[agents] New AIJ Article on Belief and Truth in Hypothesised Behaviours

[Stefano V. Albrecht]

Dear Colleagues,

I would like to draw your attention to a new article on "Belief and 
Truth in Hypothesised Behaviours" in the Artificial Intelligence 
Journal, which I believe may be of broader interest:

http://authors.elsevier.com/a/1Sh7u-c5CAxC

(arXiv: http://arxiv.org/abs/1507.07688)

This work addresses important questions of *type-based interaction* 
methods, in which an autonomous agent maintains beliefs over a set of 
hypothesised behaviours ("types") for other agents. Addressed questions 
include: (1) correctness of posterior beliefs, (2) long-term impact of 
prior beliefs, (3) optimal task completion with incorrect hypothesised 
types, and (4) detection of incorrect hypothesised types.

In light of the recent debate on *robust AI*, (3) and (4) should be of 
particular interest.

Best,

-- 
Dr. Stefano V. Albrecht
Department of Computer Science
The University of Texas at Austin
http://svalbrecht.de