Conformal Prediction Sets Improve Human Decision Making

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cresswell, Jesse C., Sui, Yi, Kumar, Bhargava, Vouitsis, Noël
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910478597357568
author Cresswell, Jesse C.
Sui, Yi
Kumar, Bhargava
Vouitsis, Noël
author_facet Cresswell, Jesse C.
Sui, Yi
Kumar, Bhargava
Vouitsis, Noël
contents In response to everyday queries, humans explicitly signal uncertainty and offer alternative answers when they are unsure. Machine learning models that output calibrated prediction sets through conformal prediction mimic this human behaviour; larger sets signal greater uncertainty while providing alternatives. In this work, we study the usefulness of conformal prediction sets as an aid for human decision making by conducting a pre-registered randomized controlled trial with conformal prediction sets provided to human subjects. With statistical significance, we find that when humans are given conformal prediction sets their accuracy on tasks improves compared to fixed-size prediction sets with the same coverage guarantee. The results show that quantifying model uncertainty with conformal prediction is helpful for human-in-the-loop decision making and human-AI teams.
format Preprint
id arxiv_https___arxiv_org_abs_2401_13744
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Conformal Prediction Sets Improve Human Decision Making
Cresswell, Jesse C.
Sui, Yi
Kumar, Bhargava
Vouitsis, Noël
Machine Learning
Human-Computer Interaction
In response to everyday queries, humans explicitly signal uncertainty and offer alternative answers when they are unsure. Machine learning models that output calibrated prediction sets through conformal prediction mimic this human behaviour; larger sets signal greater uncertainty while providing alternatives. In this work, we study the usefulness of conformal prediction sets as an aid for human decision making by conducting a pre-registered randomized controlled trial with conformal prediction sets provided to human subjects. With statistical significance, we find that when humans are given conformal prediction sets their accuracy on tasks improves compared to fixed-size prediction sets with the same coverage guarantee. The results show that quantifying model uncertainty with conformal prediction is helpful for human-in-the-loop decision making and human-AI teams.
title Conformal Prediction Sets Improve Human Decision Making
topic Machine Learning
Human-Computer Interaction
url https://arxiv.org/abs/2401.13744