Preference-Conditioned Language-Guided Abstraction

Learning from demonstrations is a common way for users to teach robots, but it is prone to spurious feature correlations. Recent work constructs state abstractions, i.e. visual representations containing task-relevant features, from language as a way to perform more generalizable learning. However,...

Full description

Saved in:
Bibliographic Details
Published in:2024 19th ACM/IEEE International Conference on Human-Robot Interaction (HRI) pp. 572 - 581
Main Authors: Peng, Andi, Bobu, Andreea, Li, Belinda Z., Sumers, Theodore R., Sucholutsky, Ilia, Kumar, Nishanth, Griffiths, Thomas L., Shah, Julie A.
Format: Conference Proceeding
Language:English
Published: ACM 11.03.2024
Subjects:
Behavioral sciences
Correlation
human preferences
Human-robot interaction
learning from human input
Robots
state abstraction
Visualization
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract Learning from demonstrations is a common way for users to teach robots, but it is prone to spurious feature correlations. Recent work constructs state abstractions, i.e. visual representations containing task-relevant features, from language as a way to perform more generalizable learning. However, these abstractions also depend on a user's preference for what matters in a task, which may be hard to describe or infeasible to exhaustively specify using language alone. How do we construct abstractions to capture these latent preferences? We observe that how humans behave reveals how they see the world. Our key insight is that changes in human behavior inform us that there are differences in preferences for how humans see the world, i.e. their state abstractions. In this work, we propose using language models (LMs) to query for those preferences directly given knowledge that a change in behavior has occurred. In our framework, we use the LM in two ways: first, given a text description of the task and knowledge of behavioral change between states, we query the LM for possible hidden preferences; second, given the most likely preference, we query the LM to construct the state abstraction. In this framework, the LM is also able to ask the human directly when uncertain about its own estimate. We demonstrate our framework's ability to construct effective preference-conditioned abstractions in simulated experiments, a user study, as well as on a real Spot robot performing mobile manipulation tasks.CCS CONCEPTS* Computing methodologies → Artifcial intelligence; Learning from demonstrations.
AbstractList Learning from demonstrations is a common way for users to teach robots, but it is prone to spurious feature correlations. Recent work constructs state abstractions, i.e. visual representations containing task-relevant features, from language as a way to perform more generalizable learning. However, these abstractions also depend on a user's preference for what matters in a task, which may be hard to describe or infeasible to exhaustively specify using language alone. How do we construct abstractions to capture these latent preferences? We observe that how humans behave reveals how they see the world. Our key insight is that changes in human behavior inform us that there are differences in preferences for how humans see the world, i.e. their state abstractions. In this work, we propose using language models (LMs) to query for those preferences directly given knowledge that a change in behavior has occurred. In our framework, we use the LM in two ways: first, given a text description of the task and knowledge of behavioral change between states, we query the LM for possible hidden preferences; second, given the most likely preference, we query the LM to construct the state abstraction. In this framework, the LM is also able to ask the human directly when uncertain about its own estimate. We demonstrate our framework's ability to construct effective preference-conditioned abstractions in simulated experiments, a user study, as well as on a real Spot robot performing mobile manipulation tasks.CCS CONCEPTS* Computing methodologies → Artifcial intelligence; Learning from demonstrations.
Author Griffiths, Thomas L.
Sucholutsky, Ilia
Bobu, Andreea
Peng, Andi
Li, Belinda Z.
Sumers, Theodore R.
Shah, Julie A.
Kumar, Nishanth
Author_xml – sequence: 1
  givenname: Andi
  surname: Peng
  fullname: Peng, Andi
  organization: MIT
– sequence: 2
  givenname: Andreea
  surname: Bobu
  fullname: Bobu, Andreea
  organization: Boston Dynamics AI Institute
– sequence: 3
  givenname: Belinda Z.
  surname: Li
  fullname: Li, Belinda Z.
  organization: MIT
– sequence: 4
  givenname: Theodore R.
  surname: Sumers
  fullname: Sumers, Theodore R.
  organization: Princeton
– sequence: 5
  givenname: Ilia
  surname: Sucholutsky
  fullname: Sucholutsky, Ilia
  organization: Princeton
– sequence: 6
  givenname: Nishanth
  surname: Kumar
  fullname: Kumar, Nishanth
  organization: MIT
– sequence: 7
  givenname: Thomas L.
  surname: Griffiths
  fullname: Griffiths, Thomas L.
  organization: Princeton
– sequence: 8
  givenname: Julie A.
  surname: Shah
  fullname: Shah, Julie A.
  organization: MIT
BookMark eNotjMFKw0AQQFdQUGvOXjz4A6kzO7s72WMp2goBPei5TDazZUFTSdKDf29FT4_Hg3dtzofDoMbcIiwRnX-ggBCZlxTIRYIzU0WOjQNgIGv9pammqXTgXURwvrky9nXUrKMOSev1YejLXE7H_r6VYX-UvdabY-lPvuqmeZT0W2_MRZaPSat_Lsz70-Pbelu3L5vn9aqtxTZxrntrkw2MTMKKqUGhxJa7zCIampAgBmKvKArBOvSas3qbbSCNGT0tzN3ft6jq7mssnzJ-7xBCgAhIP2nuRFc
ContentType Conference Proceeding
DBID 6IE
6IL
CBEJK
RIE
RIL
DOI 10.1145/3610977.3634930
DatabaseName IEEE Electronic Library (IEL) Conference Proceedings
IEEE Xplore POP ALL
IEEE Xplore All Conference Proceedings
IEEE Electronic Library (IEL)
IEEE Proceedings Order Plans (POP All) 1998-Present
DatabaseTitleList
Database_xml – sequence: 1
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
EISBN 9798400703225
EndPage 581
ExternalDocumentID 10660901
Genre orig-research
GroupedDBID 6IE
6IL
ACM
ALMA_UNASSIGNED_HOLDINGS
APO
CBEJK
LHSKQ
RIE
RIL
ID FETCH-LOGICAL-a289t-d22c267173a7e1c81a3c727bf7aae686c096375e1ae062415effe52f263e9f153
IEDL.DBID RIE
ISICitedReferencesCount 4
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001239977500061&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
IngestDate Tue May 06 03:31:43 EDT 2025
IsDoiOpenAccess false
IsOpenAccess true
IsPeerReviewed false
IsScholarly false
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-a289t-d22c267173a7e1c81a3c727bf7aae686c096375e1ae062415effe52f263e9f153
OpenAccessLink https://dl.acm.org/doi/pdf/10.1145/3610977.3634930
PageCount 10
ParticipantIDs ieee_primary_10660901
PublicationCentury 2000
PublicationDate 2024-March-11
PublicationDateYYYYMMDD 2024-03-11
PublicationDate_xml – month: 03
  year: 2024
  text: 2024-March-11
  day: 11
PublicationDecade 2020
PublicationTitle 2024 19th ACM/IEEE International Conference on Human-Robot Interaction (HRI)
PublicationTitleAbbrev HRI
PublicationYear 2024
Publisher ACM
Publisher_xml – name: ACM
SSID ssib054910458
Score 1.9524171
Snippet Learning from demonstrations is a common way for users to teach robots, but it is prone to spurious feature correlations. Recent work constructs state...
SourceID ieee
SourceType Publisher
StartPage 572
SubjectTerms Behavioral sciences
Correlation
human preferences
Human-robot interaction
learning from human input
Robots
state abstraction
Visualization
Title Preference-Conditioned Language-Guided Abstraction
URI https://ieeexplore.ieee.org/document/10660901
WOSCitedRecordID wos001239977500061&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV07T8MwELagYmACRBBvZWB1iR-xnREhCkNVZQCpW-WcL1KXgkrK7-fspO3EwGZblk-WH_f87hh7sA4K4bXjjW8tpw9P8wYCcG9cJYIAErkTUHhqZzM3n1f1AFZPWBhETMFnOI7N5MsPn7CJpjJ64cYUVURrHVprerDW9vKQniOi029I3yN0-ahiKnFrx8ooXaUw5339lMQ-Jif_JHzKsj0QL693LOaMHeDqnMl6VxuE07zQZxsK-XQwPfLXzTJQ_6mJZoyEW8jYx-Tl_fmND6UPuCcNqONBSpAmesi9RQFOeAUkaTSt9R6NM0Cah7IlCo-FiUw4Rn-UspVGYdXSL3bBRiuifcly0SC6thHOodYQT8N4qyzQkgp86a9YFje8-OqzWyy2e73-Y_yGHUti7DEOS4hbNurWG7xjR_DTLb_X9-lMfgErY40t
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LSwMxEB6kCnpSseLbHrymbh6bZI8i1opr6aFCbyWbzEIvVfrw9ztJt-3Jg7ckhAwhj3l-MwAPxvqMO2VZ5WrD6MNTrPLBM6dtwQP3JHInoHBpBgM7HhfDBqyesDCImILPsBubyZcfvvwqmsrohWudFRGttZ8rJbI1XGtzfUjT4dHt1yTw4Sp_lDGZuDFdqaUqUqDzroJKYiC943-SPoH2DorXGW6ZzCns4ewMxHBbHYTRvLDONxQ6ZWN8ZK-raaD-UxUNGQm50IbP3svouc-a4gfMkQ60ZEEIL3T0kTuD3FvupCdZo6qNc6it9qR7SJMjd5jpyIZj_EcuaqElFjX9Y-fQmhHtC-jwCtHWFbcWlfLxPLQz0nhaUnqXu0toxw1Pvtf5LSabvV79MX4Ph_3RRzkp3wbv13AkiM3HqCzOb6C1nK_wFg78z3K6mN-l8_kFt56QdA
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Abook&rft.genre=proceeding&rft.title=2024+19th+ACM%2FIEEE+International+Conference+on+Human-Robot+Interaction+%28HRI%29&rft.atitle=Preference-Conditioned+Language-Guided+Abstraction&rft.au=Peng%2C+Andi&rft.au=Bobu%2C+Andreea&rft.au=Li%2C+Belinda+Z.&rft.au=Sumers%2C+Theodore+R.&rft.date=2024-03-11&rft.pub=ACM&rft.spage=572&rft.epage=581&rft_id=info:doi/10.1145%2F3610977.3634930&rft.externalDocID=10660901