AI-Assisted Human Labeling: Batching for Efficiency without Overreliance

要旨

Human labeling of training data is often a time-consuming, expensive part of machine learning. In this paper, we study "batch labeling", an AI-assisted UX paradigm, that aids data labelers by allowing a single labeling action to apply to multiple records. We ran a large scale study on Mechanical Turk with 156 participants to investigate labeler-AI-batching system interaction. We investigate the efficacy of the system when compared to a single-item labeling interface (i.e., labeling one record at-a-time), and evaluate the impact of batch labeling on accuracy, and time. We further investigate the impact of AI algorithm quality and its effects on the labelers' overreliance, as well as potential mechanisms for mitigating it. Our work offers implications for the design of batch labeling systems and for work practices focusing on labeler-AI-batching system interaction.

著者
Zahra Ashktorab
IBM Research, Yorktown Heights, New York, United States
Casey Dugan
IBM Research, Cambridge, Massachusetts, United States
Aabhas Sharma
IBM Research, Cambridge, Massachusetts, United States
Evelyn Duesterwald
IBM Research, Yorktown Heights, New York, United States
Michael Muller
IBM Research, Cambridge, Massachusetts, United States
Michael Desmond
IBM Research, Yorktown Heights, New York, United States
Christine Wolf
IBM Research - Almaden, San Jose, California, United States
Josh Andres
IBM Research Australia, Melbourne, Victoria, Australia
Narendra Nath Joshi
IBM, Cambridge, Massachusetts, United States
Kristina Brimijoin
Mrs., Hastings on Hudson, New York, United States
Werner Geyer
IBM Research, Cambridge, Massachusetts, United States
Qian Pan
IBM Research, Cambridge, Massachusetts, United States
Darrell Reimer
IBM Research AI, Yorktown Heights, New York, United States
Michelle Brachman
IBM Research, Cambridge, Massachusetts, United States
論文URL

https://doi.org/10.1145/3449163

動画

会議: CSCW2021

The 24th ACM Conference on Computer-Supported Cooperative Work and Social Computing

セッション: Data Work and AI

Papers Room B
8 件の発表
2021-10-27 22:30:00
2021-10-28 00:00:00