Designing a Dementia Caregiver-Centered Recommendation and Context-Aware AI Media Analytical Application: A Comprehensive System Requirements Analysis

Figures & data

Figure 1. Co-design session with caregivers on AI-based recommendation and media analysis application.

Figure 2. Application structure: overall system.

Table 1. Overall system: non-functional requirements.

Requirement	User’s perspective	Technical description
Usability and simplicity	The application should be designed to be user-friendly, intuitive, and easy to understand, enabling users with limited tech literacy to utilize it without requiring extensive instructions or training.	The user interface (UI) design should be kept minimalistic, intuitive, and user-friendly to minimize the likelihood of user errors and ensure ease of use. Only the most important information should be displayed on the UI.
No preparation effort	The application should be seamlessly integrated into a reminiscence session without requiring any additional preparation effort or work.	The application should have the capability to generate all reminiscence session contents in real-time during a session, eliminating the need for time-consuming preparation efforts and the creation of biography forms.
Security	The system should ensure that unauthorized users are prevented from accessing both the application and the stored data.	The use of an encrypted user login should ensure that the application and stored data are protected from unauthorized access.
Adaptability	New requirements that may arise during the course of research should be accommodated in the application’s design.	The code of the application should be designed in a way that allows for easy adaptation to accommodate the evolving needs of people living with dementia.
Privacy	To ensure the highest level of protection for the personal data stored within the application, it is crucial to comply with the data protection regulations that are in effect within the country where it is being used.	The application should comply with the data protection guidelines according to the General Data Protection Regulation (GDPR). (Union & Union, 2016)

Table 2. Profiling: non-functional requirements.

Requirement	User’s perspective	Technical description
Time saving profile creation	The procedure should be designed to be time-saving and capable of being carried out without the need for preparing a biographical sheet.	The profiling process should be kept simple, requiring as little information as possible. However, it should take into account important events in the biography of a person living with dementia to ensure that valuable information is not lost.

Figure 3. Application mockup: profile creation process.

Table 3. Profiling: functional requirements.

Requirement	User’s perspective	Technical description
Profile creation	The only information needed to create a profile is the name of a person living with dementia. There should be an option to select some known biographical information about a resident.	A profile should be capable of being created simply by entering a resident’s name, with the optional ability to select life themes from a list of already known universal life themes. There should also be an optional facility to specify these using free text input.
Profile deletion	When a person living with dementia leaves a facility, their profile should be removable.	When a person living with dementia leaves a facility, the entire profile with all associated data should be removed from the application.
Profile content generation	Profile content and important biographical information should be automatically generated over time.	Following a reminiscence session, the content of a profile should be assigned a personalized Media Rating, which can be changed if necessary. The rated media should then be stored within the profile and categorized accordingly. The categorization of media allows for retrospective adjustments, ensuring flexibility. To keep track of the rating history, a caregiver-specific rating history, that documents which caregiver rated which media item, should be maintained.
Profile layout	For easier profile orientation, the profile content should be divided into the following areas: like, dislike, not sure, and personal media.	The media should be assigned to a profile area (like, dislike, and not sure) based on the media rating conducted.
Personal media	It should be possible to add personal media to a profile.	The media should be added to a profile by a simple file upload. For privacy reasons, personal profile content should only be accessible via a resident’s profile and should not be made available to the general public.

Figure 4. Reminiscence session approach.

Table 4. Media search: functional requirements.

Requirement	User’s perspective	Technical description
Selection of people living with dementia	Each person living with dementia should have their own profile. It should be possible to select people living with dementia for a reminiscence session by clicking on a profile and searching with text. Choosing a single profile should be possible if a one-to-one session is preferred. In the case of group therapy, several profiles should be selectable.	Profiles should be listed alphabetically and should be selectable by clicking on them. Selected profiles should be highlighted. Multiple selections should be possible for both group and individual therapy. A free text search should be available to find people, and search results should be displayed in a list format.
Media search and selection	The media search function should be accessible by entering a free text query or by selecting an example life theme provided by the application. Each search iteration should return four results. The option to search for more content should be available, allowing users to find additional relevant media based on their preferences and interests.	The entered text should be sent as a search request to public media databases. Selected life themes should be treated in the same way as free text. Search results should be analyzed for suitability for people living with dementia. Four media options should be displayed for selection, with the option to load more, enhancing the user’s ability to find appropriate media content for their reminiscence sessions.
Media Suggestions	To facilitate the media search, the application should be capable of automatically generating media suggestions based on the participating people living with dementia. Additionally, a list of sample life themes should be available to simplify the search process.	The media suggestions should be generated by a recommendation system, based on the profile content of the participating people living with dementia.
Media tray	The media tray should be used to conduct and support a reminiscence session with the application. The media resulting from the media search should be displayed here. Clicking on a media will display it in full-screen mode. If it is an audio or video file, a multimedia menu should be provided. Additional media should be added to the media tray during a session.	The media tray should store all identified media in a scrollable list. Clicking on a medium should allow selection and full-screen viewing. Audio or video files should have a multimedia menu. New media should be able to be added during a session while existing media remains in the tray.

Table 5. Media rating: functional requirements.

Requirement	User’s perspective	Technical description
Media rating	Media ratings are important but should be optional. The rating should be provided for each participating person and each medium used in a session.	Media ratings should be used as feedback for the recommendation system and to fill the profiles of the people living with dementia with media content, but they should not be mandatory.
Media rating options	Media rating should be done on a per-resident basis using three different options: “Like,” ”Dislike,” and ”Not Sure.”	To keep the media rating as simple as possible, the rating should be done by clicking on a button (“Like,” “Dislike,” and “Not Sure”).
Media rating dislike [optional]	If a medium is disliked, an additional query should be answered. To keep it as simple as possible, the query should be button-based.	If a medium is rated as disliked, additional information should be provided as to why the medium did not fit, whether it was not dementia-friendly or did not meet the biographical needs or interests of the people living with dementia. This information should be used to identify which component needs improvement (media analysis or recommendation component), allowing for continuous enhancement of the application’s performance and content selection.

Figure 5. Application mockup: media rating.

Table 6. Media storage: functional requirements.

Requirement	User’s perspective	Technical description
Participating people living with dementia selection	To conduct a reminiscence session with previously stored media sessions, it should be possible to select the participating people living with dementia.	Selecting the participating people living with dementia is a prerequisite for the recommendation and media rating components.
Meta information	Each media item stored in the application should have associated meta information that should be used internally by the system and should not directly affect the user.	The system should have meta-information about stored media to meet the needs of the recommendation system and media analysis. storing this information prevents duplicate analysis and helps to provide accurate media suggestions. Meta information can include any details about a media item.
Templates	The application should provide pre-composed media templates based on life themes.	Pre-defined media templates should be stored in the application together with their associated meta-information.

Table 7. Media analysis framework: functional requirements.

Requirement	User’s perspective	Technical description
AI	The media screening process to determine the suitability of media for people living with dementia should be implemented using AI techniques.	AI, machine learning (ML), and deep learning (DL) techniques should be employed for media analysis.
Adaptable	As additional relevant media features could be identified in research over time, it should be ensured that the AI system can adapt to changing circumstances.	The system should be adaptable to integrate new contextual information without requiring a time-consuming re-training of the entire network.
Explainable	The decisions about whether or not media is dementia friendly should be understandable to the user.	The system should be able to explain what it does and why, as it will be used directly in the context of people in need of care, ensuring transparency and understanding of the criteria used by the system.
Customizable	Users should be able to customize the system according to the specific needs and preferences of each person living with dementia, ensuring that the system addresses the unique needs and sensitivities of each person living with dementia.	A user should be able to define rules to change the media results, to meet the specific needs of each person living with dementia.
Context awareness	The Media Analysis process should utilize a context-aware AI approach to meet the other media analysis requirements: adaptable, explainable, and customizable. All information related to a person living with dementia should be considered as context information during the analysis.	Incorporating contextual information into a learning system significantly enhances the value of its results, seamlessly imbuing it with explainability, adaptability, and customizability. To fully harness the benefits of all its features, integrating context into the system necessitates the creation of a modular ensemble learning system.
Image filtering criteria	The AI system should be able to process all types of media: images, audio and videos.	The AI system should be able to analyze both syntactic and semantic media features, especially in the area of images, focusing on parameters such as contrast, colors, brightness, object identification, number of objects, object size, and semantic image content.

Table 8. Recommendation system: functional requirements.

Requirement	User’s perspective	Technical description
Personalized	The system should generate media suggestions tailored to the entire session group, ensuring that everyone’s biographical needs and preferences are taken into account.	The recommendation system should be highly flexible, allowing for tailored suggestions for both individuals and groups of people. It should also provide the option to adjust the number of remaining participants to account for varying group sizes.
Media category	The media suggestions made by the system should be related to images, audio, and videos.	The system’s ability to make recommendations should not be limited to a specific type of data, as it should be capable of making recommendations for a wide range of media, including images, audio, and video.
Explainable	A user should be able to understand why a media suggestion has been made.	A user should be able to understand the reasoning behind a media suggestion. To maintain transparency in the decision-making process, especially when working directly with individuals, the recommendation system should be explainable.

Union, E., & Union. (2016). Regulation (EU) 2016/679 of the European parliament and of the council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing directive 95/46/EC (general data protection regulation). https://eur-lex.europa.eu/eli/reg/2016/679/oj (Official Journal of the EuropeanL 119/1).

 Google Scholar