- Email: [email protected]
An effectiveness study of a CAD system augmented by audio feedback
Comput & Graphics, Vol. 2. pp. 231-233. Pergamon Press. 1977 Printed in Great Britain
AN EFFECTIVENESS STUDY OF A CAD SYSTEM AUGMENTED BY AUDIO FEEDBACK JOHN J. ALLAN lilt and ANTHONY M. CHIC* Center for Special Studies. Southern Methodist University, Dallas, TX 75275, U.S.A.
(Received 14 July 1976) Abstract--The question addressed is the effectiveness of man-computer communicationin computer-aideddesign (CAD). Many physical systems that design engineers develop on interactive CAD systems, in reality, produce noises that would give information about the system's performance. The results of a controlled man--computerstudy are presented here. The task for the subjects was to design the acoustical treatment for a noisy laboratory. A control group worked in a conventional manner with the CAD system. The study groupreceivedcontinuousaudio feedbackindicativeof the degree of noisereductionof their design. The statistically significantresults are presented. ~rrRomJc'rtON There is no CAD system which can be called the nearoptimal design system. One reason for this is the interface problems between man and machine. One of these problems, man-machine communication channel availability, was the focal point of investigation of this research. The two common reasons for the computer being an essential part of design are massive data storage ability and computing speed. Other less obvious reasons for the importanceofcomputingindesignhavebeendiscussed[1], One of these reasons is the improvement in communication between the designer and his problem.This work focuses on what effects an extra communication channelhas during CAD activity, CAD systems often use an interactive graphics console for man-computer communication. As reported to the International Institute for Applied System Analysis[2], interactive graphics is currently used in two general areas: visual scanning of data, and the input of design data. In most current CAD systems, the computer's ability to transmit information is limited to alphanumeric characters and line graphics. Information transmitted in this manner utilizes the eye as the information reception organ. Other physical outputs, such as motion, heat, or sound can also be controlled by the computer if a proper interface is used. Because it is possible to implement multi-channel input and output, future man--computer communication should not be limited to touch input from the man and visual output from the machine,
can be disturbed by sound even though he does not voluntarily want to hear it. Therefore. most warning systems use audio signals in addition to visual signals as a means of communication. As pointed out by Grandjean, "acoustic stimuli (noise and speech) detract more than any other sensory stimulus" [3]. One part of the solution to the problem of mancomputer communication is to increase the number of channels of communication. The visual channel has been in use for a long time. Various techniques have been developed to expand the channel capacity of the eye. Some written language can be replaced by line graphics; some line graphics can be replaced by real-time, continuously moving pictures. Nevertheless, a designer sitting in front of a CAD terminal cannot be looking at the picture and reading written instructions on his desk at the same time. Hence, most interactive graphic CAD terminals are designed to present one concept at a time. This limitation is still unnecessarily imposed on CAD systems. And yet, until now the ear has not been investigated as an information reception device in CAD systems.
rm~ SYS'mMm'ECTIV~mSS STL~Y This section discusses the effect of an extra communication channel on the designer: the audio feedback channel from machine to man. An experiment was designed to compare the responses of designers to different systems. Comparison was made between an interactive CAD system augmented by audio feedback and a system without audio feedback (see Fig. 1). The subjects of study were divided into two groups. All subjects were assigned the same task. Both groups AVmOlnmaSACXntOM r,t~Cm~ TO MAN were allowed to wear the headphones during the exAudio feedback is observable everywhere in daily life; periment, but only one group could get audio feedback. for example buzzing when the ignition key is forgotten. The design problem was to design the acoustical It is there because of its interference ability. A person treatment for a noisy computer room. The noise level in the actual room was very high [68 dB(A)]. Partitions had been installed in the room as means of isolating and tDr. Allan is Professor of Mechanical Engineering,Director, absorbing part of the noise in the room. Noise reduction Center for SpecialStudies, Southern MethodistUniversity,Dallas, in the room depended strongly on the orientation of the TX 75275, U.S.A. *Mr. Chiu is Design Engineer, Didde-GlazerInc., Emporia, KS partitions. 66801, U.S.A. NOW, keep in mind that this was a man-computer 231
232
J.J. ALLANIII and A, M. CI-IIU
I I
b
TM
• ~ "x,.~th . . . . .
al
~
/
~
"
.
any noise was not an indication of the designed noise reduction. In a pilot study, noise from the headphones was produced by an audio amplifier. The gain of the amplifier output was controlled by a light sensitive variable resistor. A light emitting diode was used as the light source, which was in turn controlled by the analog signal from the NOVA's D/A converter. However, such a noise source was not an effective source of information. In the second study, a computer-controllable function generator was used as the indicator of various noise levels. were:Thevariables to be investigated in this experiment
(I) The response of the designer--measurement was Fig. 1. Designerwith acoustic feedback at the start of the design done on the designer's response by counting the number task. of DELETE commands, the number of INSERT commands, and the time taken for the final design. interface study, and not an exercise in acoustical design. (2) The final design of the room--noise reduction and A detailed mapping of the room's noise level was per- cost of installation. formed (see Fig. 2). The room was divided into 72 locations (it could have been finer, but the size of the aF.st~rs ~ cowcLoslows room was too small to effectively divide any further). The pilot study The partitions had to be placed by the designer within a A pilot study with 27 subjects was conducted. Each certain distance of the computer, so that the partition subject was observed closely during his twenty minutes could affect the noise level when inserted. When one of of interaction with the computer. The responses of each the partitions was deposited in one of the eight grid subject were recorded individually for future reference. rectangles nearest to a computer, the cost was calAfter screening the data, only 18 data sets were used culated, for statistical analysis. Of the 18 data sets analysed (8 In the model, the highest noise level was 66 dB, and from the audio feedback group, and 10 from the control the lowest noise level was 46 dB (noise from air con- group), there was a slight increase in the number of ditioning, lights and outside .traffic). interactions and a decrease in final noise level by the The experiment required the use of two minicomputers: group whose headphones were active. But the results a NOVA 840 and an IMPLAC PDS-I. The latter served were not statistically significant. as the interactive interface between man and computer. Those who had used the headphones in the pilot study The designer could initiate 6 commands through the use agreed that: of the light pen. The NOVA was placed in an idle loop (1) they were more cautious about cutting the noise while the graphics manipulation was being performed in level down than the cost of installation because they the IMLAC. At the command of the IMLAC, the could tell the merit of a noise reduction through the NOVA was activiated to do the necessary acoustical and headphones; cost computation and return the computed results to the (2) the eyes experienced less fatigue because they designer, could use the headphones as a source of information One group of the designers was allowed to use the instead of staring at the CRT continuously. headphones as a source of noise reduction information, The data collected from the pilot study could not in addition to the numerical display on the CRT. The statistically identify the audio augmented group from the other group was also allowed to use the headphones, but control group. The basic problem was the hardware which generated the noise. Also, it was decided to require that: 65 (I) persons who were used to working in a noisy 65 environment should not be chosen as subjects of study in 6a the second study; 6s 65 65 [ ~ 7 ] (2) prospective subjects should be briefed only in front of the terminal to avoid pre-conception before 6,*.5 approaching the design problem; (3) subjects should be informed that they are taking 6 ~ 64.~ part in a study, not a test; they are the designers, and they are to evaluate a computer system, but not to be 71 evaluated. 66 Also, instead of using dollar cost and noise level 67 64.s 62 6~.5 reduction, an index of performance was established in 6866ff , , order to avoid the emotional reaction of a subject to Fig..2. Noise level measurements, cost. Some subjects said that the cost was too high even
An effectiveness study of a CAD system augmented by audio feedback though the noise reduction was considered optimum, They had even stopped inserting partitions which were necessary for a good design, The second study Alterations were made in the experiment for the second study. There were new results from these changes. The changes are discussed below, (1) The allowable time limit was reduced. There was very little learning shown by the subjects in the pilot study after the first 15 minutes, (2) The indicative noise source was changed to a constant volume, variable frequency tone. (3) A performance index was used, and was calculated from the number of partitions in the design, the total number of allowable partitions, the original noise level without partitions, the final noise level with partitions, the highest noise level measured, and the lowest
noise level measured in the actual room that had been modelled for these studies, CONCLUSIONSANDRECOMMENDATIONS Analysis of the mean of the dependent variables reflected some significant results for this experiment. The noise levels arrived at in the design by the two groups were significantly different, Another factor, number of deletions/number of insertions, which measures the frequency of decision error, was found to be significant. Since such differences were not observable in the pilot
233
study it may be concluded that, with the knowledge from the second study, the audio feedback channel can be an effective communication channel in a CAD system. One drawback of this research is the lack of theoretical foundation. Even though the experimental data seem to indicate a certain direction, the extent of such system effectiveness cannot yet be generalized. This generalization will be an important factor in the design of future CAD systems. It is recommended that future study involve the use of a prediction model, so that the quality of information can be related to the number and types of communication channels used. Future research should also identify the classes of design problems which need. and do not need, audio feedback.
t~_.~NC,V.S 1. John J. Allan lII, Foundations of the many manifestations of computer augmented design, Proc. IFIP Working Conference Principles of CAD, October 1972, pp. 27-54. North-Holland, Amsterdam. 2. J. Hatvany, W. M. Newman & M. A. Sabin, Report to IIASA on a World Survey of Computer-Aided Design. International Institute for Applied System Analyses, Laxenburg, Austria (July 1974). 3. E. Grandjean, Fitting the Task to the Man, An Ergonomic Approach.Taylor & Francis, London (1971). 4. Anthony M. Chiu, An effectiveness study of a CAD system augmented by audio feedback. Master's thesis, Department of Mechanical Engineering,The University of Texas at Austin (1976).
AN EFFECTIVENESS STUDY OF A CAD SYSTEM AUGMENTED BY AUDIO FEEDBACK JOHN J. ALLAN lilt and ANTHONY M. CHIC* Center for Special Studies. Southern Methodist University, Dallas, TX 75275, U.S.A.
(Received 14 July 1976) Abstract--The question addressed is the effectiveness of man-computer communicationin computer-aideddesign (CAD). Many physical systems that design engineers develop on interactive CAD systems, in reality, produce noises that would give information about the system's performance. The results of a controlled man--computerstudy are presented here. The task for the subjects was to design the acoustical treatment for a noisy laboratory. A control group worked in a conventional manner with the CAD system. The study groupreceivedcontinuousaudio feedbackindicativeof the degree of noisereductionof their design. The statistically significantresults are presented. ~rrRomJc'rtON There is no CAD system which can be called the nearoptimal design system. One reason for this is the interface problems between man and machine. One of these problems, man-machine communication channel availability, was the focal point of investigation of this research. The two common reasons for the computer being an essential part of design are massive data storage ability and computing speed. Other less obvious reasons for the importanceofcomputingindesignhavebeendiscussed[1], One of these reasons is the improvement in communication between the designer and his problem.This work focuses on what effects an extra communication channelhas during CAD activity, CAD systems often use an interactive graphics console for man-computer communication. As reported to the International Institute for Applied System Analysis[2], interactive graphics is currently used in two general areas: visual scanning of data, and the input of design data. In most current CAD systems, the computer's ability to transmit information is limited to alphanumeric characters and line graphics. Information transmitted in this manner utilizes the eye as the information reception organ. Other physical outputs, such as motion, heat, or sound can also be controlled by the computer if a proper interface is used. Because it is possible to implement multi-channel input and output, future man--computer communication should not be limited to touch input from the man and visual output from the machine,
can be disturbed by sound even though he does not voluntarily want to hear it. Therefore. most warning systems use audio signals in addition to visual signals as a means of communication. As pointed out by Grandjean, "acoustic stimuli (noise and speech) detract more than any other sensory stimulus" [3]. One part of the solution to the problem of mancomputer communication is to increase the number of channels of communication. The visual channel has been in use for a long time. Various techniques have been developed to expand the channel capacity of the eye. Some written language can be replaced by line graphics; some line graphics can be replaced by real-time, continuously moving pictures. Nevertheless, a designer sitting in front of a CAD terminal cannot be looking at the picture and reading written instructions on his desk at the same time. Hence, most interactive graphic CAD terminals are designed to present one concept at a time. This limitation is still unnecessarily imposed on CAD systems. And yet, until now the ear has not been investigated as an information reception device in CAD systems.
rm~ SYS'mMm'ECTIV~mSS STL~Y This section discusses the effect of an extra communication channel on the designer: the audio feedback channel from machine to man. An experiment was designed to compare the responses of designers to different systems. Comparison was made between an interactive CAD system augmented by audio feedback and a system without audio feedback (see Fig. 1). The subjects of study were divided into two groups. All subjects were assigned the same task. Both groups AVmOlnmaSACXntOM r,t~Cm~ TO MAN were allowed to wear the headphones during the exAudio feedback is observable everywhere in daily life; periment, but only one group could get audio feedback. for example buzzing when the ignition key is forgotten. The design problem was to design the acoustical It is there because of its interference ability. A person treatment for a noisy computer room. The noise level in the actual room was very high [68 dB(A)]. Partitions had been installed in the room as means of isolating and tDr. Allan is Professor of Mechanical Engineering,Director, absorbing part of the noise in the room. Noise reduction Center for SpecialStudies, Southern MethodistUniversity,Dallas, in the room depended strongly on the orientation of the TX 75275, U.S.A. *Mr. Chiu is Design Engineer, Didde-GlazerInc., Emporia, KS partitions. 66801, U.S.A. NOW, keep in mind that this was a man-computer 231
232
J.J. ALLANIII and A, M. CI-IIU
I I
b
TM
• ~ "x,.~th . . . . .
al
~
/
~
"
.
any noise was not an indication of the designed noise reduction. In a pilot study, noise from the headphones was produced by an audio amplifier. The gain of the amplifier output was controlled by a light sensitive variable resistor. A light emitting diode was used as the light source, which was in turn controlled by the analog signal from the NOVA's D/A converter. However, such a noise source was not an effective source of information. In the second study, a computer-controllable function generator was used as the indicator of various noise levels. were:Thevariables to be investigated in this experiment
(I) The response of the designer--measurement was Fig. 1. Designerwith acoustic feedback at the start of the design done on the designer's response by counting the number task. of DELETE commands, the number of INSERT commands, and the time taken for the final design. interface study, and not an exercise in acoustical design. (2) The final design of the room--noise reduction and A detailed mapping of the room's noise level was per- cost of installation. formed (see Fig. 2). The room was divided into 72 locations (it could have been finer, but the size of the aF.st~rs ~ cowcLoslows room was too small to effectively divide any further). The pilot study The partitions had to be placed by the designer within a A pilot study with 27 subjects was conducted. Each certain distance of the computer, so that the partition subject was observed closely during his twenty minutes could affect the noise level when inserted. When one of of interaction with the computer. The responses of each the partitions was deposited in one of the eight grid subject were recorded individually for future reference. rectangles nearest to a computer, the cost was calAfter screening the data, only 18 data sets were used culated, for statistical analysis. Of the 18 data sets analysed (8 In the model, the highest noise level was 66 dB, and from the audio feedback group, and 10 from the control the lowest noise level was 46 dB (noise from air con- group), there was a slight increase in the number of ditioning, lights and outside .traffic). interactions and a decrease in final noise level by the The experiment required the use of two minicomputers: group whose headphones were active. But the results a NOVA 840 and an IMPLAC PDS-I. The latter served were not statistically significant. as the interactive interface between man and computer. Those who had used the headphones in the pilot study The designer could initiate 6 commands through the use agreed that: of the light pen. The NOVA was placed in an idle loop (1) they were more cautious about cutting the noise while the graphics manipulation was being performed in level down than the cost of installation because they the IMLAC. At the command of the IMLAC, the could tell the merit of a noise reduction through the NOVA was activiated to do the necessary acoustical and headphones; cost computation and return the computed results to the (2) the eyes experienced less fatigue because they designer, could use the headphones as a source of information One group of the designers was allowed to use the instead of staring at the CRT continuously. headphones as a source of noise reduction information, The data collected from the pilot study could not in addition to the numerical display on the CRT. The statistically identify the audio augmented group from the other group was also allowed to use the headphones, but control group. The basic problem was the hardware which generated the noise. Also, it was decided to require that: 65 (I) persons who were used to working in a noisy 65 environment should not be chosen as subjects of study in 6a the second study; 6s 65 65 [ ~ 7 ] (2) prospective subjects should be briefed only in front of the terminal to avoid pre-conception before 6,*.5 approaching the design problem; (3) subjects should be informed that they are taking 6 ~ 64.~ part in a study, not a test; they are the designers, and they are to evaluate a computer system, but not to be 71 evaluated. 66 Also, instead of using dollar cost and noise level 67 64.s 62 6~.5 reduction, an index of performance was established in 6866ff , , order to avoid the emotional reaction of a subject to Fig..2. Noise level measurements, cost. Some subjects said that the cost was too high even
An effectiveness study of a CAD system augmented by audio feedback though the noise reduction was considered optimum, They had even stopped inserting partitions which were necessary for a good design, The second study Alterations were made in the experiment for the second study. There were new results from these changes. The changes are discussed below, (1) The allowable time limit was reduced. There was very little learning shown by the subjects in the pilot study after the first 15 minutes, (2) The indicative noise source was changed to a constant volume, variable frequency tone. (3) A performance index was used, and was calculated from the number of partitions in the design, the total number of allowable partitions, the original noise level without partitions, the final noise level with partitions, the highest noise level measured, and the lowest
noise level measured in the actual room that had been modelled for these studies, CONCLUSIONSANDRECOMMENDATIONS Analysis of the mean of the dependent variables reflected some significant results for this experiment. The noise levels arrived at in the design by the two groups were significantly different, Another factor, number of deletions/number of insertions, which measures the frequency of decision error, was found to be significant. Since such differences were not observable in the pilot
233
study it may be concluded that, with the knowledge from the second study, the audio feedback channel can be an effective communication channel in a CAD system. One drawback of this research is the lack of theoretical foundation. Even though the experimental data seem to indicate a certain direction, the extent of such system effectiveness cannot yet be generalized. This generalization will be an important factor in the design of future CAD systems. It is recommended that future study involve the use of a prediction model, so that the quality of information can be related to the number and types of communication channels used. Future research should also identify the classes of design problems which need. and do not need, audio feedback.
t~_.~NC,V.S 1. John J. Allan lII, Foundations of the many manifestations of computer augmented design, Proc. IFIP Working Conference Principles of CAD, October 1972, pp. 27-54. North-Holland, Amsterdam. 2. J. Hatvany, W. M. Newman & M. A. Sabin, Report to IIASA on a World Survey of Computer-Aided Design. International Institute for Applied System Analyses, Laxenburg, Austria (July 1974). 3. E. Grandjean, Fitting the Task to the Man, An Ergonomic Approach.Taylor & Francis, London (1971). 4. Anthony M. Chiu, An effectiveness study of a CAD system augmented by audio feedback. Master's thesis, Department of Mechanical Engineering,The University of Texas at Austin (1976).