ASSISTIVE AI TECHNOLOGY AND THE ELDERLY

Presenter(s)
Donald Egolf
University of Pittsburgh
Department of Communication, 1117 CL
Pittsburgh PA 15260
Day Phone: 412-624-6763
Fax: 412-624-1878
Email: ratchet@pitt.edu


This presentation's purpose is to discuss the developments in artificial intelligence designed to assist the elderly. The most recent research will be reviewed.

INTRODUCTION

Pollack (2006) notes that in the world today approximately 10% of the world’s population is over the age of 60, and that by 2050, this percentage will more than double with the greatest increase seen among people age 85 and over. In the United States in the year 2000, people age 65 and over made up 12.3% of the population; in 2030, they will make up 19.2% of the population. Striking about these statistics is that there is just not growth in the absolute number of adults over 65, but the proportion of the total population that this group occupies will increase as well. As a result, fewer people will be available to serve as caregivers for the elderly.

Can technology help to relieve this personnel shortage by assisting the elderly in their quest to remain healthy and independent as long as possible? The purpose of this paper is to address this question. In particular, the paper will look at artificial intelligence (AI) approaches that have been designed to help elders with cognitive impairments.  Organizing the paper will be the three-part model proposed by Pollack (2006). Pollack suggests that AI systems can assist the elderly with cognitive impairments in three ways: (1) by providing assurance to the elderly, (2) by helping the elderly compensate for their impairments, and (3) by assessing an elder’s cognitive status. These ways will be discussed under the subheadings of assurance, compensation, and assessment.

AI SYSTEMS

Assurance

Assurance systems are designed primarily to ensure the elder’s safety and to reduce the burden on caregivers. Mercury switches and heat, light, weight, and motion sensors are used, for example. More recent and more sophisticated systems utilize RFIDs (Radio Frequency Identification Tags). These tags were first used commercially to monitor and control inventory. They are used in elder AI systems to monitor an elder’s movements so that the elder or a caregiver can be warned if the elder is in harm’s way. The RFIDs can be worn by the elder and thus the elder is a message sender and in this role RFIDs can allow the elder to control the immediate environment, or they can be placed in the elder’s environment to communicate to the elder. Messages can be sent not only to the elder and the elder’s caregiver, but also can be sent to a central processor for analysis, trend analysis, for example. Some messages trigger an emergency response, when the elder falls, for instance. Other messag!
es arouse curiosity, say when an elder who is usually active in the kitchen at a certain time of day, but is not on a given day. Philipose et al. (2006), for example, have used RFIDs to facilitate elders’ performances of their activities of daily living.

Compensation

Compensation systems help elders navigate, manage schedules, and recognize faces and objects. One of the most advanced compensation projects is the Nursebot Project described by Matthews et al. (2002). Two robotic devices were described. The first, PEARL, is a mobile robotic personal assistant. She is similar to ET, standing 4.5 feet tall; has a three-dimensional head with camera-imbedded eyes, movable eyelids, eyebrows, and lips; and she has a head that flexes and rotates. PEARL can recognize and synthesize speech, she can recognize faces and objects in the environment, she is powered and can navigate alone or lead others, and has handlebars and a basket so she can serve as a walker/guide/companion. With her navigation software PEARL can lead her elder companions to very precise locations. Finally, Pearl has a computer-driven scheduler and can remind elders of appointments, medications, and can ask questions about food and water intake, for example. The second robot designed !
on the Nursebot project is the Intelligent Mobility Platform (IMP), an off-the-shelf walker modified to be power driven and computer controlled.  The IMP can learn the topography of an environment, navigate through that environment, and is equipped with a seat, basket, handlebars, and brakes for its user.  Other compensatory systems expand the range of the elder. Opportunity Knocks (Liao et al., 2004), for example, uses GPS and is designed to remind the elder and the elder’s caregivers of the elder’s whereabouts when the elder is outdoors.

Assessment

Assessment systems are designed to assess a person’s condition, particularly the elder’s cognitive condition. Jimison (2003), for example, has elders play a computer-based solitaire game, and each time they play, their performances can be compared to a set of norms to check for any changes in cognitive functioning. The advantage of this system is that it gives almost daily feedback on the elder’s cognitive performance. It is therefore not susceptible to the possible errors that can accrue when an elder is tested infrequently and on the day of a given test may give an atypical performance.

ANECDOTAL REPORTS

Two anecdotal reports support the efficacy of the AI research efforts. Shellenberger (2002) became a “patient” in a high tech nursing home (Oakfield Estates in Milwaukie, Oregon). At the home sensors dotted the walls, cameras marked boundaries, beds were wired to detect patients’ movements, and Shellenberger wore a lanyard around her neck that served as a room key, alarm, and location monitor. Shellenberger’s brother Dave was her “caregiver” 2300 miles away, and he was able to tell her about her every move even when she was sleeping. Did all the monitors bother Shellenberger? No. She said that she quickly forgot about them and that she gladly tolerated them because they gave her freedom of movement.

Jajeh (2004) visited the Longwood Retirement Community in Oakmont, Pennsylania where the robot PEARL is one of the residents. Jajeh found that PEARL is one of the most popular residents there, whizzing around reminding residents to do things (e.g., take meds) and guiding them from location to location. Because PEARL can recognize faces, she can give very personalized service. “She’s won the hearts of the elderly,” said Jajeh.

SUMMARY AND CONCLUSIONS

Certainly the above assurance, compensation, and assessment examples as well as the related anecdotal reports suggest that great progress is being made in efforts to help elders maintain levels of health and independence that would not be possible without AI systems. Continued research should further advance this progress. At the same time caution is in order. Elders should not be abandoned to technology.  Practitioners in augmentative and assistive technology fields know that often after devices are purchased for deserving clients there is inadequate training and support given in the use of the devices.  This issue must not be forgotten. It is possible too that as each new generation enters elderhood, that the new generation will be more technologically sophisticated, and that the incorporation of AI assistive devices in the daily lives of the new generation’s members will become routine. But in all cases, AI technology should serve to complement and enhance the interpersonal!
experiences of AI users, and overall, should serve to maintain and improve the users’ quality of life.

REFERENCES

Jajeh, D. (2004). Robot nurse escorts and schmoozes with the elderly. http://www.intel.com/employee/retiree/circuit/robot.htm,  August 24.

Jimison et al. (2003). Adaptive interfaces for home health. In the 2nd International Workshop on Ubiquitous Computing for Pervasive Healthcare. http://www.healthcare.pervasive.dk/ubicomp2003/papers/.

Liao et al. (2004). Learning and inferring transportation routines. In the Proceedings of the 19th National Conference on Artificial Intelligence. 348-353.

Matthews et al. (2002). Robotic assistive technology for community-residing older adults and persons with disabilities: an inter-institutional initiative for students in the health and technology fields. .In the Proceedings of the International Conference on Engineering Education.

Philipose et al. (2006). Inferring ADLs from interactions with objects. IEEE Pervasive Computing. In press.

Pollack. (2006). Intelligent technology for an aging population: the use of AI to assist elders with cognitive impairment. A preprint to appear in AI Magazine in 2006.

Shellenberger. (2002). Can technology ease elder-care concerns? http://www.careerjournal.com/columnists/workfamily/2200719 

Go to previous article
Go to next article
Return to 2006 Table of Contents