Early-career Carnegie Mellon researcher auditing economic misalignment and preference elicitation in LLM-based agents
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
This axis lacks sufficient data in the current snapshot, so no score is assigned. It would measure publication volume, venue quality, and consistency of research production over time; existing signals (h-index 1, i10-index 0) are too thin to estimate.
Impact—
Impact data is unavailable in this snapshot, so no score is shown. This axis would measure citation-based influence — citations per paper, h-index trajectory, and uptake by other researchers — and cannot yet be estimated for this early-stage record.
Focus0/100
The focus score of 0 reflects a deliberately broad portfolio: declared fields split exactly 50/50 between machine learning — AI alignment / preference elicitation and AI safety / LLM agent behavior, and topic clustering spreads across five areas, led by Ethics and Social Impacts of AI (8), followed by Names, Identity, and Discrimination Research (6), AI and HR Technologies (6), Experimental Behavioral Economics (4), and XAI (3).
Collaboration48/100
With a collaboration score of 48 and 10 distinct coauthors, the profile sits mid-range: work is team-based but the network is still compact. The frequent-coauthor list includes Anagha Tiwari, Alexander G. Gray, Nick Feamster, Alex Imas, Alex Kale, Aman Priyanshu, Supriti Vijay, and Niloofar Mireshghallah; on both Evolve-reviewed 2026 papers the researcher holds a middle, non-corresponding author position, consistent with contributing within larger teams.
Activity81/100
The activity score of 81 is the strongest signal in the profile: 15 works within the past three years, with indexed output rising from 6 works in 2025 to 9 in 2026, a steep early ramp. Both cataloged papers appeared as 2026 arXiv preprints, indicating a currently active, high-tempo phase concentrated in the last twelve months.
Reproducibility38/100
Reproducibility scores 38, computed from a mean of 38 across the 2 papers Evolve has reviewed. Readiness differs by project: 'Et Tu, Brute?' is rated medium — a rigorous synthetic audit whose ecological validity remains unproven — while 'GUIDE' is rated low, being in silico only with no uncertainty reported. The small base of 2 papers makes this estimate fragile.
Research summary
Brian Jabarian is a researcher at Carnegie Mellon University (US) working at the intersection of machine learning alignment and AI safety, with an emphasis on how AI systems shape human economic decisions. The OpenAlex-resolved bibliometric snapshot shows a young, fast-moving profile: 15 works in the past three years, with 6 works recorded in 2025 rising to 9 in 2026, and an h-index of 1 (i10-index 0) — consistent with an early-stage record. The portfolio splits evenly across two declared fields: machine learning — AI alignment / preference elicitation, and AI safety / LLM agent behavior. Topic clustering groups most output under Ethics and Social Impacts of AI (8), followed by Names, Identity, and Discrimination Research (6), AI and HR Technologies (6), Experimental Behavioral Economics (4), and XAI (3). Two Evolve-reviewed 2026 preprints anchor the profile: 'Et Tu, Brute?', a computational audit finding 8 of 13 LLM agents steer wealthier users toward pricier options even against stated preferences, and 'GUIDE', a preference-elicitation engine combining LLM dialogue, Bayesian particle filtering, and symbolic priors. Reproducibility readiness is modest (mean 38 across 2 papers), and on both papers the researcher holds a middle, non-corresponding author position within larger teams.
Collaboration pattern
The collaboration pattern is team-based but compact: 10 distinct coauthors and a mid-range collaboration score of 48. Recurring partners include Anagha Tiwari, Alexander G. Gray, Nick Feamster, Alex Imas, Alex Kale, Aman Priyanshu, Supriti Vijay, and Niloofar Mireshghallah. On both Evolve-reviewed 2026 papers the researcher appears in a middle, non-corresponding position, suggesting contribution within larger CMU-led teams rather than solo or corresponding-author leadership so far.
Research keywords
Highlights
Auditing economic misalignment in personal AI agents
GUIDE: a hybrid engine for preference elicitation
Timeline
- 2025First indexed output appears in the OpenAlex snapshot: 6 works recorded in 2025, the earliest year with activity in the profile.
- 2026Indexed output rises to 9 works, the strongest single year in the snapshot, with 10 distinct coauthors appearing in the collaboration network.
- 2026'Et Tu, Brute? Economic Misalignment in Personal AI Agents' posted to arXiv (middle author, non-corresponding); Evolve reviews it as a rigorous synthetic audit with medium reproducibility readiness.
- 2026'GUIDE: Generative Utility Inference and Decision Engine' posted to arXiv (middle author, non-corresponding); Evolve notes strong early turns, in-silico-only evaluation, and low reproducibility readiness.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometrics are derived from OpenAlex only; no ORCID is linked, so this profile is name-resolved from public bibliographic data and carries some name-disambiguation risk.
- The output and impact axes lack sufficient underlying data in this snapshot; the reported h-index (1) and i10-index (0) come from a very thin record.
- Both Evolve-reviewed papers are 2026 arXiv preprints with middle, non-corresponding author positions; no peer-reviewed venue record appears in the catalog.
- The reproducibility score (mean 38) is based on only 2 reviewed papers, so the estimate is fragile.