titleVisualization of Decision-Theoretic plans
lNameDekhtyar
fNameAlex
phone62387
emaildekhtyar@calpoly.edu
additional
departmentComputer Science
proj_descPlanning is a commonplace human activity, which is typically abstracted in terms of choosing
sequences of actions that lead to a specific goal, or maximize certain rewards.
Decision-theoretic planning is a subarea of artificial intelligence that has as its goal building
software for planning in different domains.
In simple domains, planning is deterministic: the exact effects of a ll actions are known in advance.
However, most real domains in which decision-theoretic planning is desired have actions whose
exact effects cannot be predicted in advance. For example, it is the role of an academic advisorto
suggest courses for students that best fit their goal of obtaining a specific degree (or receiving a
specific education). However, an action of take a course does not have a deterministic effect:
students may pass the course with various degree of achievement (A, B ,C), they may fail the
course, or withdraw from the course completely. When long-term plans are made in domains such
as academic advising, the liklihood of specific outcomes of actions also needs to be taken into
account.
This leads to a new formal notion of a plan (or policy), and a drastically new approach to building
and evaluating plans. Methodology for building good decision-theoretic plans in the presence of
uncertainty exists and is well-studied, however, the problem of presenting computed plans to
end-users is not.
The proposed project will concentrate on building software for presenting decision-theoretic plans
to their consumers. We will use academic advising as a sample domain , and Cal Poly students as
consumers. The goals of the multi-disciplinary team working on the p roject will be to:
(a) learn and understand the notion of a decision-theoretic plan;
(b) develop a student\\\'s cognitive model of a decision-theoretic plan in the academic advising
domain (this stage may potentially involve a number of experimental studies to assess how Cal
Poly students evaluate academic plans);
(c) design and implement software for display of decision-theoretic plans that incorporates the
developed cognitive model;
(e) evaluate the utility of the implemented software.
inter_descThe project is designed to be a collaborative effort between computer
science/mathematics/statistics students and students studying psycho logy, cognitive and
behavioral sciences. The psychology/cog.sci students will take lead in designing the experiments
and building the cognitive model of advising plans and communicating this model to the students
the CSC/Math/Statistics students. They will also conduct usability testing of the developed softwre.
The CSC/Math/Sats students will design the plan presentation software and implement it, as well as
participate in other activities within the project.
While the final deliverable of the project is a software product, th e success of the project depends
highly on the successful collaboration between two sides.
An ideal project team will involve 3-4 students, and will include one already chosen computer science
student and at least one student in psychology/cognitive/behavioral science.
The project in particular is intended for Computer Science students interested in artificial
intelligence and human-computer interaction. However, because the project already has
computer science students, we will need psychology students interested in cognition and
information, as well as, math and statistics students interested in the applications of probabilistic
methodologies.
links
students2
majorsMATH, STAT, PSY, EDUC
desired_res
date_added2008-10-24 12:09:19