Mashrur Rashik

If you're attending CHI 2025, please stop by our presentation on April 28th from 5:32 PM to 5:44 PM.

This work demonstrates the potential of combining AI and human-centered design to transform chronic disease management by enabling supportive, personalized tools that empower patients and enhance clinical decision-making.

Furthermore, two clinicians and a movement disorder specialist confirmed the usefulness of the collected data in supporting care for Parkinson's patients.

Our findings showed that Patrika effectively interpreted user input and maintained coherent, health-related personalized conversations with participants. Participants responded meaningfully, and Patrika successfully captured both direct symptom references and indirect personal anecdotes.

To evaluate Patrika, we conducted a two-week deployment with eight Parkinson's patients. Insights from this first study revealed several usability and performance issues, which informed improvements to the system. Following these improvements, we ran a second user study involving nine patients.

The system overview of Patrika. First, the NLU module parses the user’s input and forwards the information to the LLM intent interpreter as a prompt. The LLM intent interpreter predicts the user’s intent from the prompt and sends it to the response generation module. The response generation module uses the predicted intent to trigger an appropriate journaling rule and selects the next relevant journal probe. To personalize this probe, the response generation module queries the journal module using the journal probe and conversation context. The journal module uses a BM25 retriever to retrieve relevant history from the user’s conversation history that correspond to the journal probe and the ongoing conversation. This conversation history is then used to personalize the journal probe, which is sent to the user through Alexa.

In this work, we present Patrika, an AI-enabled journaling system for Parkinson's patients. The system employs cooperative conversation principles, clinical interview simulations, and personalization to provide a more effective and user-friendly journaling experience.

The work was done by @mashrurrashik.bsky.social Shilpa Sweth, Nishtha Agrawal, Saiyyam Kochar, @narges-mahyar.bsky.social, Ali Sarvghad from UMass Amherst, Kara M Smith from UMass Chan Medical School, Fateme Rajabiyazdi from Carleton University, & @vsetlur.bsky.social from Tableau Research.

This figure shows an anonymized snippet of conversation between a patient with Parkinson’s disease and Patrika. Here: (A) The patient invokes their journal using a voice command to initiate the conversation. (B) Patrika responds with a personalized greeting [name] and prompts the patient to start recording. (C) Patrika detects “tremor” in the patient’s previous entry and initiates collecting additional data via asking relevant follow-up questions. (D) Patrika utilizes the conversation history with the patient to personalize the follow-up question, asking about symptoms previously co-occurring with tremors and their impact on their routine activities. (E) Patrika utilizes conversation history to identify and compare the severity of recurring symptoms. For instance, in a previous entry, the patient reported experiencing tremors with no change in severity or overall experience up to that point. Patrika compares this information to the most recent tremor episode, which occurred in the middle of the night. (F) Patrika confirms that the data has been recorded and then prompts the user to either end the conversation or initiate a new one.

I am excited to share our @chi.acm.org 2025 paper, “AI-Enabled Conversational Journaling for Advancing Parkinson’s Disease Symptom Tracking.”

Link to preprint: arxiv.org/pdf/2503.03532

CHI program schedule: programs.sigchi.org/chi/2025/pro...

This work was awarded the Honorable Mention award!