The Study Of Intrusion-Based Robust User Experience Design On Predictive Interface

With the rapid development of artificial intelligence technology,computer systems have gradually become more aware of and able to learn about users and environments.Driven by AI technology,more and more systems could facilitate users by predicting their intentions and behaviors.Follow this trend,predictive interface user experience design should become a hot research topic.Due to the uncertainty of the predicted results,the traditional user experience design methods for deterministic systems will also face a lot of challenges.First,conventional design methods do not think of whether the case is right or wrong.While a wrong prediction without a robust experience design can easily cause negative interference and even harm the users;Secondly,the design targets of conventional design methods are usually passive interactive systems.The effects of the system's proactive behaviors on the users is not considered;Third,conventional design methods are used to subdivide users and scenarios,and generate personalize design for these typical users and scenarios.While the intelligent system's understanding of the users and scenarios is much more abstract and detailed,which is hard to match with the understanding of the experience designer;Fourth,conventional design methods generally discuss how to understand users from the perspective of designers,without considering how to make the system better understand and learn the users'behaviors and habits.With the increasingly mature artificial intelligence technology,more proactive and predictive interactive systems are becoming technically possible.However,from the perspective of user experience design,the academia has not yet provided targeted design guidance for the above issues.To solve the above problems,a robust design method of predictive user experience based on intrusion degree is proposed in this paper.With the study of single low-intrusion design model,interaction behavior modeling and training,design element extraction,and dynamic multi-intrusion interface design model,the feasibility and effectiveness of this method is verified.This research is mainly divided into four parts as follows:The first part proposed the single low-intrusion interaction model,followed by an empirical design and evaluation study(1)The single low-intrusion interaction model is proposed.The main features of the strategy and its role in improving the robustness of user experience are discussed;(2)The single low-intrusion interaction model is applied on smart TV input method design;(3)The feasibility and effectiveness of the single low-intrusion interaction model are verified by computer simulation and user usability testThe second part constructed the interaction rule training and prediction model for rapid prototyping.(1)Taking contextual intelligence recommendation products as an example,the contextual recommendation process is modeled based on the Markov decision process;(2)Through questionnaire survey and user interview,the typical contextual status features and recommendable actions were identified,the acquisition and divergent design method of behavior experience is proposed,and the experience replay pool is constructed;(3)The reinforcement learning model for rapid prototyping was constructed and trained based on experience replayThe third part extracted the design elements,and identified the corresponding priority.(1)Taking mobile app recommendation as an example,the preliminary design elements and the corresponding effects of on users were identified through questionnaire survey and user interview(2)The design elements were further extracted through eye-tracking experiment.The final design element set and priority ranking were generated(3)The extraction method is extended to all recommendable actions,and finally the extraction of design elements of all recommended behaviors is realized,which can be used as the design material for the subsequent dynamic multi-intrusion designIn the fourth part,based on the results of the second and third part,the dynamic multi-intrusion interface design model was carried out(1)The dynamic multi-intrusion interaction design model is proposed and its effect on the robustness of user experience in uncertain systems is discussed and analyzed;(2)The intrusion degree of current mobile interaction components was evaluated According to the evaluation results,the components with appropriate intrusion level were selected for the next design progress;(3)Based on the design elements with priority of part 3,the multi-intrusion interfaces for all actions(of part 2)were carried out.The designs were validated by the predictive model(of part 2),and were optimizedBased on the above studies,a model training and design evaluation system for predictive interface is constructed.It supports the training and predictive evaluation of dynamic multi-intrusion designs to facilitate the optimization iteration processes.