Martha Pollack ’79, a professor at the University of Michigan, spoke at Dartmouth last Tuesday about computer science developments in assistive technology – a field dedicated to helping elderly people and patients with cognitive deficiencies with their daily lives.

Pollack began her lecture by noting that our country is getting older; by 2050, 40% of our population will be over 60 years of age.  Many other countries are facing similar trends in aging.  Because there is a strong correlation between sensory, emotional, and cognitive ability and age, assistive technology will play an increasing role in the next half a century and beyond as a greater proportion of the population will develop cognitive impairments associated with age.

Assistive technology can help by providing assurance for a person’s safety, helping to compensate for cognitive deficits, and allowing for better diagnosis of impairments.  Pollack noted that assistive technology is a “body of work that combines fundamentally interesting questions in computation with societal relevance.”

A prime example of assurance is Carewatch, which relies on alarms to notify caregivers if an Alzheimer’s patient gets out of bed or leaves the room.  Caregivers have reported that not having to check on patients to make sure they are still in bed has had a huge impact of their lives.

Pollack has devoted a large portion of her research to a system called Autominder – a personalized schedule management program that will help patients compensate for cognitive deficits.  Autominder does not issue reminders to patients to take their medicines or use the restroom at the same time every day.  Rather, it uses intelligent reminders that combine alarms if they are set to go off at nearly the same time.

The architecture of Autominder consists of two main functions: the plan manager attempts to fit all alarms into a certain time boundary and intelligently decides when to issue reminders.  For example, if a patient needs to brush his teeth, take a bath, and eat breakfast in the morning, the plan manager will organize these alarms.  First, it knows that the patient cannot eat breakfast and take a bath at the same time he eats breakfast.  It also knows that the patient should brush his teeth after he eats.  Lastly, the plan manager knows that the patient will be in the bathroom when he takes a bath and brushes his   teeth.  Therefore, the plan manager will issue two reminders: first to eat breakfast, then to brush his teeth and take a bath.

The goal of the client modeler is to recognize what activities are being performed by the patient.  Future research in this field will need to tackle significant challenges, including human activity recognition.  Pollack speculates that once we start recording human activities there will also be an increased need for mechanisms for privacy and security and encryption and data recognition policies.