Neura
Neura – AI-assisted mental health assessment for older adults
A concept project focused on designing clear, low-burden assessment and reporting flows for caregivers and healthcare professionals. Neura explores how conversational signals and passive data can surface meaningful mental health insights without requiring constant user input.
Company
Northeastern University (Acedemic Project)
Timeline
11/2024-12/2024
Team
2 UIUX Designer · 1 Devoloper
Role
UIUX Designer
Focus
Interaction design, assessment flows, web & app UI
Platform
Mobile App + Web Dashboard
Key Users
60 and above older adults
Problem & Constraints
The Problem:
Older adults often experience gradual changes in mental health that are difficult to notice early. Existing tools tend to be either highly clinical or require frequent, active input, which makes long-term engagement challenging. At the same time, caregivers often lack clear, actionable signals to understand when attention or intervention is needed.
Key Constraints:
-
Users have low tolerance for complex interactions
-
High risk of false positives
-
Emotional sensitivity during assessment
Product Strategy
Neura was designed around reducing user burden while still supporting reliable decision-making for caregivers.
Instead of relying on frequent surveys or constant alerts, the system prioritizes conversational and passive assessment.
Insights are derived from patterns over time rather than isolated moments, and user-facing experiences are clearly separated from caregiver-facing reports to match different responsibility levels.
Escalation is intentionally limited and triggered only when sustained risk thresholds are met, helping maintain user trust and emotional safety.
My Role & Contribution
My Role:
As the Product Designer, I focused on defining core product flows, interaction patterns, and decision logic across mobile and web experiences.
Key Contributions:
-
Defined assessment and escalation flows aligned with GDS-style screening logic
-
Designed conversation-based check-ins to reduce assessment fatigue
-
Structured caregiver reports to highlight trends rather than raw data
-
Designed emergency response states with clear triggers and user control
System Context
Supporting the Product Decisions
Neura operates within a multi-device ecosystem where mental health signals are collected passively through daily life. Health data such as heart rate, sleep, and activity is captured via wearables, while emotional cues emerge through natural conversations on voice assistants and mobile devices.
This context reduces the need for frequent, explicit assessments, which are often burdensome for older adults. Instead of relying on active self-reporting, the system synthesizes ambient data and conversational signals to surface mental health risks.
This environment directly shaped key product decisions, including how users enter the experience, how assessments are conducted, and how insights are separated between users and caregivers.
Based on this system context, we made the following key product decisions to balance low user burden with reliable mental health insights.
Key Product Decisions Enabled by AI
Decision 1
Conversation as the Primary Entry Point
Freeform Conversation
Embedded Screening
Behavioral Activation Prompt
Rationale:
Older adults often feel discomfort or resistance when asked to complete explicit mental health questionnaires. Direct screening can feel clinical or emotionally demanding, especially in everyday home settings.
Neura treats natural conversation as the primary interaction surface. Assessment signals are embedded within dialogue rather than presented as separate tasks.
Impact on the product:
-
Assessment feels gradual and non-intrusive.
-
Users can engage without committing to a formal “mental health task.”
Decision 2
Separate User and Caregiver Experiences
Same data, different responsibility levels
User Side
Caregiver Side
Rationale:
Users and caregivers have fundamentally different needs. Older adults require reassurance, autonomy, and minimal cognitive load, while caregivers need clarity, trends, and risk indicators to support decisions.
Combining these needs into a single interface would either overwhelm users or underserve caregivers. Neura separates these experiences while using the same underlying data.
Impact on the product:
-
Users interact with supportive conversations and simple feedback.
-
Caregivers access synthesized reports and alerts only when thresholds are met.
Decision 3
Use threshold-based escalation instead of continuous monitoring
Normal
Mild
Moderate
Severe
Supportive Responses
Self-guided Activities
Caregiver Notified
Constant monitoring or frequent alerts can reduce trust and increase anxiety. Instead, Neura uses a threshold-based model that escalates support only when sustained patterns indicate elevated risk.
Low or moderate signals result in supportive responses and self-guided activities. Higher-risk patterns trigger caregiver visibility and intervention.
This approach balances safety with user autonomy and avoids alert fatigue.
Decision 4
Interfaces That Respond to Emotional Context
Smartwatch
Mobile
Web
Rationale:
During emotional distress, dense information and complex interfaces can become barriers. Neura reveals information progressively based on context, readiness, and device.
Across smartwatch, mobile, and web experiences, feedback focuses on trends rather than raw data, and actions are framed as suggestions rather than requirements.
This decision guided UI hierarchy, layout simplicity, and interaction pacing across platforms.
Smart Watch Interfaces
The smartwatch functions as a passive signal layer, offering low-effort emotional check-ins without interrupting daily routines.
It surfaces simple emotional cues rather than detailed data, gently prompting awareness and reflection.
Mobile Interfaces
The mobile app serves as the primary interaction layer, balancing emotional awareness, reflection, and conversational support.
The home screen prioritizes emotional state over metric exploration.
Mood trends are presented separately from interventions, encouraging reflection without prompting immediate action.
Conversational insights are summarized into lightweight themes to remain human-readable and approachable.
Web Interfaces
The web interface is designed for synthesis and review, supporting caregivers or providers with summarized behavioral patterns and risk indicators rather than raw conversational data.
It emphasizes behavioral patterns and risk indicators rather than raw health metrics, supporting faster interpretation and decision-making.
Reports are shared at the user’s discretion, with automatic escalation only at severe risk levels.
Validation & Feasibility (Conceptual)
Early validation focused on understanding emotional load, trust, and escalation behavior rather than clinical accuracy.
Competitor analysis
Competitor research highlighted gaps between over-clinical tools and overly shallow wellness apps.
Storyboard
Storyboards were used to test perceived autonomy and escalation timing.
Voice-Flow Prototype
Voice flow prototypes explored how assessment logic could be embedded into natural dialogue.
Reflection & Next Steps
1
Maintaining Healthy Boundaries with AI Support
One key challenge moving forward would be ensuring that users do not become overly reliant on automated emotional support.
Future iterations should explore clearer boundaries between emotional check-ins, self-guided reflection, and human intervention.
2
Validating Long-Term Engagement
While the current design focuses on low-burden interactions, long-term engagement patterns would need further validation.
Additional research would help determine how frequently users are comfortable with conversational check-ins over extended periods.
3
Balancing Caregiver Awareness and Signal Fatigue
For caregivers, future improvements could focus on refining escalation thresholds and prioritizing trend-based insights over isolated events, reducing signal fatigue while preserving timely intervention.