A Very Brief Overview of AI in Assistive Technology

                I found an intriguing article by Martha E. Pollack that emphasizes the socioeconomic concerns with regard to an aging population, in contrast to the last blog post from the IT Perspective. The article began by establishing the need for further innovation in intelligent assistive technology and follows through to survey some of the technology being deployed. (I should also mention that this research is a few years old, so some of the information is a little dated; however, the concepts remain very relevant .)

Currently, over 10 percent of the world’s population is over the age of 60. Despite the overall problem of worldwide population growth, there will continue to be an disproportionate increase of the old and the oldest old age groups. This is especially a concern considering the strong correlation between age and cognitive impairment involving Alzheimer’s, in particular. There will also be fewer young people to assist in their endeavors.  The author states that “although the shift is most dramatic in the more industrialized regions of the world, a significant growth in the percentage of older adults is expected in virtually every country.”

                With advances in technology that will help assist, compensate, and assess the impaired, society as a whole may become sounder.  After all, institutionalization comes as a great financial burden.  The article claims that over 132 billion dollars are spent on U.S. nursing home bills a year, of those 60 percent are covered by the government.  Many of the elderly prefer to age at home, but fail in performing everyday activities or feel socially isolated. The process of aging is quite difficult for the caregivers as well. Many people do not have the available time, and even then struggle to be consistently attentive. It is often more feasible to rely on today’s technology even with non-computerized inventions such as lift chairs or ergonomic handles.  Nonetheless, there is much AI –based research being performed with the goal of curbing some of the difficulties involved with cognitive and physical impairments.

                As I implied, research can be placed into three main categories:  those that assist, compensate, or assess.  With advances it will be easier to assure safety in performing daily activities by assisting or assessing deviations. Systems may even warn the user or caregiver of any errors. If for instance, a person falls or forgets to take their medicine, a caregiver’s phone could be called. Activity monitoring is essential in any of the methods.

                One method is using radio frequency identification tags. RFID chips are definitely a major concern when it comes to privacy, but the convenience in data collecting is tremendous. By tagging objects or clothes, detailed data can be collected about the user’s interactions, locations, and physical well-being. This way a system can infer specific daily activities. Most technology used for recognition use dynamic Bayesian networks that filter and derive probabilities. For example, the PROACT uses a DBN and the user wear s a glove that includes an RFID reader. Such as making tea can be broken down into a three-step process. In step two the system may derive that there is a high probability of using a tea kettle. Time can be monitored along with it. Based on trends, the system can target unusual activity.

                Because wandering is a significant concern, systems such as Opportunity Knocks employ GPS tracking. The system may learn typical behavior and preferences. Deviations can be spotted based on normal trends. The IMP or intelligent mobility platform uses similar techniques. It uses a “semi-Markov” model having three layers:  metric position, topology, and current activity. This system makes a map as it goes through a partially observable environment and monitors trends. A user can then select rooms/locations for sequentially displayed directions—and of course, it uses a large arrow to show the current direction needed.

        Planning systems are in high demand as well. Many of the elderly simply struggle to remember what they need to do next.  Interesting enough, many modern systems don’t even require exact times to be inputted.  If a person is typically awake and eating at 8 a.m. every morning, schedules can be shifted accordingly or a caregiver may be notified with lack of activity. If a person needs to take a pill an hour after eating, then it can be scheduled based on when sensors find the user has eaten. Some even will avoid specific times that are normally set aside. If a user likes a particular television show, the ‘eat’ notification may occur before or after the typical time. The current focus is for them to be even more dynamic. Critics argue that the systems are too fixed with regard to the decision logic; problems may occur if a user becomes ill or the user ages more.

       Personally, I had barely considered that some of the elderly struggle to simply following sequences. Simple tasks, like washing hands, have become not all that simple. Many try to dry their hands before even washing them. I also think advances in assistive technology will revolve around AI. Nonetheless, I remain generally unconcerned. As the need for the technology increases, I am confident that even better solutions will be employed. Hopefully the technology can then be applied to an even wider audience, further improving everyday conveniences.