The curse of dimensionality in Voting Advice Applications: reliability and validity in algorithm design

ABSTRACT

The proliferation of Voting Advice Applications (VAAs) offers voters a simplified means by which to estimate their alignment with the candidates contesting a given election. But are the outputs generated by such applications both reliable and valid? While they differ greatly in design and degree of sophistication, most VAAs share a distinctive element that serves as their defining feature: an aggregation algorithm. Aggregation algorithms are the source of a VAA's legitimacy and yet no formal framework for their evaluation has as yet been agreed upon. We posit a dimensionality reduction technique as a corrective to recognised shortcomings in the dominant approaches to VAA design. We then test our model within a proposed framework for evaluating the validity and reliability of VAA algorithms.