Workshop on the Psychological Reality of Lisp

Yale University U.S.A

11-12 March, 1981

Report by: R.M. Duck Lewis

The first Workshop on the Psychological Reality of LISP was recently held at Yale University under the sponsorship of the Sloan Foundation, with about twenty invited participants. The Workshop was conceived and organized by Roger Shank.

The program consisted of informal presentations by some of the participants, each followed by long and often lively discussion. A brief description of the program follows:

ROGER SHANK opened the workshop with an overview of his own research, in which an NLU program is treated as a literal theory of human language understanding. He had slowly begun to realize, he said, that this approach could only make sense if one made the additional assumption that there was an underlying mental reality to LISP, and he therefore began research to find proof of this reality. He became aware of other researchers doing the same, and conceived this workshop to promote an interchange of ideas on the topic.

JAMES RODERICKSON: `Evidence for the mental reality of cons cells.' Roderickson described a series of experiments he and his colleagues had performed in a attempt to show that human memory representations are based on cons cells and pointers. Subjects were given simple list manipulation problems and reaction times in choosing the correct answer from an array of four choices were measured. The results were in accord with the prediction that time taken on a problem would be proportional to the number of LISP conses that would be required to do the problem. One surprise was that in some cases a subject would take longer than predicted; it was assumed that this was due to garbage collection of used cons cells, and these data points were therefore not included in the analysis.

NORMA TINSTEIN: `Evidence for the mental reality of pointers.' After the lunch break, Roderickson's colleague Tinstein continued the report on the above mentioned experiments. Given the reality of cons cells, it was then necessary to show the reality of pointers. This was investigated indirectly, through a study of the problems of broken lists, dangling pointers and circular lists. For example, a subject would be asked to perform a rplaca or rplacd which would result in a circular list. When asked to say what the list was, subjects would go into an infinite loop, which, in one extreme case, could only be halted by a blow from the experimenter. Regrettably, the subject was hospitalized for two days, and the University's Human Experimental Subjects Review Board terminated the experiments before adequate data could be generated.

MARTIN HAMMOND, a speech therapist from New York, spoke for half an hour on the `The psychological ontogeny of sibilant speech impediments' before anyone realized that he was talking about the psychological reality of a different type of lisp; whereupon he was removed from the workshop by the security guards.

ELLEN JOHNSON: `MACLISP or Stanford LISP?' Johnson considered the question of which of the many dialects of LISP now is use was closest to the mental LISP. At present, only intuition and anecdotal evidence are available, but the indications are that MACLISP is closer to real life. Matters considered include whether taking the car and cdr of nil result in an error, or yield a value of nil, and whether the default number base is octal or decimal. The cardinality of the chiral philanges is considered significant in the second question.

H. TOUCAN: `A mental model of errors in LISP programming.' According to Toucan, the three most common errors in LISP programs are (1) an incorrect number of arguments to a function; (2) trying to take the car or cdr of an atom; and (3) attempting to evaluate an unbound atom that should have been quoted but wasn't. Toucan described a model of the human mind as the execution of a LISP program in which mental errors are explained as instances of these LISP programming errors. For example, the oft studied `tip of the tongue' phenomenon, when one is unable to complete the retrieval of a word from the mental lexicon, is caused by a type 3 error: a missing quote mark at the time when the word is initially entered in the lexicon. A surprising consequence of the model is that the human mind has a 44 1/2 bit address space.

The final conference session was an evaluation of the workshop, and discussion of future plans. It was decided that another workshop should be held in two years. In the meantime, an elected committee would look into the possibility of organizing a formal society, possibly know as the Association for the Reality of Artificial Intelligence. In addition, Academic Press has expressed an interest in publishing a journal tentatively named Cognitive Programming; a committee to liase with the publisher was elected.

The social program was just as successful as the rest of the workshop. Instead of the usual boring conference banquet, Schank organized a game of touch football (Home v. Visitors), which the Yale team won 14-9.

                                                                                                                             

Copyright 1981: AISE Quarterly, Issue 40-41, Spring Summer 1981