Presentation Notes from 11/9, Melissa Kay House Beth Meyer: Adaptive Performance Support: User Acceptance of a Self-Adapting System Introduction RUA = Retail user assistant adaptive help for use of existing retail point of sale device ie. cash register in department stores RUA-assistance on a display that was added to the existing system uses a task model to show the cashier exactly how to perform each step of the operation layaway point-of-sale discount detects common errors & guides cashier through correcting them it's both textual and video cashier can go to the next step while the video is still playing, though RUA can adapt to the needs of the user as the cashier gets more familiar with the system and makes fewer mistakes it adapts the help goes from more detailed to less detailed as the user shows he knows what's going on or becomes more confused Design Issues and Constraints intended to test the RUA in a live retail environment didn't want negative transfer ie. didn't want to teach people things that they wouldn't use on the same systems once the testing was over couldn't change any existing procedures or operations this reduced the ways the RUA could act as an intelligent agent gave users info on how to correct errors couldn't correct the errors for the users couldn't interrupt users to ask the OK to adapt the displays this was an interruption in the working environment and could cause problems also couldn't add any tasks to the user's work So the system was self adaptation controlled and implemented entirely by the system no input from the user risky design decision 'cause this can make the user feel frustrated and out of control Evaluation Results evaluated in laboratory usability tests, demonstrated in a store training room wasn't field tested due to changes in the field testing environment observed new 7052 users as they did a lot of transactions three different groups, five people in each only; no aiding and RUA with aiding fixed at the level of greatest detail and RUA with fully adaptive aiding each person performed eight trials of each of these transactions cash sales with dollar amount markdowns cash salkes with percentage markdowns cash sales to employees check sales cash returns layaway sales generally, performance times of the 7052 only group were faster it may be surprising that additional help slowed users down this just showed that they were USING the help shows the effectiveness of the error detection features the users were better able to detect errors with the RUA but then they'd spend time correcting them only group didn't find errors and thus didn't spend the time to correct them RUA groups made fewer errors than people in the 7052 only group RUA improved the accuracy, but at some cost to initial speed however, as the help level adapted to give less help, performance time dropped dramatatically the fixed level RUA group performance times tapered off more gradually user performance information both RUA groups liked the RUA fully adaptive RUA groups adapted displays without user control unanimously favorable in opinion of RUA subjects ether didn't remember changes in the displays or stated that the changed didn't affect their performance fixed level RUA one participant complained that the videos were not useful once he had learned how to use the system this is very interesting 'cause users generally don't like self adaptation by the system at all, but they liked it here Discussion Need to do more research with more subjects with a greater number of trials however, adaptation of the RUA did not disrupt performance and was accepted by the users very different from previous research done on adaptive interfaces when the user didn't have reasons for designing the RUA like this were practical constraints not allowed to give the users significant additional tasks when there's a high workload, giving the user more responsibility can have negative effects on performance USER CONTROL SEEMS MORE IMPORTANT FOR SOME TYPES OF ADAPTATION THAN FOR OTHERS Why was self-adaptation acceptable to RUA users when i t was not acceptable to users of other systems?-Previous Evaluations of Self-Adapting Systems in some systems, when the system adapted the user's strategy for using the system had to be completely changed one system (study by Hockley) that had varying levels of user guidance had complaints by the users because the users got different levels of guidance for different commands, and this was possibly different enough to change the stimulus-response relationship between the user and the system users didn't have to change the stimulus-response relationship with RUA Properties of RUA Adaptation Why did users so readily accept the adaptation of the RUA system? when people first learn a task, they take instructions about how to perform a task and translate them into actions Anderson's ACT* theory of learning cognitive skills classifies knowledge declarative knowledge (facts, categories, history) procedural knowledge (how to perform tasks, skills) info from external world is always declarative knowledge first when they try the task, they use general translation processes to make this into procedural knowledge then the user can use the procedural knowledge directly to complete the task the next time RUA seems to support this process RUA displays have summary info that would cue the user to do the right thing new users and lower performers also get detailed info at first users watched the videos and instructions very carefully, but then as they got used to what they were doing they used the instructions less thus, when the detailed information was removed they often didn't even notice RUA allows for chunking of tasks novices think of a propcess as a sequence of actions experts think of it as more of a group of actions SO all displays have summaries of the task more detailed displays have information that could prompt actions later on Conclusion: usually preferable to give the users control over adaptation these are guidelines for when user control is not possible ****************************************************************************** Melissa Kay House | This life is a test. Graduate Student, Information Design and Technology|If this had been an actual Georgia Institute of Technology | life, you would have gt0959a@prism.gatech.edu | instructions on where to | go and what to do.