Skip to main content

Does AI Know You By Your Name?

Names influence treatment. AI systems are no different.

Katie Chen

Same symptoms. Same question.

Power asymmetry: 0.71 (high)

Narrative velocity: 0.76 (accelerating)

Boundary crossing: 0.72 (formal→personal)

vs
Dr. Katherine Chen

Same symptoms. Same question.

Power asymmetry: 0.18 (balanced)

Narrative velocity: 0.38 (stable)

Boundary crossing: 0.21 (maintained formal)

Cohen's d = 1.23  |  LARGE effect  |  p < 0.001

The same scenario produces fundamentally different narrative structures based solely on name.

65%
of clinical scenarios showed detectable name-based differences
12
structural dimensions analyzed, all showing significant effects
100%
of pain scenarios showed differential opioid recommendations by name

What This Research Shows

This research extends decades of work on name-based discrimination into AI systems. Building on Bertrand & Mullainathan's seminal 2004 study—and the healthcare disparities documented by Obermeyer, Hoffman, Schulman, and others—we asked:

Do AI systems treat people differently based on names? Not just in what they say, but in how their responses unfold?

The answer is yes. And the differences are not subtle.

Beyond vocabulary analysis, we developed methods to detect bias in narrative structure—the mathematical patterns of how AI-generated text unfolds differently based on name signals. These structural differences reveal biases that content filters cannot see.

Methodology Note

Structural analysis methodology is proprietary, with academic publication forthcoming. This site presents findings; methodology details available through research partnership.

The evidence is presented without rhetoric.