Aspen Hopkins

First author of Contra Labs' flagship paper, The Human Creativity Benchmark (Jun 2026), listed with a dual Contra (New York) and MIT affiliation. She is the most credentialed ML researcher publicly attached to Contra Labs, and she may be the unnamed 'Head of Research', which is unconfirmed.

Background

PhD student in Aleksander Madry's lab and the Center for Deployable Machine Learning at MIT (CSAIL profile). She co-authored 'Emergent World Representations' (ICLR, the Othello-GPT paper), 'Designing Data: Proactive Data Collection and Iteration for ML' (2023) and 'Machine Learning Practices Outside Big Tech' (AIES 2021), plus 2025–26 interpretability work (steering, sparse autoencoders). Her early work includes VisuaLint, a visualisation-annotation paper, which shows an HCI and visual-design strand.

What they run now

Methodology behind the Human Creativity Benchmark: phase decomposition (ideation, mockup, refinement) and treating evaluator disagreement as signal. The benchmark separates convergence on professional standards from legitimate divergence in taste.

Career

2026 – presentResearcher; first author of HCB, Contra (Contra Labs) — exact title not public
—PhD researcher, MIT (Madry Lab / CDML)
—Massachusetts Institute of Technology, PhD (in progress or completed; not confirmed)

On the record

Disagreement as signal · 2026-06

In creative evaluation, rater disagreement is informative: agreement is high on prompt adherence and low on visual appeal, separating shared standards from taste. source

Phase-specific rankings · 2026-06

Model rankings invert across ideation, mockup and refinement, so a single leaderboard hides where models fail. source

Connections

Alexandria Minetti — co-author (HCB)
Allison Nulty — co-author (HCB, TASTE)
Anoop Pakki — co-author (HCB)
Angad Singh — co-author (HCB)
Ben Huffman — Contra CEO

Why it matters here

The HCB methodology (phase decomposition, disagreement-as-signal) is the most copyable and defensible part of Contra Labs. She is the person who can explain what actually works in rater design for creative tasks. An academic collaboration or advisory relationship with someone like her would give a new entrant research credibility.

How to reach

Academic channels: the MIT CSAIL profile and the corresponding address on the arXiv paper. Ask a methodology question, for example how disagreement was modelled or how evaluators were selected across 13 countries. Speculative.

What we could not establish

Her exact Contra title, and whether she is employee, contractor or collaborator
Whether she is the unnamed 'Head of Research'
PhD status
Personal site: aspenhopkins.com is NOT hers (it is an unrelated domain)
X and LinkedIn not verified

Background

What they run now

Career

On the record

Connections

Sources

Ben Huffman

Gajus Kuizinas

Alexandria Minetti