DEVELOPMENT OF AN EASY-TO-USE APPLICATION WITH ARTIFICIAL INTELLIGENCE TO SUPPORT DIGITAL INCLUSION FOR THE ELDERLY

Isabel Rio-Torto, Vinícius Marques da Silva Ferreira, Carlos Alberto Nunes Cosenza, Kilmer Pereira Boente, Hamilton Burratto Júnior, Paulo Sérgio da Silva Santos, Margarida Ferreira, Gabriel dos Santos RaimundoAracê 2025 · DOI

TL;DR

FacilitaApp is an Android-based solution designed to bridge the digital divide for the elderly by integrating Generative AI (Google Gemini) with accessibility-first design. By simplifying complex interfaces and providing real-time security verification, the project aims to foster autonomy and digital literacy for a rapidly aging global population.

What Problem Does This Paper Address?

As the global population ages—with Brazil’s elderly population expected to triple by 2100—digital exclusion has become a significant barrier to human dignity and citizenship. The authors identify a "perfect storm" of challenges facing seniors:

  • Physical & Cognitive Limitations: Vision impairment (presbyopia), motor tremors, and hearing loss make standard mobile interfaces (small buttons, low contrast) nearly unusable.
  • Psychological Barriers: A pervasive "fear of breaking" devices and "digital shame" leads to withdrawal from technology.
  • Security Risks: High vulnerability to online fraud and scams creates a perception of the internet as a dangerous environment.
  • Design Gap: Most modern applications are built with an "imperative" mindset that prioritizes feature density over intuitive, accessible navigation.

Methods

The researchers developed FacilitaApp using a modern, robust tech stack tailored for maintainability and accessibility:

  • Declarative UI: The app was built using Kotlin and Jetpack Compose. Unlike traditional XML-based imperative paradigms, this declarative approach allows for better state management and more cohesive UI components that react automatically to data changes.
  • Architecture: The project follows the MVVM (Model-View-ViewModel) pattern, ensuring a clean separation between the user interface and the underlying logic.
  • AI Integration: The core "intelligence" is powered by the Google Gemini 1.5 Flash model, utilized for contextual chat support and automated news summarization.
  • API Ecosystem:
    • Google SafetyNet API: For real-time website reliability verification.
  • SerpAPI: To curate and recommend popular, reliable apps from the Play Store.
  • GNews API: For news integration.
  • TTS/STT: Native Android Text-to-Speech and Speech-to-Text for eyes-free/hands-free interaction.
  • Design Philosophy: The development was grounded in Universal Design (creating products usable by all) and Participatory Design (incorporating feedback from the target demographic).

Datasets

The paper does not involve training a novel machine learning model from scratch; instead, it leverages state-of-the-art pre-trained models via API. However, it utilizes several critical data sources for its foundation and functionality:

  • Demographic Data: Extensive use of data from the IBGE (Brazilian Institute of Geography and Statistics) and the United Nations to map aging trends and internet usage gaps.
  • Live Data Streams: Real-time data from the GNews API and Google Play Store (via SerpAPI) to populate the app’s features.
  • Safety Benchmarks: The Google Safe Browsing database (accessed via SafetyNet) serves as the benchmark for identifying malicious URLs.

Key Insights and Ideas

  • Security as Accessibility: The authors argue that security features (like URL checking) are actually accessibility features for the elderly. Reducing the "fear of fraud" is just as important as increasing font size for digital inclusion.
  • The Power of Summarization: For users with slower information processing or cognitive fatigue, the ability of an LLM to condense long news articles into "bullet points" significantly lowers the barrier to staying informed.
  • State Management is Key: By using StateFlow and Hilt (for dependency injection), the developers ensured the app remains performant and "bug-free" even when handling complex asynchronous calls like AI chat or security scans.
  • Multimodality is Non-Negotiable: The transition from text-only to voice-and-text (TTS/STT) is what transforms a standard app into a "virtual assistant" that feels like a companion rather than a tool.

Key Results

The paper details the successful implementation of a functional prototype with the following results:

  • Integrated AI Chat: Successful deployment of a Gemini-powered assistant capable of handling natural language queries to help users navigate the digital world.
  • Automated Security Shield: A functional verification system that warns users before they visit potentially malicious sites, directly addressing the "fear of fraud."
  • Simplified Navigation: A UI featuring high-contrast elements, large fonts, and a "Useful Apps" section that bypasses the cluttered standard Android app drawer.
  • Optimized News Consumption: A dedicated tab where Gemini-summarized articles reduce reading time and cognitive load for the user.

Limitations and Future Work

The authors candidly acknowledge that while the technical foundation is solid, there is room for growth:

  • User Testing: The current phase focused on development; the next critical step is extensive field testing with elderly participants to measure real-world UX impact.
  • Voice Recognition Nuance: The STT system currently struggles with non-standard pronunciations or regional expressions common among older generations.
  • Offline Functionality: Future iterations aim to include caching and offline support to ensure the app remains useful in areas with poor connectivity.
  • Onboarding: The team plans to develop a specialized "onboarding" process to guide first-time users through the app’s AI features without overwhelming them.

Summary updated

Abstract

The research included the elderly population through the development of the FacilitaApp mobile application for Android, using features such as a virtual assistant, automatic content reading, website reliability verification, news summaries, and accessibility features, aligned with the principles of Universal Design and Participatory Design, based on the lessons learned throughout the research. The proposal arose from observing the difficulties faced by this audience in using digital technologies, such as vision problems, motor limitations, fear of virtual fraud, and low familiarity with complex interfaces. Thus, the project seeks to promote citizenship, autonomy, and human dignity for older adults in the digital environment.
BibTeX
@inproceedings{riotorto2025development,
  author = {Isabel Rio-Torto and Vinícius Marques da Silva Ferreira and Carlos Alberto Nunes Cosenza and Kilmer Pereira Boente and Hamilton Burratto Júnior and Paulo Sérgio da Silva Santos and Margarida Ferreira and Gabriel dos Santos Raimundo},
  title = {{DEVELOPMENT OF AN EASY-TO-USE APPLICATION WITH ARTIFICIAL INTELLIGENCE TO SUPPORT DIGITAL INCLUSION FOR THE ELDERLY}},
  booktitle = {Aracê},
  year = {2025},
  doi = {10.56238/arev7n11-202},
}
Figure