Protocol Analysis of Man-Computer Languages: Design and Preliminary Findings [report]

John F. Heafner
1975 unpublished
This document approved for public release and sale; distribution unlimited. IT DISTRIBUTION STATEMENT fei ihm mbtirtci anfarad In Slack 20, (/ rf/ffaranf froai Rmpa'i) H SURRLEMENTARY NOTES I» KEV WORDS rCanHmia en rwraraa •/<*• II nacaaaarr ""< (dam/fr by Mock number) application-oriented language design, man-computer language design, man-machine communication, message proce»sing, military message processing, protocol analysis, statistical analysis of computer languages, user performance. }0
more » ... STRACT (Canllm-an ravaraa «id» II nacaaaary and (dam/fj-br Mac* miftarj (OVER) DO | J AN^l 1473 EDITION OF 1 NOV «1 IS OBSOLETE S/N 0102-014-SeOI |. UNCLASSIFIED SECURITY CL A»»|FIC ATiON OP THIS PAGE r'tian Data Bnlararf) ■*-1 ■ p-UNCLASSIFIED »CU«ITV CLAMIFICATlON OF THIS PAOCfWlMn OaM CRMfMO ABSTRACT This report describes an on-gomg study in man-machine communications. The study's main premise is that in developing man-computer languages one should consider the users 1 needs and habits as well as features of the computer service. The problem in doing so is that the designer does not have sufficient quantitative information about the users to enable him to specify languages permitting near-optimal performance. The study proposes and tests a method to achieve a closer fit between users and their computer languages by involving potential users m the design process. Token languages of several syntactic forms are defined. Then, -esearch hypotheses are stated concerning the users' perferences ragardmg the language s ructure and vocabulary. Next, an experiment design is described, based on a statistical mudei of observations of commands entered by users as they perform a standardized task. The method is tested by protocol analysis with subjects who are potential users. In the protocol analysis, subjects vocally stated commands in each of the token languages as they performed the standardized task. These respondents were requested to change the granmar of each language (during the task) to make it natural for them to use. Their task inputs were used to test the hypotheses. The report concludes that the method of modelling users and then testmg draft languages is useful m language design, since th?re was a consensus of users' opinions as to specific language improvements.
doi:10.21236/ada013568 fatcat:ay274od7ynhsjexn5fav4jktjy