Beyond Single Ground Truth: Reference Monism as Epistemic Injustice in ASR Evaluation

TL;DR

This paper critiques the reliance on a single ground truth in ASR evaluation, highlighting how it causes epistemic injustice, especially for speakers with aphasia, and proposes a range-based evaluation approach.

Contribution

It introduces a philosophical framework and formal metric to measure the injustice caused by reference monism, and empirically demonstrates the variability in WER based on transcription conventions.

Findings

WER varies with transcription conventions.

Reference monism causes epistemic injustice for speakers with aphasia.

Proposes WER-Range to better capture performance across conventions.

Abstract

Automatic speech recognition (ASR) evaluation compares system output to ground truth transcripts, with Word Error Rate (WER) quantifying the distance between them. But ground truth transcripts are not discovered - they are produced by human annotators following conventions that encode normative assumptions about which speech features matter. Different conventions (verbatim, non-verbatim, legal) produce different transcripts of identical speech and judge the same ASR output differently. This paper argues that reference monism - enforcing a single transcription convention as ground truth - commits epistemic injustice. Speakers with aphasia, whose speech includes clinically meaningful disfluencies, are systematically disadvantaged when evaluated against "clean" references that treat those disfluencies as errors. The harm is not merely differential performance, but that evaluative infrastructure lacks interpretive resources to recognize their contributions as legitimate. We develop a philosophical framework introducing the hermeneutical gap, formalize Epistemic Injustice Distance (EID) to measure reference monism's cost, and demonstrate empirically using AphasiaBank that WER varies depending on which convention defines ground truth. We propose WER-Range: reporting performance across legitimate conventions rather than assuming a single correct answer.