and a member of the Computer Science field. I think about new ways of doing democracy (e.g., citizens’ assemblies), about how to fairly allocate resources (e.g., capacity to host refugees), and developing AI for heterogeneous users (e.g., generative social choice).

Before joining Cornell, I did my undergrad at Saarland University in Germany, got my PhD in computer science at Carnegie Mellon University, and did postdocs at Harvard, the Simons Laufer Mathematical Sciences Institute (formerly: MSRI), and UC Berkeley. You can find my CV here.

Contact:

E-Mail:
mail-removethispart@paulgoelz.de
Address/Office:
Paul Goelz
Rhodes 221
136 Hoy Rd
Ithaca, NY 14850, USA

I enjoy getting to know new people and chatting about research. Whether you're a researcher, a student interested in research, a practitioner, or just curious, please don’t hesitate to reach out!

Papers:

W2
Distortion of AI alignment: Does preference optimization optimize for preferences?.
Paul Gölz, Nika Haghtalab, and Kunhe Yang.
In submission.
W1
Putting fair division on the map.
In submission.
C15
Generative social choice.
C14
Monotone randomized apportionment.
J5
In this apportionment lottery, the House always wins.
Paul Gölz, Dominik Peters, and Ariel Procaccia.
Operations Research (2025). Supersedes C12.
J4
Dynamic placement in refugee resettlement.
Operations Research (2024) Communications of the ACM research highlight. Supersedes C8.
C13
Now we’re talking: Better deliberation groups through submodular optimization.
Jake Barrett, Kobi Gal, Paul Gölz, Rose Hong, and Ariel Procaccia.
O1
Mini-public selection: Ask what randomness can do for you.
Bailey Flanigan, Paul Gölz, and Ariel Procaccia.
Harvard Kennedy School Ash Center Policy Briefs (2023).
T2
Social choice for social good: Proposals for democratic innovation from computer science.
Paul Gölz.
Ph.D. thesis, 2022.
J3
Approval-based apportionment.
Markus Brill, Paul Gölz, Dominik Peters, Ulrike Schmidt-Kraepelin, and Kai Wilker.
Mathematical Programming (2022). Supersedes C6.
C12
In this apportionment lottery, the House always wins.
Paul Gölz, Dominik Peters, and Ariel Procaccia.
EC 2022. Superseded by J5.
C11
Fair allocations for smoothed utilities.
C10
Envy-free and Pareto-optimal allocations for agents with asymmetric random valuations.
Yushi Bai and Paul Gölz.
J2
Fair algorithms for selecting citizens’ assemblies.
Nature (2021).
C9
Incentive-compatible kidney exchange in a slightly semi-random model.
Avrim Blum and Paul Gölz.
C8
Dynamic placement in refugee resettlement.
EC 2021. Superseded by J4.
J1
The fluid mechanics of liquid democracy.
ACM Transactions on Economics and Computation (2021). Supersedes C2.
C7
Neutralizing self-selection bias in sampling for sortition.
C6
Approval-based apportionment.
Markus Brill, Paul Gölz, Dominik Peters, Ulrike Schmidt-Kraepelin, and Kai Wilker.
AAAI 2020. Superseded by J3.
C5
Paradoxes in fair machine learning.
Paul Gölz, Anson Kahng, and Ariel Procaccia.
NeurIPS 2019 Spotlight presentation (2.5% of submissions).
C4
No stratification without representation.
Gerdus Benadè, Paul Gölz, and Ariel Procaccia.
C3
Migration as submodular optimization.
Paul Gölz and Ariel Procaccia.
C2
The fluid mechanics of liquid democracy.
WINE 2018. Superseded by J1.
C1
Synthesis in distributed environments.
Bernd Finkbeiner and Paul Gölz.
T1
Synthesis for Petri games with one system player.
Paul Gölz.
Undergraduate thesis, 2017.

News:

May: New preprint on the distortion of AI alignment: we show that standard methods like RLHF and DPO can fail at satisfying even the average user, while Nash Learning from Human Feedback achieves robust guarantees.
April: I'm co-organizing the 2nd Workshop on New Directions in Social Choice at EC'25 (July 10 at Stanford). Submit your work and join us!
January: Bailey Flanigan and I received a $40,000 Structural Democracy Fellowship, funded by the Crankstart Foundation, to update Panelot, our not-for-profit website allowing practitioners to draw citizens' assemblies.
January '24: Our research project on generative social choice has been part of OpenAI's grant program for Democratic Inputs to AI ("10 teams out of nearly 1000 applicants" awarded). Check out the paper here!
October '23: My thesis received an honorable mention at the George B. Dantzig dissertation award.
July '23: My thesis was recognized with an honorable mention at the ACM SIGecom Doctoral Dissertation Award.
April '23: I'm incredibly excited to join Cornell ORIE as an assistant professor in July 2024!
Show older
August '22: I got my PhD! You can find my thesis here.
August '21: After a lot of work, Nature published our new paper on citizens' assemblies/sortition!
April '21: I received a JPMorgan Chase AI Research fellowship!
August '20: This Fall, Anson and I are founding a new MD4SG working group on novel forms of Civic Participation.
April '20: My team finished 41st out of 10,724 at the Google Hash Code programming competition. Thanks to Alex, Andrii, and Da Qi!

People:

2nd year ORIE PhD (joint with David Williamson).
Other students I've been recently working with include Gabriel Morete de Azevedo (ORIE 1st year), Shayan Ranjbarzadeh (ORIE 2nd year), Sonja Kraiczy (Oxford), Markus Utke (TU Eindhoven), and Kunhe Yang (UC Berkeley).

Online Tools:

Panelot.org (serious).
Not-for-profit website for selecting citizens' assemblies using our state-of-the-art selection algorithms.
Check if your place is unusually sunny/cloudy/rainy/dry this year.

Teaching:

ORIE 4160/5160: Optimized Democracy.
Undergraduate/Master's elective covering mathematical and computational approaches to democracy.
Cornell, Spring 2025.
ORIE 6300: Mathematical Programming.
PhD course covering linear, convex, and nonconvex optimization.
Cornell, Fall 2024.