Routing & Autonomy
How should applicants be allocated between AI and human interviewers to maximize skill-adjusted hires, firm surplus, fairness, and candidate agency?

Information Integrity
Can real-time multimodal fraud detection keep deepfakes, synthetic voices, and sensitive-attribute leakage under control without harming conversion or equity?

Welfare Distribution & Regulation
How are gains distributed between firms and workers? What disclosure and audit regimes secure public trust?

Workflow Augmentation-Automation
Which recruitment funnel stages should be automated first, and how do compound automation effects reshape cost-per-hire, time-to-hire, human-in-the-loop tasks and carbon footprint?

Contracts, Trade & Markets
How do reskilling, new incentives, and gig platforms (e.g., AgentsOnly) reshape recruiter performance, contracts, wage dispersion, and global talent flows?

Our team embeds theory-driven natural AI field experiments directly into live RPO/BPO operations worldwide.

• Large-Scale: at least a 100K but up to 5 million+ job applicants present in our experiments.

• Linked outcomes: offers, wages, retention, recruiter productivity, applicant satisfaction, and downstream client performance.

• Unified infrastructure: randomization and causal inference at an industrial scale, with immediate firm strategic and welfare policy relevance.

Voice AI in Firms (with Luca Henkel)

• 70,000 applicants randomized between AI and human interviewers.

• AI interviews captured richer information, raised offer rates, and improved one-month retention.

• Applicants consistently chose the AI interviewer (78%) when given the option.

• Recruiters rated AI transcripts favorably, though they leaned more on standardized tests when reviewing them.

Current and upcoming projects include:

• Screening Labor with AI and Humans: Optimal Choice and Welfare (with Pëllumb Rešhidi)

• Human–AI Learning: Theory and Field Evidence from Job Interviews (with Andrew Koh)

• Automated Recruitment Evaluation (benchmarking human vs. multimodal AI prediction)

Next focus areas:

• Fraud detection pipelines (deepfakes, identity spoofing, multimodal security).

• Gig-economy staffing models and contract design (AgentsOnly).

• General equilibrium effects on wages and talent flows across BPO hubs.

• Hybrid AI–human workflows (“AI-first, human-escalation” designs).

This research program goes beyond traditional AI studies by:

• Randomizing live stages of business pipelines

• Linking every stage of the RPO/BPO pipeline (not just prediction tasks).

• Embedding dynamic human integration rules based on real-time uncertainty.

• Incorporating security-conversion layers (multimodal fraud detection).

• Operating at global industrial scale, with outcomes tracked to wages, retention, recruiter effort, and applicant satisfaction.

• Formal DUA enabled by the University of Chicago Booth Business School and PSG Global Solutions, ensuring publication rights prior any data collection.

• Original DUA signed July 2024, renewed for 5 years (2025–2030).

• Dr. Jabarian became Chief Economist at PSG in an unpaid advisory role in August 2025 to lead this new 5-year program

• All field experiments are pre-registered on AEA RCT Registry

• All field experiments are IRB-approved at the University of Chicago Booth Business School.

• Data stored and analyzed in our secure Booth Google Cloud Platform, established through a cloud partnership with Google.