Using Default Logic to Create Adaptable User Models for Behavior Support Agents

Johanna Wolff; Victor De Boer; Dirk Heylen; M. Birna Van Riemsdijk

doi:10.3233/FAIA240207

Using Default Logic to Create Adaptable User Models for Behavior Support Agents

Johanna Wolff^*, Victor De Boer, Dirk Heylen, M. Birna Van Riemsdijk

^*Corresponding author for this work

Research output: Chapter in Book / Report / Conference proceeding › Conference contribution › Academic › peer-review

Abstract

Behavior support agents can assist a user in reaching their goals by suggesting suitable actions. In order for these agents to be effective, the agent's advice should be personalized to the user's needs and preferences. However, the way context influences the user, the internal state of the user and the user's desired behavior are all subject to change while the agent is in use. If the agent is not able to adapt to these changes, this can lead to a misalignment between the user and the agent. By making the reasoning of the agent explicit, we can allow the user to directly interact with the agent's user model in order to resolve possible misalignments. We propose to use ordered default logic to reason about the user model as its defeasible nature is inherently well suited to model behavior patterns and routines which may have exceptions dependent on the context. We then analyze different misalignment scenarios and describe how we can use various belief revision techniques to update the agent's user model and resolve these misalignments.

Original language	English
Title of host publication	HHAI 2024: Hybrid Human AI Systems for the Social Good
Subtitle of host publication	Proceedings of the 3rd International Conference on Hybrid Human-Artificial Intelligence
Editors	Fabian Lorig, Jason Tucker, Adam Dahlgren Lindstrom, Frank Dignum, Pradeep Murukannaiah, Andreas Theodorou, Pinar Yolum
Publisher	IOS Press BV
Pages	350-359
Number of pages	10
ISBN (Electronic)	9781643685229
DOIs	https://doi.org/10.3233/FAIA240207
Publication status	Published - 2024
Event	3rd International Conference on Hybrid Human-Artificial Intelligence, HHAI 2024 - Hybrid, Malmo, Sweden Duration: 10 Jun 2024 → 14 Jun 2024

Publication series

Name	Frontiers in Artificial Intelligence and Applications
Publisher	IOS Press
Volume	386
ISSN (Print)	0922-6389
ISSN (Electronic)	1879-8314

Conference

Conference	3rd International Conference on Hybrid Human-Artificial Intelligence, HHAI 2024
Country/Territory	Sweden
City	Hybrid, Malmo
Period	10/06/24 → 14/06/24

Bibliographical note

Publisher Copyright:
© 2024 The Authors.

Keywords

Behavior Support Agents
Belief Revision
Default Logic
Human-Machine Alignment
Non-Monotonic Logics
User Modeling

Access to Document

10.3233/FAIA240207Licence: CC BY-NC

Persistent URL (handle)

Persistent link

Cite this

Wolff, J., De Boer, V., Heylen, D., & Van Riemsdijk, M. B. (2024). Using Default Logic to Create Adaptable User Models for Behavior Support Agents. In F. Lorig, J. Tucker, A. D. Lindstrom, F. Dignum, P. Murukannaiah, A. Theodorou, & P. Yolum (Eds.), HHAI 2024: Hybrid Human AI Systems for the Social Good: Proceedings of the 3rd International Conference on Hybrid Human-Artificial Intelligence (pp. 350-359). (Frontiers in Artificial Intelligence and Applications; Vol. 386). IOS Press BV. https://doi.org/10.3233/FAIA240207

Wolff, Johanna ; De Boer, Victor ; Heylen, Dirk et al. / Using Default Logic to Create Adaptable User Models for Behavior Support Agents. HHAI 2024: Hybrid Human AI Systems for the Social Good: Proceedings of the 3rd International Conference on Hybrid Human-Artificial Intelligence. editor / Fabian Lorig ; Jason Tucker ; Adam Dahlgren Lindstrom ; Frank Dignum ; Pradeep Murukannaiah ; Andreas Theodorou ; Pinar Yolum. IOS Press BV, 2024. pp. 350-359 (Frontiers in Artificial Intelligence and Applications).

@inproceedings{8cc5cf4b40634cd7add5d8c35c34fe11,

title = "Using Default Logic to Create Adaptable User Models for Behavior Support Agents",

abstract = "Behavior support agents can assist a user in reaching their goals by suggesting suitable actions. In order for these agents to be effective, the agent's advice should be personalized to the user's needs and preferences. However, the way context influences the user, the internal state of the user and the user's desired behavior are all subject to change while the agent is in use. If the agent is not able to adapt to these changes, this can lead to a misalignment between the user and the agent. By making the reasoning of the agent explicit, we can allow the user to directly interact with the agent's user model in order to resolve possible misalignments. We propose to use ordered default logic to reason about the user model as its defeasible nature is inherently well suited to model behavior patterns and routines which may have exceptions dependent on the context. We then analyze different misalignment scenarios and describe how we can use various belief revision techniques to update the agent's user model and resolve these misalignments.",

keywords = "Behavior Support Agents, Belief Revision, Default Logic, Human-Machine Alignment, Non-Monotonic Logics, User Modeling",

author = "Johanna Wolff and {De Boer}, Victor and Dirk Heylen and {Van Riemsdijk}, {M. Birna}",

note = "Publisher Copyright: {\textcopyright} 2024 The Authors.; 3rd International Conference on Hybrid Human-Artificial Intelligence, HHAI 2024 ; Conference date: 10-06-2024 Through 14-06-2024",

year = "2024",

doi = "10.3233/FAIA240207",

language = "English",

series = "Frontiers in Artificial Intelligence and Applications",

publisher = "IOS Press BV",

pages = "350--359",

editor = "Fabian Lorig and Jason Tucker and Lindstrom, {Adam Dahlgren} and Frank Dignum and Pradeep Murukannaiah and Andreas Theodorou and Pinar Yolum",

booktitle = "HHAI 2024: Hybrid Human AI Systems for the Social Good",

address = "Netherlands",

}

Wolff, J, De Boer, V, Heylen, D & Van Riemsdijk, MB 2024, Using Default Logic to Create Adaptable User Models for Behavior Support Agents. in F Lorig, J Tucker, AD Lindstrom, F Dignum, P Murukannaiah, A Theodorou & P Yolum (eds), HHAI 2024: Hybrid Human AI Systems for the Social Good: Proceedings of the 3rd International Conference on Hybrid Human-Artificial Intelligence. Frontiers in Artificial Intelligence and Applications, vol. 386, IOS Press BV, pp. 350-359, 3rd International Conference on Hybrid Human-Artificial Intelligence, HHAI 2024, Hybrid, Malmo, Sweden, 10/06/24. https://doi.org/10.3233/FAIA240207

Using Default Logic to Create Adaptable User Models for Behavior Support Agents. / Wolff, Johanna; De Boer, Victor; Heylen, Dirk et al.
HHAI 2024: Hybrid Human AI Systems for the Social Good: Proceedings of the 3rd International Conference on Hybrid Human-Artificial Intelligence. ed. / Fabian Lorig; Jason Tucker; Adam Dahlgren Lindstrom; Frank Dignum; Pradeep Murukannaiah; Andreas Theodorou; Pinar Yolum. IOS Press BV, 2024. p. 350-359 (Frontiers in Artificial Intelligence and Applications; Vol. 386).

Research output: Chapter in Book / Report / Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Using Default Logic to Create Adaptable User Models for Behavior Support Agents

AU - Wolff, Johanna

AU - De Boer, Victor

AU - Heylen, Dirk

AU - Van Riemsdijk, M. Birna

PY - 2024

Y1 - 2024

N2 - Behavior support agents can assist a user in reaching their goals by suggesting suitable actions. In order for these agents to be effective, the agent's advice should be personalized to the user's needs and preferences. However, the way context influences the user, the internal state of the user and the user's desired behavior are all subject to change while the agent is in use. If the agent is not able to adapt to these changes, this can lead to a misalignment between the user and the agent. By making the reasoning of the agent explicit, we can allow the user to directly interact with the agent's user model in order to resolve possible misalignments. We propose to use ordered default logic to reason about the user model as its defeasible nature is inherently well suited to model behavior patterns and routines which may have exceptions dependent on the context. We then analyze different misalignment scenarios and describe how we can use various belief revision techniques to update the agent's user model and resolve these misalignments.

AB - Behavior support agents can assist a user in reaching their goals by suggesting suitable actions. In order for these agents to be effective, the agent's advice should be personalized to the user's needs and preferences. However, the way context influences the user, the internal state of the user and the user's desired behavior are all subject to change while the agent is in use. If the agent is not able to adapt to these changes, this can lead to a misalignment between the user and the agent. By making the reasoning of the agent explicit, we can allow the user to directly interact with the agent's user model in order to resolve possible misalignments. We propose to use ordered default logic to reason about the user model as its defeasible nature is inherently well suited to model behavior patterns and routines which may have exceptions dependent on the context. We then analyze different misalignment scenarios and describe how we can use various belief revision techniques to update the agent's user model and resolve these misalignments.

KW - Behavior Support Agents

KW - Belief Revision

KW - Default Logic

KW - Human-Machine Alignment

KW - Non-Monotonic Logics

KW - User Modeling

UR - http://www.scopus.com/inward/record.url?scp=85198708668&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85198708668&partnerID=8YFLogxK

U2 - 10.3233/FAIA240207

DO - 10.3233/FAIA240207

M3 - Conference contribution

AN - SCOPUS:85198708668

T3 - Frontiers in Artificial Intelligence and Applications

SP - 350

EP - 359

BT - HHAI 2024: Hybrid Human AI Systems for the Social Good

A2 - Lorig, Fabian

A2 - Tucker, Jason

A2 - Lindstrom, Adam Dahlgren

A2 - Dignum, Frank

A2 - Murukannaiah, Pradeep

A2 - Theodorou, Andreas

A2 - Yolum, Pinar

PB - IOS Press BV

T2 - 3rd International Conference on Hybrid Human-Artificial Intelligence, HHAI 2024

Y2 - 10 June 2024 through 14 June 2024

ER -

Wolff J, De Boer V, Heylen D, Van Riemsdijk MB. Using Default Logic to Create Adaptable User Models for Behavior Support Agents. In Lorig F, Tucker J, Lindstrom AD, Dignum F, Murukannaiah P, Theodorou A, Yolum P, editors, HHAI 2024: Hybrid Human AI Systems for the Social Good: Proceedings of the 3rd International Conference on Hybrid Human-Artificial Intelligence. IOS Press BV. 2024. p. 350-359. (Frontiers in Artificial Intelligence and Applications). doi: 10.3233/FAIA240207

Using Default Logic to Create Adaptable User Models for Behavior Support Agents

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Persistent URL (handle)

Other files and links

Fingerprint

Cite this