Human Performance Modeling - Essay Example

CogTool Introduction CogTool is a UI prototyping tool that evaluates design using a human performance model in predictions. A storyboard of preferred design idea created has images, sketches or functions by use of CogTool widgets. Tasks demonstrated on the storyboard are producing valid cognitive models after pressing a button. A prediction is made based on the duration taken by a skilled user to complete the task. The challenges faced by the software are its speed and accuracy. Modeling Several models created vary depending on the modelers involved in making human performance models. The types of modeling techniques used are AI and cognitive modeling. Examples of modeling are the following: Sisyphus, Project Halo, Ambr. Project and Predicting Cognitive Performance. The differences existing in the models depend on the person’s or team’s modeling approach. Predictions related to similar performing model take place because of a similar approach which remains unknown. Comparison of modelers relies on approaches used to predict time that is spent by skilled performers on the various user interfaces (UIs). Heuristic estimation techniques are compared to predictive human performance modeling approach, the Keystroke-Level Model (KLM). Method of Study The KLM approach was compared with data from 19 novice modelers in building two UIs tasks. The data was published originated from 8 novice modelers. The variance for the data collected was approximately 20%. This evaluator effect in Human-Computer Interaction (HCI) is revealed in various techniques of HCI. The 20% evaluator effect for all the techniques faced critics on use in KLM modeling technique because of its assumption that it predicted accuracy of 20%. The differences in modeler techniques are based on expected accuracy that emphasizes on behavior representation community. Attempts to reduce variation in modelers are provided for by tool-support for KLM analyses. The human-centered design (HCD) techniques used created CogTool, a tool to construct valid KLMs. Error Detection Detection for systemic errors was done on examination of eight KLM novice modelers. 87 operators were compared with inclusion of KLM created. Many common errors were detected. Overt steps essential for task operation were left out. Accurate follow up of the Ks, Ps, and Hs in KLMs was impossible, making the task unsuccessful. 88% of modelers experienced this error. 38% of the novice modelers included extra overt operators unnecessary for the task. The novice modelers experienced difficulties in applying Moran and Newell’s heuristics when placing M operators. This was characterized by uneven placing of Ms in the models. The errors discovered were analyzed through human centered design techniques that prompted design of CogTool (Bonnie). Results of Research: CogTool Project CogTool project allows UI designers to predict human performance modeling techniques analyze ideas in design from a quantitative approach before implementing ideas on resource investment. Error analysis mentioned was used as a guideline in design of CogTool project so that the errors identified could be eliminated to the utmost. Contextual inquiry was used to understand the challenges faced by modelers and the success of the project in suiting the workflow and culture of UI designers. Competitive analysis was used to assess what had been tried. Results obtained were considered in design of the project. Continuous analysis has made the tool of importance in today’s real-world design and evaluation processes. This success enables it to be taught to HCI, UI design and Human Factor students. Using CogTool The procedure for KLM in CogTool is different from KLM done by hand. UI design is modeled on a graphical storyboard placed on widgets other than being listed by overt operators in spreadsheets separated from UI design. The widgets are in frames, representing what users see as they precede with others tasks. The frames are connected through transitions drawn from a widget to another frame. The drawing represents the user’s action that is displayed on the screen. A specific task is demonstrated by the modeler on the storyboard. This creates KLM by demonstration. The tool creates ACT-R code, interprets it to predict execution time. Ms is placed consistently as outlined by Card, Moran and Newell’s heuristics. The project has therefore transformed the modeling process to a design one since modelers determine the type of widget to use in design but not make decisions on addressing errors. The challenge left is to examine whether CogTool has minimized differences in models by novice modelers. KLMs with CogTool A blended class of 101 students was used in analysis of CogTool. The students had technical, behavioral science and fine art design backgrounds. One student was disqualified from the group because he had experience on CogTool project. The total number of students (N) that participated in the analysis was 100. They were exposed to equal time for lecture, and CogTool demonstration. The students downloaded the software from the download link. Interfaces and tasks used were from a real-world situation: cataloging books and online sharing of collections. The students were given instructions for each task per interface. Another focus on the study was on predictions resulting from modeling process but not from the modeler’s misunderstanding of task procedures and interfaces. The 100 files were analyzed to obtain the origin of variance like decisions made by modelers that created several numeric predictions. The variance developed could probably have been from; missing required steps altogether, implementation of a transition rather than a left click, using a widget instead of a button and insertion of a wrong system response time. The result produced from the 100 novice modelers indicated 164 rows of possible errors, blanks indicating correct decisions, errors the modelers made and matrix of ones. Although the modelers had 164 opportunities to create errors, 3 never differed in the selection of the right model. 46 of the novice modelers committed errors rated at 5%, 27 made 5-10% errors while the other quarter 10-20% possible mistakes. From the study, the mean and median for error probabilities were 7% and 6% respectively. The CogTool keeps track of the hand and inserts done between the mouse and keyboard. This avoids errors in the model (Bonnie). Placing of Ms automatically depends on widget choices. The automatic task by CogTool makes it user friendly to modelers. Of 1900 tasks completed, only 100 had errors. However, the errors do not affect predictions by CogTool. The implication about the errors is that they are style errors but not those of substance. The project indicates clear cut differences between text boxes and text inside the textboxes. This distinction influences the predictions of the tool. The user guide clearly states ways and differences between widgets. The error committed by modelers is as a result of ignorance. Two widgets proved to be difficult because they never mapped directly to widgets recognized by CogTool. The modelers were confused and as a result, 29 of them repeated 58 errors. The error was based on the modelers’ ability to articulate their decisions from right decisions. High chances of errors in relation to interaction styles still exist even though new interaction designs are designed. Further research to design an innovative program for CogTool in minimizing this error is needed (Bonnie). Conclusion Most errors experienced by modelers through KLM by-hand have been reduced by CogTool. The reason behind this success is due to reduced variability in models. HCD methods are used, that is, Contextual Inquiry, error analysis, usability evaluation among others, which form the basis of design of the CogTool. Work Cited Bonnie, John, E. Reducing the Variability between Novice Modelers: Results of a Tool for Human Performance Modeling Produced through Human-Centered Design, 2009. Web. 3rd Apr. 2012. . Read More