Abstract

This TREO talk introduces Voxvoro (voxvoro.com), a mobile-optimized Progressive Web App (PWA) designed to mitigate scholarly information overload by repurposing high-engagement social media interface patterns for research discovery. As academic publication volume increases, scholars often default to non-productive digital "doomscrolling" habits due to the low friction and variable reward schedules of social media platforms. Voxvoro addresses this by utilizing a local Large Language Model (Gemma3:27b via Ollama) to transform dense academic metadata into easy to digest concept cards. The Voxvoro architecture integrates a multi-stage data pipeline designed for high-signal scholarly synthesis, bridging a Python-based ingestion engine with a Next.js frontend to enable a "native-feel" mobile experience. The backend manages a hardcoded registry of 117+ high-impact journals and utilizes the OpenAlex API to fetch and reconstruct recent works from each source. AI orchestration is handled via local LLM inference, where a specialized system prompt directs the model to generate summaries constrained to three standardized areas: Core Finding, Methodology, and Implications. In addition to text, the pipeline generates conversational deep-dive scripts synthesized into high-fidelity audio using the Kokoro ONNX TTS model, stored via Cloudflare R2. Grounded in Cognitive Load Theory, the artifact uses this structured information chunking and bolded key-phrase signaling to reduce extraneous cognitive load, facilitating rapid scanning in a few seconds. The frontend leverages server-side fetching via Supabase for low-latency delivery, while the Progressive Web App (PWA) framework supports mobile-first features and direct DOI linking for immediate access to full-text manuscripts. This presentation discusses the design science process of building the PWA, the technical challenges of local multimodal orchestration, and preliminary faculty feedback regarding the replacement of non-productive digital habits with efficient scholarly engagement. The ongoing evaluation of this artifact focuses on three primary behavioral and cognitive dimensions. First, the project investigates whether the transition from unformatted abstracts to structured, bolded concept cards and audio summaries significantly reduce the extraneous mental effort required for literature screening. Second, the study explores the potential for "habit displacement," questioning if a high-signal, low-friction scholarly PWA can successfully compete for the digital "micro-moments" currently dominated by non-productive social media platforms. Finally, the research assesses the affective response of the user, specifically examining how "scrolling guilt" or digital fatigue may be mitigated when mobile scrolling, or passive listening, is shifted from entertainment-based content to professional research discovery. These questions will be addressed through a formative evaluation involving a cohort of university faculty, using self-reported engagement data and qualitative feedback to refine the artifact's design science cycle.

Download

Share

COinS