| Summary: | 碩士 === 國立成功大學 === 資訊管理研究所 === 103 === SUMMARY
Interface design of mobile computing devices has become a new issue on human-computer interaction(HCI) field. This study has two purposes, one is to explore usability, user experience and the link between interface and functionality. The other is to use the cognitive model evaluate mobile app interface, and explore if model data and subjects data are consistent or not. This study chooses the ACT-R cognitive model to evaluate that subjects operate GPS navigation APP by using a smart phone. The results of experiment 1, subjects have different user experience(UX) whether researcher give tips or not. The results of experiment 2, the prediction of touch action is good forecasting, but the prediction of sliding action is not.
Key words: ACT-R, cognitive model, GPS navigation APP.
INTRODUCTION
This study has two purposes, one is to explore usability, user experience and the link between interface and functionality. Subjects should evaluate the degree of usability of APP by writing the questionnaire subjectively. The other is to use the cognitive model evaluate mobile app interface, and explore if model data and subjects data are consistent or not. This study desired UI designers could use the model to evaluate draft. Designers could determine abandoning or continue developing through evaluating the draft by model. So it will reduce the risk of development failure. Therefore, this study chooses the ACT-R cognitive model to evaluate that subjects operate GPS navigation APP by using a smart phone.
In this study, 18 subjects were recruited. They should operate three GPS navigation APPs twice, and write SMEQ, PSSUQ and questionnaire this study made at each time. First time researcher would say nothing, and second time the researcher will tell subjects the best method. And then, subjects should repeat the best method 50 times for three GPS navigation APPs, and this data is used to compare to model data.
The results of experiment 1, subjects have different user experience(UX) whether researcher give tips or not. If researcher gives tips, subjects would get better UX. The results of experiment 2, this study uses MAPE to be as prediction accuracy. The MAPE of three APPs are 9.4%(high accuracy forecasting), 13.8%(good forecasting) and 20.9%(reasonable forecasting).
MATERIALS AND METHODS
Three APPs this study used are Google Maps, Garmin Taiwan and Mio Map. Subjects only could use right hand thumb to touch or slide screen. They couldn’t type or use two fingers. All the subjects should do both experiment.
This study used CogTool (1.2.2 version) based on ACT-R 6 to predict the execution time. CogTool is a prototyping tool, and it can predict the behavior of users. ACT-R model is developed by Anderson (1983). It is a cognitive architecture including visual attention, and motor movement and has been the basis for a number of models in HCI. So researcher load all interfaces of three APPs, and link them by setting touch or sliding action. CogTool has a step named “Think Step”, researcher set it 0.4 minutes. All the setting is already, then execute the prediction and get the prediction time.
In experiment 1, this study used MIXED procedure of SAS 9.3to run two-way ANOVA analysis, and selected α = 0.05 as statistical significance, and use the least significant difference procedure(LSD) as a post-test approach.
In experiment 2, this study used mean absolute error(MAE), root-mean-square error(RMSE) to be measures of error, and used mean absolute percentage error(MAPE), symmetric mean absolute percentage error(sMAPE), mean absolute scaled error(MASE) to be measures of error rate.
RESULTS AND DISCUSSION
The result of experiment 1, Figure 1 shows APPs have significant SMEQ scores differences in pre and post task. In pre section, Google Maps and Mio Map have significant SMEQ scores differences. Garmin Taiwan and Mio Map have significant SMEQ scores differences, too.
In pre section, subjects spend much time to find function which could complete tasks, but too much time let subjects get high mental workload. So SMEQ score is high in pre section.
Mio Map does not provide street view so subjects do not find the assigned street view. Because of that, Mio Map task is easier than Google Maps and Garmin Taiwan. So subjects feel lower mental workload when using Mio Map.
Figure 1. SMEQ score
The result of experiment 2, Figure 2 shows Model prediction and subjects data by Google Maps task. The MAE is 1102ms, RMSE is 1480ms, MAPE is 9.4%, sMAPE is 9.5%, and MASE is 29.1%
Figure 2. Model prediction and subjects data(Google Maps)
Figure 3 shows Model prediction and subjects data by Garmin Tiawan task. The MAE is 2192ms, RMSE is 2748ms, MAPE is 13.8%, sMAPE is 13%, and MASE is 60.4%.
Figure 3. Model prediction and subjects data(Garmin Taiwan)
Figure 4 shows Model prediction and subjects data by Garmin Tiawan task. The MAE is 2198ms, RMSE is 2602ms, MAPE is 20.9%, sMAPE is 18.6%, and MASE is 62.5%.
Figure 4. Model prediction and subjects data(Mio Map)
Model predicting Google Maps task is the best prediction of three APPs. The possible reason researcher thought are action types and action numbers. The mainly action type of Google Maps task is touch, of Mio is sliding. So researcher infers that model could predict touch action more exactly.
CONCLUSION
In this study, researcher has described a set of evaluation concepts and tools to predict using APPs time. Model has good forecasting in touch action, reasonable forecasting in sliding action. Prediction time could provide usability information so that researcher find out usability information from experiment 1. Google Maps gets the best prediction and the least time, but its icon desired not well. Garmin Taiwan gets the second best, but its items of lists is too much to find the goal. Mio gets the worst prediction and the second least time, but its icon is not bad to users.
|
|---|
