AI safety researcherIstanbul, Türkiye

Sofiia Lobanova

I study what models do under pressure, not only what they say. Behavioral evaluations, model incentives, and the human systems around them.

"The poetry of earth is never dead."John Keats
Model incentivesBehavioral evalsGovernance

01

Status / now

Currently, roughly.

  • AI Revealed Preferences at SPAR
  • External incentives at MARS V
  • Structured transparency and decision making at TALOS

02

Selected work

Questions with
some mileage.

Full research index
  1. 01

    Research paper

    AI Revealed Preferences

    A behavioral framework for measuring what AI systems choose under incentive pressure rather than what they report preferring.

    2026
    • model incentives
    • evaluation
    • deception
  2. 02

    Ongoing / MARS V

    Aligning AIs Via External Incentives

    Extending revealed-preference methods to isolate how external incentives change model choices.

    2026
    • alignment
    • incentives
    • behavioral evals
  3. 03

    TALOS Fellowship / ongoing

    Structured transparency and AI in decision making

    Research on making safety evaluations easier to inspect and on the role of AI in social decision-making processes.

    2026
    • structured transparency
    • safety evaluations
    • decision making
  4. 04

    Forecasting project

    AI Treaty Momentum Index

    An index for tracking political momentum toward international AI coordination.

    2025
    • forecasting
    • governance

03

Recent writing

Notes from the
messy middle.

All writing

Nothing published here yet.

Research and publications

Elsewhere in my brain

Other interests

  • urban systems
  • human behavior
  • low-resource languages
  • graphs
  • medical AI
  • model reliability