ongoing

completed

Home Health

The Home Health system, a collaboration with Bill Gaver and Michael Golembewski at the Royal College of Art, London, is a ubiquitous computing system that monitors a home's emotional climate and provides open-ended feedback about it to users. Everyday household objects are wired with sensors. The resulting sensor data is used to develop a model of the current emotional climate of the people living in the home. Once a day, the user receives a list of suggestions from the system of emotional issues that s/he might wish to consider.

This system is not intended to solve a goal-oriented task, but to provide an opportunity for reflection in a whimsical way. In order to support reflection, the system creates natural language descriptions based on the metaphor of horoscopes. Many people read horoscopes every morning to take a look at the emotional issues that may come up over the day. Similarly, our system can once a day output a natural language description of the issues in the house in a horoscope-like style: "You've been enjoying peace, but are you getting stuck in a rut? Consider exploring new interests." This 'horoscope' does not demand belief - it is open to interpretation by users in any way they choose.

In designing systems that relate to the emotional state of the house, we are entering an intimate, personal territory. The meaning of sensor data like 'the lights are out' will have very different interpretations for different families and at different times of the day. We are tailoring the system deeply to each particular family by interviewing the family and determine the local meanings of the sensed activities.

At the same time, an approach that attempts to directly identify and measure actual emotional state of the family would be inappropriate for this system, not only because it would be invasive of privacy, but also because the system is less about the actual emotional state of the family and more about helping users reflect on the meanings of home activities. We are taking two approaches to developing alternatives.

First, instead of using a physiological or statistical approach - wiring large portions of the house and looking for correlations between sensor data and tested user state - we are using an approach based on semiotics, or the philosophy of meaning. Using this approach, we will look for a small set of rich and meaningful signs that can be directly measured but represent a broader set of emotional implications. For example, slippers can sense when they are being worn; overall, this represents relaxation, but, if worn during the workday, can represent sickness and change in routine. The meaning of these signs is not universal; much depends on the nature and habits of the individual family. To uncover these meanings, we are drawing on our interview data.

Second, because we cannot expect to have a scientifically accurate model of the family's emotional state, we need to work some play into the 'horoscopes' that the system generates. We are developing natural language processing algorithms that can mimic the rhetorical openness of horoscopes - leaving room for user interpretation - while maintaining a balance between being open to interpretation and also reflecting accurately the current emotional state of the home as represented by the sensors.