Reproduce “Evidence as Opinions of Experts” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Robert Hummel, Michael Landy

This research reinterprets the Dempster/Shafer (D-S) Theory of Evidence, showing its combination rule as a statistical aggregation of "opinions of experts" that perform pairwise Bayesian updates. It demonstrates that D-S theory implicitly reduces probabilistic opinions to boolean statements and introduces an alternative method for combining probabilistic opinions with an O(n^2) parameter representation, contrasting D-S's 2^n parameters.

Key insights

  • The normalization term in the Dempster rule is operationally optional during intermediate calculations, streamlining the algebraic process.
  • The D-S combination acts as a representation of aggregate statistics derived from individual experts performing Bayesian updates, demystifying its "non-Bayesian" appearance.
  • The D-S theory implicitly reduces nuanced probabilistic opinions to binary (boolean) possibility statements at the expert level.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.