Connor Dickie

This prototype explores how attention sensing can make large language models behave more like thoughtful conversational partners. In the first demo, two Groks run in voice mode, each on its own iPad and augmented with a Raspberry Pi camera performing real-time head-pose estimation. A custom interface bridges each Pi to its paired iPad via a hacked USB mouse, dynamically muting the non-addressed AI. The result is a natural, gaze-driven turn-taking system where only the agent being looked at can hear and respond — dramatically reducing interruptions and overlap.

In a follow-up experiment, the same approach was applied to ChatGPT while watching television. The system ignored speech coming from the TV and responded only when I made direct eye contact, demonstrating how simple attention cues can solve one of the biggest usability issues in multimodal, voice driven AI.

These demos illustrate a path toward socially intelligent, gaze-aware interfaces that will be foundational for multi-agent systems and humanoid robots. The project blends computer vision, hardware hacking, and human-computer interaction research — showing how embodiment and context awareness can transform the quality of human-AI conversation.

As lead inventor of eyeLook, I developed an attention-sensing platform that uses an embedded EyeContact Sensor (ECS) to detect when a user is looking at their device and dynamically adapt media playback. The system could pause a video when the viewer looked away (seeTV) or advance text only when the reader was engaged (seeTXT).

The concepts were commercialized by Samsung as Smart Pause and Smart Scroll, premiere features on the Galaxy S4 and S5, promoted in global TV campaigns and shipped on over 90 million devices. My patented technology (US Patent 8,672,482) demonstrated how human-aware interfaces could make everyday devices more intuitive—a concept that has since become standard practice.
Today, similar attention-sensing features appear in leading products such as the Apple iPhone, which uses gaze awareness to unlock devices and keep screens awake—validating the long-term influence of this work on mainstream interaction design.

In 2016, our team pursued a $50 million patent infringement case against Samsung for unlicensed use of the technology. Although the patent was ultimately invalidated during litigation, the case highlighted the commercial relevance and foresight of the invention, as well as the challenges innovators face when bringing new interface paradigms to market.

Highlights

Inventor of the eyeLook platform — integrating computer vision, UX design, and embedded sensing for attention-aware media.

Commercialized globally as Samsung Smart Pause and Smart Scroll — featured in TV commercials and flagship product launches.

Over 90 million devices shipped with this technology.

US Patent 8,672,482 — foundational intellectual property in gaze-responsive mobile interaction.

IP strategy & enforcement leadership — led efforts in multi-million-dollar litigation for unlicensed commercialization.

Early demonstration of concepts now seen in modern smartphones, such as attention-aware unlocking and display wake control.

This project exemplifies my ability to invent, productize, and defend pioneering technologies that reshape how humans interact with machines.

ACM UIST ‘05 Paper - eyeLook: Using Attention to Facilitate Mobile Media Consumption

Patent - Method and apparatus for communication between humans and devices

Patent Infringement Suit - Queen’s University sues Samsung over alleged patent theft

Developed at the Human Media Lab (Queen’s University), the Eye Contact Sensing Phone was an early exploration of attention-aware mobile devices. The prototype integrated a low-cost EyeContact Sensor (ECS) and speech activity analysis to detect when the user was engaged in a face-to-face conversation.

By interpreting eye contact and conversational cues, the phone could determine when it was socially appropriate to interrupt—reducing disruptive notifications and adapting its behavior in real time. For example, the system could automatically silence alerts or signal to callers that the user was currently in conversation.

This work, published in CHI 2002 as Designing Attentive Cell Phones Using Wearable EyeContact Sensors, was among the first demonstrations of context-aware, socially intelligent mobile devices—anticipating today’s smartphones that use attention sensing to manage screen behavior and notification delivery.

Highlights

Published research: Designing Attentive Cell Phones Using Wearable EyeContact Sensors (CHI 2002).

Combined eye contact detection and audio analysis for context-sensitive interaction.

Explored socially intelligent mobile design—devices that understand when and how to demand attention.

Early foundation for modern attention-aware features in smartphones, such as adaptive notification modes and gaze-based display control.

References

ACM CHI. Minneapolis, USA. 2002.

Roel Vertegaal, Connor Dickie, Changuk Sohn and Myron Flickner.

Available: ACM - Attentive Cellphone

The Attentive Television was an early exploration of the Attentive User Interface (AUI) paradigm; a system that uses eye contact to mediate attention and control. Developed while I was a film student exploring the relationship between audience and screen, the prototype used an eye-contact sensor to detect when someone was watching; when attention drifted away, playback paused automatically, and when the viewer returned, it resumed.

This work demonstrated how devices could infer user intent and activity purely from visual cues, reducing friction and energy use while making interactions feel more human.

The Attentive Television became the foundation for my later research on mobile media consumption, which ultimately evolved into Samsung SmartPause, a flagship feature on the Galaxy S4 and S5 that shipped to over 90 million users worldwide.

At the intersection of social interaction and wearable computing, ECS Glasses explored how devices can detect when someone is looking at you, rather than just when you are looking at a device. The prototype—described in “Eye Contact Sensing Glasses for Attention‐Sensitive Wearable Video Blogging”—used a pair of eyeglasses augmented with a custom eye-contact sensor and wireless streaming functionality.

When the system detected another person’s gaze directed at the wearer, it could trigger contextual behaviours: for example, switching camera modes, flagging attention events, or logging social-attention metrics for use in wearable video and lifelogging applications.

Highlights

Demonstrated a novel wearable system that senses when others are looking at you—a reversal of the typical “you are looking at the screen” model.

Integrates gaze-detection and wearable electronics into eyewear form-factor, anticipating social and device-aware interactions in future wearables.

Published work provides the technical basis and design rationale for attention-sensitive wearables.

This work underscores my ability to design and build systems that sense social context, bridge hardware and interaction design, and anticipate emerging attention-aware paradigms in wearable technology.

References

ACM CHI. Vienna, Austria. 2004.

Connor Dickie, Roel Vertegaal, Jeffrey S. Shell, Changuk Sohn, Daniel Cheng and Omar Aoudeh.

Available: ACM - eyeBlog
Media: link (Boingboing.net), link (Slashdot.org)

Augmenting and Sharing Memory with EyeBlog

1st ACM CARPE, ACM Multimedia. New York, New York. 2004.

Connor Dickie, Roel Vertegaal, David Fono, Changuk Sohn, Daniel Chen, Daniel Cheng, Jeffrey S. Shell and Omar Aoudeh.

Available: ACM - eyeBlog 2.0
Media: .pdf (Globe & Mail)

This project applied my attentive computing research to a retail environment, creating a context-aware vending machine that senses and responds to human presence, engagement, and emotion using computer vision. The system detects when a user approaches, looks at, or interacts with it, and dynamically adjusts its interface and lighting to match the level of attention it receives.

Built as an independent prototype, the vending system combines proximity and motion sensing, embedded microcontrollers, and an adaptive display interface. It demonstrates how principles of attentive systems—traditionally used in HCI and personal devices—can be extended to public and commercial spaces.

Highlights

Explored how attentive computing can make public interfaces more human-responsive.

Sensors detect approach, gaze, and interaction, modulating system behaviour in real time.

Bridged embedded hardware, perception sensing, and interaction design into a single adaptive system.

Illustrates the broader paradigm of context-aware interfaces that respond intelligently to human attention and proximity.

This project represents an early exploration of how physical systems can perceive social cues and modulate their responses—a recurring thread throughout my work on attentive devices and human-aware interfaces.

References

Patent #8,594,838

Ted Selker, Connor Dickie, Matthew Hockenberry, John Wetzel and Julius Akinyemi. 2006

Exhibited at Wired NextFest NYC. 2006, PepsiCo. HQ., Purchase, NY. 2006, MIT Media Lab 2006

During my early research era, I co-designed a prototype system that redefined the office workspace as an attentive environment—one that senses where people are looking, how they’re oriented, and who they’re socially grouped with. The result: the “Attentive Office Cubicle,” a workspace that automatically mediates visual and auditory interactions between co-workers.

The prototype integrated:

Privacy-glass walls that switch between opaque and transparent based on real-time detection of joint head orientation and proximity (i.e., when two people turn toward each other, the wall becomes transparent so they can interact).

Noise-cancelling headphones tied into the system that mute background audio when the user is alone/focused, but become “transparent” to ambient sound when the system infers that a social interaction is about to occur.

We coined and used the term Social Geometry to describe how simple measures of head-orientation and body clustering can infer which people are likely interacting together—as opposed to relying on expensive eye-tracking or full gaze‐analysis systems. 

By sensing the state of attention in the space, the cubicle could fluidly shift between “focus” mode and “social” mode, supporting deep work and collaboration without explicit user commands.

This work laid a key foundation in my broader trajectory of explicitly embedding visual‐attention and social‐attention sensing into media devices and environments. It anticipates how we might design media systems that are aware of social context rather than simply user context.

Paper: UBICOMP 2004 - Attentive Office Cubicles: Mediating Visual and Auditory Interactions Between Office Co-Workers

At the Human Media Lab (Queen’s University) I was part of the team behind LookPoint: a system that used eye tracking to detect which screen or device a user was looking at, then automatically rerouted mouse and keyboard input to that device.

In the prototype, users switched input between three computers simply by looking at different monitors — removing physical device switching and enabling a more seamless, “hands-free” interaction experience. The evaluation found that eye-input for switching was 111% faster than a mouse, 75% faster than function keys, and 37% faster than using multiple keyboards.

Highlights

Demonstrated a novel interaction model: look to switch, instead of click or type.

Empirical user study showing significant speed advantages for eye-based switching between devices.

Explored seamless integration of input across multiple computing devices, anticipating environments with dense screen-ecosystems and mixed device workflows.

Exemplifies my work at the intersection of input devices, human-computer interaction, and applied gaze technology.

References

ACM OZCHI. Sydney, Australia. 2006.

Connor Dickie, Jamie Hart, Roel Vertegaal and Alex Eiser.

Available: ACM - LookPoint

Developed at the Human Media Lab, FlexCam explored how flexible OLED displays and camera arrays could merge into a single interactive imaging surface. The prototype used a thin-film FOLED display on a flexible chassis with embedded flex sensors and three FireWire cameras.

When the display was bent, software re-stitched the video feeds in real time based on sensor input — allowing users to physically control the camera’s field of view by flexing the screen.

Highlights

Published as FlexCam — Using Thin-Film Flexible OLED Color Prints as a Camera Array.

Demonstrated early “bend-to-zoom” interaction using flexible electronics.

Combined display, sensing, and computer vision into a single prototype.

Anticipated interaction models later seen in foldable and flexible smartphones, where device shape directly influences camera behavior and user experience.

References

ACM CHI. Austin, Texas. 2012.

Connor Dickie, Nicholas Fellion, Roel Vertegaal.

Available: ACM - Flexcam [PDF]
Available: from author - Flexcam [PDF]
Press: Gizmodo

In my role as Director of Research (Product) at InteraXon (Toronto, 2009–2010), I led rapid-prototyping and R&D for the Muse brainwave headband. I coached the CEO on shifting the company’s positioning—from a research‐prototype outfit into a consumer wellness brand.
I also spearheaded a cost-reduction initiative that cut unit cost by 95%, while enabling real-time neural data acquisition suitable for AI-based state detection.

Highlights

Led R&D and rapid-prototype systems for Muse — a brain-wave sensing wearable intended for consumer mindfulness and neurofeedback applications.

Transitioned the communications strategy and branding from “lab tech” to “mindfulness consumer device.”

Delivered manufacturing and cost reduction strategy: achieved a ~95 % unit-cost reduction while maintaining real-time neural data pipelines.

This project reflects my ability to bridge deep-tech R&D, product strategy, branding and cost engineering—taking neuroscience-enabled hardware from lab‐prototype into consumer-ready form.

#ScienceHack — Open Source Biotech for Everyone
As Co-Founder and CEO of Synbiota Inc., I led the creation of one of the world’s first open-source biotechnology platforms. In collaboration with Genomikon Inc. (a Synbiota company), we launched #ScienceHack — a global experiment in decentralized manufacturing that invited artists, designers, and scientists to collaboratively engineer a micro-organism that produces violacein, a potent anti-cancer molecule.

What began as a weekend hackathon evolved into a proof-of-concept for open, distributed bio-manufacturing — showing that high-impact science can thrive outside traditional labs.

Highlights:

Co-Founder & CEO, Synbiota Inc. — Pioneered web-enabled synthetic biology collaboration tools.

SXSW Interactive Accelerator Winner — Recognized for advancing open science and community-driven research.

Featured in Forbes, VICE, and CBC — Spotlighting a new model for accessible biotech innovation.

Cross-disciplinary leadership — Built and managed teams spanning biology, software, and design to deliver working prototypes and media-ready demonstrations.

This project exemplifies how I build bridges between science, storytelling, and scalable technology — turning ambitious ideas into tangible systems that inspire and empower others.

References

Winner: SXSW Interactive Accelerator

Article: "Growing Violacein Factories w Synbiota" - Britt Wray

Article: "Distributed Science" - Forbes

Article: "Kitchen counter bio hacking" - Joi Ito

Archive: #ScienceHack website

The invention of Kameraflage display technology was a spin-off from my work in computer vision for eye-tracking at MIT. I learned that both CCD and CMOS image sensors can see a broader spectrum of light than a human, and that often the firmware of digital cameras display both Ultra Violet, and Near InfraRed as human visible colours on a display.

Initially a wearable electronic fashion accessory, Kameraflage display technology eventually found it's home in advertising displays where the technology was licensed by Diageo, Activision, and AKQA.

Debut: Paris Fall Fashion Week, 2007.
Winner: Thailand Textile Institute's Top 100 Innovations, 2007.
Invited: CES Technology Fashion Show, 2011.
Patent: #8,531,308