Don't Blame the Annotator: Bias Already Starts in the Annotation Instructions

TL;DR

This paper investigates how biases in crowdsourcing instructions influence the data collected for NLU benchmarks, leading to overfitting and poor generalization, and offers recommendations to mitigate such biases in future dataset creation.

Contribution

It identifies and analyzes instruction bias in NLU benchmarks, demonstrating its impact on model performance and generalization, and provides guidelines for better benchmark design.

Findings

Instruction bias propagates patterns from instructions to data.

Models overestimate performance due to instruction bias.

Bias effects are influenced by pattern frequency and model size.

Abstract

In recent years, progress in NLU has been driven by benchmarks. These benchmarks are typically collected by crowdsourcing, where annotators write examples based on annotation instructions crafted by dataset creators. In this work, we hypothesize that annotators pick up on patterns in the crowdsourcing instructions, which bias them to write many similar examples that are then over-represented in the collected data. We study this form of bias, termed instruction bias, in 14 recent NLU benchmarks, showing that instruction examples often exhibit concrete patterns, which are propagated by crowdworkers to the collected data. This extends previous work (Geva et al., 2019) and raises a new concern of whether we are modeling the dataset creator's instructions, rather than the task. Through a series of experiments, we show that, indeed, instruction bias can lead to overestimation of model performance, and that models struggle to generalize beyond biases originating in the crowdsourcing instructions. We further analyze the influence of instruction bias in terms of pattern frequency and model size, and derive concrete recommendations for creating future NLU benchmarks.