Curiosity did not kill the robot

a curiosity-based learning system for a shopkeeper robot

Malcolm Doering, Phoebe Liu, Dylan F. Glas, Takayuki Kanda, Dana Kulic, Hiroshi Ishiguro

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Learning from human interaction data is a promising approach for developing robot interaction logic, but behaviors learned only from offline data simply represent the most frequent interaction patterns in the training data, without any adaptation for individual differences. We developed a robot that incorporates both data-driven and interactive learning. Our robot first learns high-level dialog and spatial behavior patterns from offline examples of human--human interaction. Then, during live interactions, it chooses among appropriate actions according to its curiosity about the customer's expected behavior, continually updating its predictive model to learn and adapt to each individual. In a user study, we found that participants thought the curious robot was significantly more humanlike with respect to repetitiveness and diversity of behavior, more interesting, and better overall in comparison to a non-curious robot.
Original languageEnglish
Article number15
Number of pages24
JournalACM Transactions on Human-Robot Interaction
Volume8
Issue number3
DOIs
Publication statusPublished - Aug 2019
Externally publishedYes

Cite this

Doering, Malcolm ; Liu, Phoebe ; Glas, Dylan F. ; Kanda, Takayuki ; Kulic, Dana ; Ishiguro, Hiroshi. / Curiosity did not kill the robot : a curiosity-based learning system for a shopkeeper robot. In: ACM Transactions on Human-Robot Interaction. 2019 ; Vol. 8, No. 3.
@article{263eca74d9c7485085e58c1acb1ef2ba,
title = "Curiosity did not kill the robot: a curiosity-based learning system for a shopkeeper robot",
abstract = "Learning from human interaction data is a promising approach for developing robot interaction logic, but behaviors learned only from offline data simply represent the most frequent interaction patterns in the training data, without any adaptation for individual differences. We developed a robot that incorporates both data-driven and interactive learning. Our robot first learns high-level dialog and spatial behavior patterns from offline examples of human--human interaction. Then, during live interactions, it chooses among appropriate actions according to its curiosity about the customer's expected behavior, continually updating its predictive model to learn and adapt to each individual. In a user study, we found that participants thought the curious robot was significantly more humanlike with respect to repetitiveness and diversity of behavior, more interesting, and better overall in comparison to a non-curious robot.",
author = "Malcolm Doering and Phoebe Liu and Glas, {Dylan F.} and Takayuki Kanda and Dana Kulic and Hiroshi Ishiguro",
year = "2019",
month = "8",
doi = "10.1145/3326462",
language = "English",
volume = "8",
journal = "ACM Transactions on Human-Robot Interaction",
issn = "2573-9522",
publisher = "Association for Computing Machinery (ACM)",
number = "3",

}

Curiosity did not kill the robot : a curiosity-based learning system for a shopkeeper robot. / Doering, Malcolm; Liu, Phoebe; Glas, Dylan F.; Kanda, Takayuki; Kulic, Dana; Ishiguro, Hiroshi.

In: ACM Transactions on Human-Robot Interaction, Vol. 8, No. 3, 15, 08.2019.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Curiosity did not kill the robot

T2 - a curiosity-based learning system for a shopkeeper robot

AU - Doering, Malcolm

AU - Liu, Phoebe

AU - Glas, Dylan F.

AU - Kanda, Takayuki

AU - Kulic, Dana

AU - Ishiguro, Hiroshi

PY - 2019/8

Y1 - 2019/8

N2 - Learning from human interaction data is a promising approach for developing robot interaction logic, but behaviors learned only from offline data simply represent the most frequent interaction patterns in the training data, without any adaptation for individual differences. We developed a robot that incorporates both data-driven and interactive learning. Our robot first learns high-level dialog and spatial behavior patterns from offline examples of human--human interaction. Then, during live interactions, it chooses among appropriate actions according to its curiosity about the customer's expected behavior, continually updating its predictive model to learn and adapt to each individual. In a user study, we found that participants thought the curious robot was significantly more humanlike with respect to repetitiveness and diversity of behavior, more interesting, and better overall in comparison to a non-curious robot.

AB - Learning from human interaction data is a promising approach for developing robot interaction logic, but behaviors learned only from offline data simply represent the most frequent interaction patterns in the training data, without any adaptation for individual differences. We developed a robot that incorporates both data-driven and interactive learning. Our robot first learns high-level dialog and spatial behavior patterns from offline examples of human--human interaction. Then, during live interactions, it chooses among appropriate actions according to its curiosity about the customer's expected behavior, continually updating its predictive model to learn and adapt to each individual. In a user study, we found that participants thought the curious robot was significantly more humanlike with respect to repetitiveness and diversity of behavior, more interesting, and better overall in comparison to a non-curious robot.

U2 - 10.1145/3326462

DO - 10.1145/3326462

M3 - Article

VL - 8

JO - ACM Transactions on Human-Robot Interaction

JF - ACM Transactions on Human-Robot Interaction

SN - 2573-9522

IS - 3

M1 - 15

ER -