The Attentive Room

The Attentive Room is a post-smartphone productivity system that transforms your environment into a cognitive ally. It uses a network of sensors—microphones, thermal cameras, air quality monitors, and wearables—to infer your focus state in real time. Based on this, it subtly adjusts lighting (color temperature, brightness), sound (nature soundscapes, white noise), and even scent diffusers to match your activity: cool blue light for deep work, warm amber for reading, lavender for wind-down. You set intentions through voice commands or by placing a physical token (e.g., a wooden block labeled 'Focus') on a shelf. The system integrates with your calendar and task manager, pre-emptively preparing the room for meetings or creative blocks. Feedback is entirely ambient: a gentle hum when you're off-task, a vibrating floor tile when it's time for a break. There are no screens, no notifications—just a room that breathes with your productivity.

The system learns your personal rhythms over time, using end-of-day self-assessments to refine its models. It can even suggest schedule adjustments: "Your peak focus window is 9-11 AM—maybe reschedule your stand-up?" Multi-user support allows a shared home office to adapt to each person's presence via UWB tags. The Attentive Room is not about control; it's about creating an environment where focus happens naturally.

Modern productivity tools are ironically distracting. Smartphones, smartwatches, and desktop apps bombard us with notifications, pulling us out of flow. We spend more time managing our attention than actually working. Meanwhile, our physical environment—lights, noise, temperature—is often static or manually adjusted. There is no system that proactively shapes the environment to support cognitive needs without requiring screen interaction. The Attentive Room fills this gap by replacing digital nudges with ambient cues, making productivity feel effortless.

• Create a room-scale system that adapts to user's focus state without explicit input.
• Eliminate screen-based notifications; replace with environmental cues like light and sound.
• Integrate with calendar and task lists to adjust environment for scheduled activities.
• Learn user's peak focus times and suggest schedule adjustments.
• Support multiple users in the same room (e.g., home office for couple).

• Not a wearable device; it's about the room.
• Not a brain-computer interface; no invasive sensing.
• Won't replace planning tools; it integrates with them.
• Not for public spaces; designed for personal rooms.
• Does not enforce tasks; it's an enabler, not a boss.

Sensors: Ubisense UWB tags for location, BOSCH BME680 for air quality, MEMS microphones for voice/activity, thermal cameras (FLIR Lepton) for facial temp (cognitive load). Actuators: Philips Hue lights (color temp & brightness), Sonos speakers (soundscapes), Aroma360 scent diffuser, haptic floor tiles (subwoofers with vibratory feedback). Edge processing: NVIDIA Jetson Orin for local ML inference, privacy-first audio processing (on-device wake word, no cloud). Integration: Calendar APIs (Google, Outlook), Todoist/Notion APIs. Physical token: Wooden blocks with NFC tags for state intention (Focus, Relax, Meeting).

The system is an edge AI loop running on a Jetson. Sensor data (audio envelope, thermal facial temperature, air quality, location) is fed into a multi-modal neural network that outputs a focus state: 'flow', 'distracted', 'tired', 'overstimulated', 'neutral'. This state, combined with calendar events and user-set intentions (via token or voice), determines actuator settings using a rule-based engine with learned biases. The user can override any actuator via voice (e.g., "lights brighter") or by adjusting the token. Each evening, the user rates their day's focus (1-5), which becomes a training signal for the model. The system also correlates state changes with task completion rates from integrated task managers to refine predictions.

• Privacy concerns with always-on audio/visual sensing; need robust local processing and clear consent.
• User resistance to environmental changes without explanation; must provide subtle feedback for actions.
• Over-personalization causing dependency or discomfort when the system misreads state.
• Technology not mature enough for reliable cognitive state detection from non-invasive sensors; false positives could annoy users.
• Cost of hardware (Jetson, sensors) may be prohibitive for mainstream adoption.

• How to ensure the room doesn't become creepy? Need explicit opt-in and clear feedback on what is sensed.
• What is the minimal set of sensors to infer focus accurately without being intrusive?
• Should the room have a 'manual mode' if AI is consistently wrong?
• How to handle edge cases like visitors or pets?
• What is the best way to measure 'focus' without subjective self-report?

The state of ambient computing hardware is still too expensive and fragmented for a polished consumer system; reliable cognitive state inference from non-invasive sensors is not yet production-ready. The team shifted focus to a related startup that requires less hardware R&D, leaving this plan on the shelf.

Based on research from Microsoft's Ambient Intelligence lab and the MIT Media Lab's 'Doppel' project. Key insight: ambient feedback should be below conscious awareness to be effective.