Digital Empathy: Bringing Emotional Intelligence to AI Conversational Agents

Empathy is a cornerstone of meaningful human interaction, and its absence in AI agents can undermine user trust, personalization, and true alignment with human goals. This effort systematically examines what “digital empathy” entails for conversational AI—identifying the contexts where empathic responses matter most, defining practical mechanisms for simulating empathic understanding, and assessing how such behaviors can enhance both personalization and users’ own capacity for empathy. Through iterative studies and principled design guidelines, the project seeks to delineate when and how empathic strategies improve alignment without overstepping ethical boundaries, ultimately charting a responsible path toward AI systems that respectfully recognize, adapt to, and even nurture human emotions.

Promoting Workplace Wellbeing in the New Future of Work

Hybrid schedules, ubiquitous AI tools, and fluid time-zones are reshaping how—and when—we work, creating new opportunities but also new strains on productivity and wellbeing. This effort tracks the evolving rhythms of knowledge workers by combining calendar data, interaction logs, and unobtrusive sensing to reveal how shifting routines map onto focus, stress, and performance. By translating these insights into personalized dashboards and just-in-time nudges, the project explores how technology can align work patterns with human needs, ultimately fostering healthier, more sustainable ways of working in an era of constant change.

Emotional Navigation: Bringing Emotional Intelligence to the Driving Experience

[web, video]

Before automobiles were invented and widely adopted, animals like horses were the most common mode of transportation. While this change brought significant improvements in terms of reliability and efficiency, it also removed a core component: the emotional relationship that existed between the person and the animal. While largely ignored, the emotional states of drivers are quite important, as they influence not only driving behavior but also the safety of all road users. For instance, driving can be quite an emotionally stressful experience and, while certain amounts of stress help the driver to remain alert and attentive, too much or too little can negatively impact driving performance and safety. Furthermore, stress in large doses has been linked to a large array of adverse health conditions such as depression and various forms of cardiovascular disease.

Global Vitals: Democratizing Physiological Sensing

Pervasive physiological monitoring offers the possibility of better capturing the health condition of people. This information is not only useful to help inform medical diagnosis, but also to help track chronic health conditions and therapeutic interventions as well as understanding the emotional state of people. While great strides have been made to provide comfortable physiological monitoring, traditional methods still require attaching electrodes to the skin, buying expensive custom-made hardware, and/or interrupting daily activity to collect measurements. Moreover, these sensors tend to measure a limited set of metrics at a time. In contrast, Global Vitals offers an alternative solution that provides accurate, low-cost and comfortable multi-physiological parameter with the goal of democratizing physiological sensing.

SenseGlass: Using Google Glass to Sense Daily Emotions

SenseGlass is a tool that uses Google Glass to measure the emotional state of the wearer during daily life. SenseGlass collects peripheral information using the embedded sensors of Google Glass (i.e., head motions, vocal features, and skin conductance) and provides feedback on the user's emotional state. In particular, SenseGlass monitors the physiological arousal and valence of the user using machine learning techniques and makes the user aware of their emotional state through a dynamic wallpaper on the Google Glass screen. By helping people increase their emotional self-awareness, SenseGlass aims to help people recognize their emotional state, especially negative states, and take the appropriate actions for improved wellbeing.

BioPhone: Physiology Monitoring from Peripheral Smartphone Motions

Can we measure physiological parameters like heart rate and breathing rate just by having a smartphone in our pocket or bag? This project explores this possibility by analyzing the subtle motion signals captured by a smartphone's accelerometer while the user carries it peripherally (e.g., in a pants pocket). We collected data from participants performing various activities (sitting, walking) while carrying a smartphone and compared the estimated physiological rates with ground truth measurements. Our findings suggest that while challenging, it is sometimes possible to extract estimates of heart and breathing rates from peripheral smartphone motions, particularly during periods of relative stillness like sitting. This research highlights the potential, and the challenges, of using ubiquitous devices like smartphones for passive physiological monitoring.

Mood Meter: Counting Smiles in the Wild

Mood Meter is a large-scale, long-term monitoring system designed to automatically detect smiles from people interacting with a public display. We deployed Mood Meter in the lobby of the MIT Media Lab for several months, collecting data from thousands of interactions. The system uses computer vision techniques to identify faces and analyze facial expressions in real-time, specifically focusing on detecting smiles. By tracking the frequency and duration of smiles, Mood Meter provides insights into the overall positive affect of the community interacting with the display. This work demonstrates the feasibility of using unobtrusive sensing in public spaces to gauge collective mood and explores potential applications in areas like environmental design and social wellbeing assessment.