John L. Pollock
Department of Philosophy
University of Arizona
Tucson, Arizona 85721
This is to be done in your three-member groups. It is a combination coding and writing assignment. The general task is to construct some interesting scenario involving a combination of perception, temporal projection, and causal reasoning, and encode it and run it using MAKE-SIMULATION-PROBLEM. In clear prose, describe the problem and explain what you think the right answer is and why. Then run it in OSCAR and compare OSCARÕs answer with yours. If they differ, and you still think your answer is right, try to explain where OSCAR goes wrong.
The general form of MAKE-SIMULATION-PROBLEM is as follows:
"First, Fred looks red to me. Later, I am informed by Merrill that I am then
wearing blue-tinted glasses. Later still, Fred looks blue to me. All along, I know that the
probability is not high of Fred being blue given that Fred looks blue to me, but I am
wearing blue tinted glasses. What should I conclude about the color of Fred?"
The above should be a brief explanation of the problem, enclosed in quotation marks.
The reason-schemas used in the problem. You can use any of the reason-schemas contained in the file PERCEPTION-CAUSES_3.31a, and you can make up new ones if your problem requires them. For that purpose, use the examples in PERCEPTION-CAUSES_3.31a for guidance. I did that myself, for example, in problems 9 – 11.
(1 "(the color of Fred is red)" 0.8)
(20 "(Merrill reports that I_am_wearing_blue_tinted_glasses)" 1.0)
(30 "(the color of Fred is blue)" 0.8)
The inputs have the form (time-of-input input-formula strength). The time-of-input is an integer, input-formula is a string (i.e., enclosed in quotation marks), and strength is a number between 0 and 1 that gives us the degree of justification associated with the conclusion (drawn by PERCEPTION) that the input-formula is true.
In OSCAR, times are measured by cycles of reasoning. If there is an input whose time is n, then at the nth cycle OSCAR receives that as a new premise.
("((the probability of (the color of Fred is blue) given
((I have a percept with content (the color of Fred is blue)) & I_am_wearing_blue_tinted_glasses)) <= .8)"
("(Merrill is a reliable informant)" 1.0)
The premises are Ōbackground informationĶ about the problem situation. This is where you put any probability information needed by the problem, or other general information.
("((the color of Fred is red) at 50)" 0.55)
Interests (there can be more than one) are of the form (interest-formula interest-strength). Interest-formula must be in quotation marks. Interest-strength is a number between 0 and 1. If you are having trouble getting OSCAR to draw the conclusion you expect, it may be because you have set the interest-strength too high. Conclusion strengths decay by temporal projection, so if the time intervals involved in the problem are large, the conclusion strengths may become weaker than you realize. You can turn (display-on) to monitor the reasoning and see whether it is working as expected but getting conclusions that are too weakly justified to discharge the interests.
DonÕt forget the closing parentheses. In LISP, everything is enclosed in parentheses, and the number of right and left parentheses must match.
Once you have formulated your problem, you load it into OSCAR by putting it into a file and loading the file. Then run (simulate-oscar n) where n is the number you chose for the problem-number (probably 1).
Turn in your results by email to email@example.com by Monday, February 14. The prose part can be a Microsoft Word file, an rtf file, or a pdf file. Attach a complete run of OSCAR doing your problem with (display-on). Note that if you have trouble with ACLÕs small display buffer, you can have ACL save the contents of the display to a file using the command (dribble filename). This saves new display contents, not anything already in the buffer. (dribble) turn that off when you are through. With LISPs other than ACL, it will probably not be necessary to use DRIBBLE.