Project Ideas
Your project may be either
- A paper that overviews in some depth a topic touched upon in the
course. It should be between 8 and 15 pages in length, and include a
bibliography. It should be based either on a book or at least three
papers that you identify as both relevant and significant for the
topic. An excellent resource to use to find papers is CiteSeer, please
note the new URL:
Citeseer
Note that citation counts are a reasonable way to determine the influence
of a paper - a paper that is rarely or never cited is unlikely to have been
influential! Your paper could take the form of an overview or synthesis,
and/or suggest how a theory that is discussed in non-computer science terms
could actually be grounded computationally, and/or describe a potential
research project one could do based on the topic.
- A programming project that implements some of the models of cognition
discussed in the course. Group projects are welcome.
In either case, please let me know what you will be doing by February 20th.
If you are having trouble deciding please email me to set up a time to talk.
Projects are due March 17 (one week after the last class).
Paper Projects
The following are only suggestions - if you have other ideas please discuss
them with me. For cases where I mention specific books you can borrow a
copy from me.
- The role of metaphor in language and cognition. Suggested book: Metaphors
We Live By, by G. Lakoff and M. Johnson
- Neurological evidence of the role of emotion in cognition. Suggested
book: Descartes' Error, by Antonio Damasio
- Game theory and the philosophy of language. Suggested book: Convention,
by D. K. Lewis
- Logics for representing knowledge and belief. Suggested starting
point: papers by Joseph Halpern (Cornell U)
- Social and/or emotional intelligence in human and machines.
Suggested starting point: papers (or books) by psychologist Howard Gardner
on multiple intelligences; papers by Roz Picard (MIT) on affective computing
- Grand "architectures" for artificial intelligence.
Suggested starting points: papers on Prodigy and SOAR
- What are intentions? Suggested starting points: papers by Michael
Bratman & Michael Georgeff
- Bounded rationality. Suggested starting point: papers by Stuart
Russell
- A great book from the 1950's that lays out the foundations of AI: The
Sciences of the Artificial, by Herbert Simon.
- A great book from the 1950's that warns of the dangers of AI: The Human
Use of Human Beings, by Norbert Wiener
- Your own idea here
Programming Projects
- Build a "softbot" (a software agent) that you communicate with
by typing in speech acts (in your own formal notion), that then infers what
you want done and does it. The softbot may live in a simple simulated
world (e.g. the blocks world) or the Unix world (e.g., it has the ability to
run some commands such as "cp" and "lpr"). You
should look at the papers by Etzioni and Weld on the softbot project to get
some ideas of the issues that arise - but don't feel that you need to follow
their example.
- Even more ambitiously: do the above with a natural language
interface. Feel free to only consider a small subset of English, and
to steal parsers etc from the web.
- Switchboard is a large corpus of short dialogs that have been hand
annotated according to many different schemes. These include an
annotation according to speech acts. Investigate this coding scheme, and
try your favorite classification learning algorithms to see if you can do
better than chance at learning to classify speech acts. If you are
interested in doing this let me know ASAP so I can find a local copy of the
entire Switchboard corpus. I found a
small sample of 35 dialogs on the web here, where the
coding manual
is here.
- Put the microphone in your laptop to some use: build a system that
analyzes the sound it picks up to try to guess whether: (1) the room is empty;
(2) you are typing; (3) you are in a conversation.
- Even more ambitiously: use the above as part of a messaging system, that
interrupts you only if you are not in a conversation.
- The GUIDE project Don Patterson will talk about uses RFID tag information
to infer activities of daily living in the home. Create a similar
(simplified) system that could recognize some of the activities you carry out
in your office. I do not have a glove, but I have a iPAQ based RFID tag
reader and a set of tags I could lend you to get you started! GUIDE uses
pretty heavy duty probabilistic machine - your system might try something
simpler - e.g., edit distance.
- Write a program that creates original jokes.
- Your own idea here. (Almost) anything goes -
this is your chance to do something strange!