The Open University
Designing a Personal Awareness Agent to Ameliorate Information Overload
Abstract
dc:description.abstractUbiquitous devices provide users with content and notifications that often blur the distinction between work and personal activities and can lead to information overload (IO). This research aimed to support users of multiple collaborative and social systems who may experience this, by designing and evaluating a modular personal software agent - the Awareness Agent - to support the user and ameliorate the effects of overload. Looking at the issue of information overload through a lens of the CSCW concept of Awareness, a survey of users showed that there may be no one size fits all solution, with people perceiving and reacting to IO in different ways - but with common themes emerging showing a need for more control over the algorithms that determine what users see. A set of personas was devised based on this input, which were used to inform and evaluate the design for the agent. With respect for user control and ownership of Artificial Intelligence (AI) interaction being at the heart of the work, a modular framework was designed, able to pull in messages from different sources and make use of varied commodity cognitive computing services to manage that content for the user. A system called User-Directed Machine Learning (UD-ML) was introduced as part of this framework, to enable the user to control the lifecycle of the AI that makes decisions about the content that they see. A novel system for generating and using synthetic messages and conversations was created in order to address practical issues relating to access to confidential and walled-off data in academic studies. A technique for using Generative AI-driven synthetic 'virtual study participants' was also introduced with twin purposes of easing the workload on human participants in high data volume studies, and to more broadly investigate how AI-driven evaluation techniques can form part of the solution to IO. The process of evaluating the Awareness Agent combined human and synthetic study participants to draw conclusions on the efficacy of both UD-ML and the synthetic evaluation techniques themselves. Key findings highlight the importance of balancing AI systems with user control to foster trust and usability. The prototype Awareness Agent showed strong potential as a basis for a practical solution to mitigate information overload, demonstrating effective capabilities in filtering and consolidating information across multiple domains. The techniques for synthetic content and evaluation were found to be an effective part of the study process and are an asset that can be re-used in other research contexts.
Degree
thesis:*- Name dc:type.qualificationname
- phd
- Level dc:type.qualificationlevel
- doctoral
- Grantor dc:publisher.institution
- The Open University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Goddard, David
Rights
- Language dc:language
- en