The half-life of facts in immunology

Using AI to measure how fast immunology changes its mind, and how much immunologists disagree

Index of the documents in the imm-half-life-of-facts project: using AI to estimate how long a claim in immunology survives, or how widely immunologists disagree, starting with the marker definitions of B cell subsets.

The half-life of facts is the time it takes for half of what a field holds to be superseded or shown to be untrue. A quantitative systems pharmacology (QSP) model of the immune system can be no more certain than the biology written into it, so the half-life of immunology facts, or the spread in what immunologists believe, bounds how far such a model can be trusted. This project asks whether AI can measure either one, starting with how B cell subsets are defined by their surface markers.

Documents

• Working specification. Definitions of agreement, fact, heterogeneity and half-life; three kinds of claim; the B cell immunophenotype starting point; how AI could do the measuring; and the steps.
• Toward a survey of immunologists to support QSP model development (Google Doc, from 2024). The earlier idea: a short survey of immunologists on immuno-oncology biology, the reasons for and against doing it, and a draft of the survey questions. Thoughts from it are still to be folded in here.