[ACT-R-users] Symposium on Human-Like Learning

[Christian Lebiere]

 [Posted for Christopher J. MacLellan, Georgia Institute of Technology,
cmaclell at gatech.edu ]


*Title:* Symposium on Human-Like Learning



*Description:* Recent machine-learning research has made incredible
progress across a wide range of tasks. While many systems can achieve
human-like performance, one area that is currently under explored is how to
realize *human-like learning* capabilities within these systems. For
example, machine learning typically employs batch training and requires
more data and computation than people to achieve similar capabilities. The
resulting models are effective, but difficult to update in the face of new
data without costly retraining. In contrast, humans excel at rapidly
assimilating new information on the fly from a limited number of examples.
More research is needed to investigate human-like capabilities, such as
efficient, incremental learning, and to explore the design of artificial
systems that can also exhibit them.



*General Themes:*

- Identification of key characteristics of human-like learning to target in
AI/ML research, and what makes them challenging for current approaches;

- Ongoing and proposed research into how to create artificial systems that
exhibit human-like learning;

- Approaches for evaluating such systems; and

- Exploration of the broader context and impacts of this research, such as
how human-like learning systems might complement/benefit current machine
learning systems and humanity.



*Relevant Topics (not exhaustive):*

   - Cognitive architectures
   - Interactive task learning
   - Extended/continual learning
   - Probabilistic programming
   - Concept formation
   - Analogical and case-based learning
   - Logic-based learning
   - Simulated students
   - Human-like neural network learning



*Format:* The symposium will consist of a small number of invited speakers,
followed by approximately 20 technical presentations and a poster session.
Each talk will be allotted 30-minutes (20 for talk and 10 for discussion).
There will also be coffee breaks and time for broader reflections and
discussions.



*Submission: *Authors will submit abstracts, which the organizing committee
will use to decide on session topics and presentations. Speaker abstracts
will also be shared with attendees. Authors of submissions that are not
presented as talks will be invited to participate in a poster session
during the first day. In choosing presenters, the committee will give
preference to submissions that are more closely aligned to the overarching
theme, while also trying to give coverage to different aspects of the theme.



Please format abstracts using the AAAI Author Kit (
https://aaai.org/authorkit24-2/) and submit via easychair (
https://easychair.org/my/conference?conf=sss24).



*Organizing Committee:*

- Christopher J. MacLellan, Georgia Institute of Technology,
cmaclell at gatech.edu

- Ute Schmid, University of Bamberg, ute.schmid at uni-bamberg.de

- Douglas Fisher, Vanderbilt University, douglas.h.fisher at vanderbilt.edu

- Randolph M. Jones, Soar Technology, LLC, rjones at soartech.com



*For more info:* https://humanlikelearning.com/aaai24-ss/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/act-r-users/attachments/20231115/dd6d01f4/attachment.html>