A new pilot project in Newton, MA, is testing an AI-driven cognitive screening platform that leverages real-time voice transcription, automated time-marker extraction, and adaptive task design to detect early indicators of memory change.
The system captures fine-grained speech and timing signals across multiple languages, enabling high-resolution cognitive analysis while prioritizing participant comfort, privacy, and ethical data use.
NEWTON, MA – The Cognitive Health Tests pilot has officially launched in Newton, Massachusetts, introducing a short, noninvasive, technology-powered cognitive screening experience designed for research and exploratory purposes. The platform evaluates whether artificial intelligence, multilingual voice transcription, speech signal processing, and automated time-based analytics can generate meaningful insights into early cognitive change, while maintaining strict standards for confidentiality, safety, and voluntary participation.
Unlike traditional paper-based or clinician-administered cognitive tests such as the NMSE or MoCA, the platform captures continuous voice input during task performance and applies real-time transcription and timestamping to analyze not only task accuracy, but also response latency, speech rate, hesitation patterns, pausing behavior, and verbal fluency. These high-resolution time markers allow the system to detect subtle changes in cognitive processing speed, memory retrieval, and language formulation that conventional scoring methods may overlook.
The platform delivers a series of brief, digitally administered tasks that assess short-term memory, attention, processing speed, language, reasoning, and orientation. By combining adaptive task design with automated speech analysis, the system enables fine-grained behavioral measurement without increasing participant burden. Support for multilingual voice input expands accessibility and enables cross-linguistic cognitive assessment, creating opportunities for broader deployment and population-scale research.
Using machine learning models, the platform analyzes task accuracy alongside speech dynamics, response timing, and micro-pauses to generate layered insights into cognitive performance. This multidimensional approach transforms traditional screening into a data-rich, software-driven process, offering deeper interpretability while preserving a low-stress, short-duration experience. Tasks remain intentionally brief and adaptive, enabling participants to engage comfortably and withdraw at any time without consequence.
“This pilot explores how modern AI, speech recognition, and time-series analytics can redefine early cognitive screening,” said project lead and founder Jonathan Xue. “By transcribing voice input in real time and extracting detailed timing signals, we gain visibility into how cognition unfolds moment by moment, not just whether answers are correct. This opens the door to more sensitive, scalable, and accessible screening across languages and environments.” Xue added, “Equally important is our emphasis on privacy-first engineering and ethical system design, which guide every stage of development and deployment.”
The system is engineered with privacy-first architecture. No personal identifiers, demographic data, medical records, or raw audio recordings are retained. Instead, speech is processed in real time, converted into anonymous numerical indicators such as latency metrics, fluency scores, and timing distributions, and immediately discarded. Only aggregate, de-identified outcomes are stored, and all data is used solely to evaluate platform performance, reliability, and usability.
The pilot builds upon earlier technical trials conducted at Sunrise of Newton in Massachusetts and Sunny View in Cupertino, California, with additional deployments planned in Dublin and San Francisco, California. These early implementations validated system stability, multilingual speech processing, and participant usability, supporting continued refinement of the platform’s AI models, transcription accuracy, and time-series analytics.
The Cognitive Health Tests pilot reflects a broader effort to advance early memory detection through responsible technology innovation. By integrating cognitive science, machine learning, multilingual speech recognition, and automated timing analysis within a privacy-centric architecture, the project aims to shape next-generation approaches to scalable, software-driven cognitive screening.
The Cognitive Health Tests pilot is an exploratory research initiative focused on evaluating an AI-enabled, low-friction cognitive screening platform. The project emphasizes ethical data handling, participant comfort, and transparent system design to support responsible progress in cognitive health technology
For more information about the Cognitive Health Tests pilot, visit: https://cognitive-test-app.vercel.app/
