Current and Future Varieties of Human-AI Collaboration
March 31 - April 2, 2025
San Francisco Airport Marriott Waterfront | Burlingame, CA, USA
Submissions to the symposium are currently closed. Accepted contributions will be announced soon.
There is general agreement that AI has the potential to revolutionize many aspects of society, but existing systems lack the flexibility and robustness of human thought. Before they can assist humans effectively, they must be able to interact with people as genuine collaborators. This suggests that we should explore more fully the ways that humans and machines can work together, each drawing on strengths that offset the other's weaknesses. We should also explore concerns about risk, trust, and safety that arise in collaborative settings.
There is general agreement that AI has the potential to revolutionize many aspects of society, but existing systems lack the flexibility and robustness of human thought. Before they can assist humans effectively, they must be able to interact with people as genuine collaborators. This suggests that we should explore more fully the ways that humans and machines can work together, each drawing on strengths that offset the other's weaknesses. We should also explore concerns about risk, trust, and safety that arise in collaborative settings.
Focus of the Symposium
The Symposium focuses on varieties of human-AI collaboration – both existing and potential – that reflects the entire space of task-oriented interactions. An illustrative but incomplete list includes team settings in which:
A human oversees and directs AI agents on some task
An AI agent advises a human on executing some complex task
Humans and AI agents cooperate on a some task as equal members
An AI system coordinates a set of humans and agents
Humans and AI agents handle subtasks usually done by one person.
These differ in the roles played by members, their responsibilities, and their interactions. Teams can also differ in size, organization, and how they address task components. These design choices cut across different applications and AI paradigms.
Expected Presentations
Presentations will discuss application domains that can benefit from collaborative AI, implemented systems that participate in human-AI teams, and remaining challenges. Rather than focusing on details of AI systems, submissions and presentations should discuss the:
Problem domain and overall objective of human-AI collaboration
Human and AI team members, including abilities and expertise
Roles and responsibilities of team members for component tasks
Objectives, expectations, and interactions among team members
Incorporation of features like transparency, trust, and dependability
Responses to benefit-risk tradeoffs in AI-human collaboration; and
Measures of task performance for both the overall team and members.
This will make talks more accessible to participants with different backgrounds, encourage discussion of general principles for human-AI teaming, and forge a stronger research community.
Topics of interest
The topics of interest for the symposium include:
Dimensions along which human-AI teams can vary, along with existing examples that illustrate them;
Possible team organizations that have not been explored but that show promise for effective interaction;
Ways that team designs can balance the benefits of including AI team members with the risks;
Influence of features like AI transparency, interpretability, and explainability on team effectiveness;
Principles for team design, including how to assign roles and responsibilities to people vs. machines;
Modalities suited to support human-AI communication (e.g., text, speech, graphics) in different settings;
Social norms that facilitate meaningful yet feasible interactions between human and AI teammates;
Impacts of these and other factors on the engineering life cycle, including development and deployment;
Socio-technical perspectives on how AI and people shape each others’ behavior during collaboration; and
Criteria for effectiveness of human-AI teams, ways to measure them, and likely tradeoffs among them.
Together, these insights will offer a road map for future research on human-AI collaboration, including open issues and priorities for addressing them.
Submission format
Authors should submit extended abstracts of proposed talks.
Submissions should be one to two pages, including figures and references, using 11-point font in any PDF format.
We will prioritize abstracts that address the facets of human-AI collaboration listed above.
We do not plan to compile a proceedings volume, although revised abstracts will be available on the meeting Web site.
Authors are welcome to reuse content they plan to publish or that has already appeared.
Submissions will be accepted through EasyChair only. The submission link will be available soon.
Important dates
Submission deadline: Friday, February 7, 2025, Anytime On Earth
Notifications to authors: Friday, February 21, 2025, Anytime On Earth
Revised abstracts due: Friday, March 14, 2025, Anytime On Earth
Symposium dates: March 31, 2025, to April 2, 2025
Organizers
David W. Aha, Navy Center for Applied Research in AI (https://www.linkedin.com/in/david-w-aha/)
Marco Brambilla, Politecnico di Milano, Italy (https://www.marco-brambilla.com/)
Matthew Johnson, InstituteforHumanandMachineCognition(https://www.ihmc.us/groups/mjohnson/)
Pat Langley, Georgia Tech Research Institute (http://www.isle.org/∼langley/)
William Lawless, Paine College
Venue
San Francisco Airport Marriott Waterfront in Burlingame, CA. USA
1800 Old Bayshore Hwy, Burlingame, CA 94010
This event is part of the AAAI Spring Symposium Series
https://aaai.org/conference/spring-symposia/sss25/