Class of ’84 | With the baby-boom generation approaching its golden years, the population of older Americans is poised to more than double by the year 2030. At the same time, the pool of people willing and able to take care of the elderly is shrinking. For seniors with cognitive problems, who are often forgetful or disoriented, the scarcity of available caregivers is particularly troublesome.
Enter Pearl. A personal robotic assistant, Pearl’s mission is to make life easier for the elderly and infirm. Pearl will never replace human companionship, but with her quiet voice, sassy red lips, and “chest”-mounted touch-screen, she is well suited to attend to many of her clients’ daily needs. She is able to physically accompany them to medical appointments, remind them to eat and go to the bathroom, and even deliver their medication at the right time.
Pearl is the product of a team of engineers from four universities, but her intellectual prowess is the brainchild of artificial intelligence expert Dr. Martha Pollack GEE’84 Gr’86. Pollack, who teaches electrical engineering and computer science at the University of Michigan, designs cognitive orthotics, sophisticated tools that help people with decreased memory and executive function (a group of mental activities that organize and plan the flow of behavior) live as independently as possible.
She recalls a meeting a few years ago when the discussion turned to how artificial intelligence could help design systems useful to elders. A nurse explained that a serious problem for older adults is keeping track of their daily activities. Pollack, who had spent nearly a decade developing theories to enable computer systems to plan their own actions, was intrigued. She thought, “One thing I really know how to do is to model and monitor the execution of plans, and that’s what these people are having trouble with.”
At a cost of about $100,000, Pearl’s usefulness, when fully developed, is probably limited to institutional settings where she and her progeny can ease staffing crunches. To make the technology more readily available, Pollack is also working on a smaller device called Autominder. Though it uses the same software that makes Pearl smart, Autominder looks more like a personal digital assistant (PDA) than a full-scale robot. (Pollack would ultimately like to see the technology incorporated into a watch, pendant, or other wearable device.) Unlike reminder systems on the market today, which are essentially talking alarm clocks, Autominder is able to reason about whether and when to issue reminders.
For example, a diabetic may need to eat every three hours and to take his medication on a full stomach. Current technology can prompt him to eat and take his medicine at 9 a.m. and then to have lunch at noon. “But people don’t live their lives on fixed schedules,” Pollack notes. Autominder solves this problem by issuing reminders based on the user’s actual activities, not in accordance with a predetermined timetable. It can also recognize that, in some cases, a reminder may not be necessary at all. So if the diabetic remembered to eat and take his medicine shortly after getting up in the morning, regardless of the exact time, Autominder would keep quiet. But if he had breakfast and an hour passed, it would remind him to take his medication as well.
For Pollack, the appeal of artificial intelligence is an extension of her long-time interest in language. At Penn, she studied natural-language processing, which involves enabling computer systems to speak and read natural languages (such as English and Japanese) rather than computer languages. “I was incredibly interested in cognition and thinking, and how people did those things,” Pollack remembers.
Since then, the fields of artificial intelligence and cognitive science have largely gone their separate ways, and Pollack’s approach to artificial intelligence has also evolved. Once interested in artificial-intelligence (AI) algorithms in the abstract, she now looks for ways to put those algorithms to work.
“What really excites me is to ask what we can do to an AI algorithm to make a device that helps someone live life better, rather than asking what we can possibly do with the algorithm after we already have it,” she explains.
While the current Autominder prototype relies on input from the user to determine whether it should issue a reminder, Pollack sees future versions ascertaining that information themselves by monitoring the user’s activities with environmental sensors, motion detectors, and the like.
Mindful of the fact that many older people are not technologically savvy, Pollack stresses that her goal is to create a simple system, more akin to operating a toaster than a VCR. She is currently field-testing early versions of Autominder with younger victims of traumatic brain injury, many of whom suffer impairments similar to those experienced by older adults, and she expects that the device could be available commercially within the next decade.
“All of the studies show that people want to stay at home longer, and it’s significantly less expensive than moving into an institution,” Pollack explains. “It’s a win-win situation.”
—Stefanie A. Doebler
