Helping the Elderly with Machine Learning

According to 2012 estimates of the U.S. Census Bureau, the U.S. population aged 65 and over is projected to be 83.7 million in 2050. I am sure that the majority of you are no strangers to the idea of a depleting social security, and the lack of ability of the “younger” generations to support our aging population in the U.S.. As the population in the U.S. ages, however, these are just a couple of the issues they face.

In fact, this article from the Wall Street Journal does a great job of personifying some of the personal struggles senior citizens face; in particular, those in rural areas with limited access to social programs and family. One of the saddest aspects of caring for senior citizens means tearing them from what they know to be their home and their space to live in senior care facilities. Often, the declining health of our senior citizens, however, impede that they maintain their independence. As we age, we face health issues we most likely did not have to face before. For example, our vision and hearing significantly deteriorate (I’m sure you have had to raise your voice when speaking to an older adult before!), our ability to walk and make controlled movements also deteriorates, as well as our general dexterity.

Machine learning, surprisingly or not, can help at least alleviate some of the challenges that come with caring for a senior citizen and/or prolonging their independence. Below, I provide 5 different examples in which machine learning can help improve the life of older adults.

  1. Providing help for the visually impaired. Missouri is one of 15 states to implement a program to help people with disabilities (e.g., being visually impaired is one example) have access and be able to use telecommunications. Missouri’s Telecommunication Access Program (TAP) provides technology and seminars for people with disabilities (many of whom are older adults). The smartphones, tablets, and other devices provided by TAP help older adults use telecommunications and have a wider access to technology. The Columbia Tribune (in this article) introduces Barb Griffin, a 69 year old woman who can only see lights and shadows. She extensively uses her smartphone, which reads to her everything that is on the screen, to check the weather, check the color of her clothes (she describes that she previously needed help dressing because she could no longer match her clothes), and even help her paint. The current programs play an assistive role in the lives of seniors now. However, with advances in technology, and in particular, computer vision, it would be possible for these smartphones and tablets to do more for senior citizens. For example, the smartphone can become a personal assistant and care taker — it could reccomend walking routes based on expected pedestrian traffic, it could not only advise what the weather is, but even provide a suggestion for what a person should wear (e.g., “there seems to be a breeze outside, your light blue jacket would be perfect!”). It could also describe to the senior citizen, not only what is on the screen, but also what is on the street. Imagine a narrator describing the sights of the neighborhood as an older adults walks with his/her sight seeing dog. the walk would certainly be much more enjoyable and interesting, thus encouraging the senior citizen to walk more often and remain healthy for longer. All of these additional features would use machine learning (recognizing sights without previously being told what to expect, suggesting clothes worn by the user, determining an expected pedestrian traffic). The technology now is not far from my suggestions, and I suspect many of these features would be available in the near future.
  2. Providing active alerts on the safety of the senior citizens. Many are familiar with the “red button” or the emergency signal in many nursing homes. However, these devices require that the user is a. conscious enough to press the button (or pull the string) and b. within physical reach and physical condition to press the button. Many times, however, when accidents occur, this is not the case. An innovative solution presented here in Hong Kong’s version of Computer World, proposes an active home-based elderly care system. The article describes various sensors positioned in different rooms of an older adult’s home. These sensors would be able to monitor presence and motion. A computer system (to which these sensors would be connected) would “observe” the activities of the adults during a “learning period.” The computer would then be able to identify daily habits of the adult and be able to generate alerts when something is up (e.g., something unusual occurs). This technology still needs further development before becoming commercially available, but it is easy to see its advantages over the alert system. Through the proposed active care, an older adult may continue to carry on independently without having his/her family worry about missing a fall, having him/her fall unconscious without someone else finding them, etc. The proposed active care would instead generate an alert to which care takers can respond to by checking on the older adult.
  3. Providing navigation help with people with dementia. In the past year at least a few silver alerts were generated by the authorities in my cities. We can only expect these types of alerts to increase as our population ages, but their repeated occurrence doesn’t make them any easier to handle. Older adults often suffer from dementia, Alzheimers, and other memory loss disorders. Sometimes these older adults go on walks only to find themselves lost. It doesn’t seem far fetched for an application (e.g., on a smartphone) or on a “red button” type alert system that “follows” the user (e.g., the older adult) and learns their daily routine, their most visited places, and where they like to walk. The system would then be able to sense when the older adult has deviated from their known routes and may offer assistance (for example, “You are trying out a new route! Let me know when you want to head back home”) Using friendly tones instead of “You are deviating from your route, turn around” will help users still feel in control and may make them more prone to using the application. The application can also let the adult know when they are “too far” (for example, 3 miles) away from home and nudge the user to return home before getting further. In some circumstances (e.g., the older adult doesn’t turn to go back home and they are far away), the application may send an alert to a family member or home taker and may, in some implementations, send the location of the older adult to the family members.
  4. Much hype has been building around the “connected home,” in which many of the home appliances are connected to the internet and accessible through an application on your phone. There are many examples of this (think: hue lights from Phillips, water heater applications, nest thermometer, among others). Senior citizens sometimes forget to regulate the appliances in their home safely. For example, they may accidentally increase or decrease the heat to the home (without realizing that this may be unsafe — particularly in WI winters). Including safeguards in stovetops (especially gas stovetops), heating units, lamps, and the like can prevent devastating accidents. It is foreseeable that a “connected home” system would be able to learn the user’s daily habits (habits of senior citizens) and automatically turn off or at least alert users or care takers when certain conditions are present (e.g., the stove has been on for three hours, the heating system is below/above a reccomended given the time of year, etc.)
  5. Lastly, health is another aspect that strongly impacts a senior citizen’s quality of life. Many different health issues affect different demographics of senior citizens. One of these health concerns includes arthritis. This paper provides some research and indication that machine learning can improve the diagnosis and treatment of arthritis. While machine learning has been used to diagnose different types of disorders and illnesses, it is relatively new (as explained in the hyperlinked paper) to also determine how well a specific treatment plan is working on a disorder. It would be feasible for some of these tools to be accessible to medical professionals in the near future. For example, using machine learning (e.g., regression models and classifiers) a medical practitioner may be able to decide whether a patient is progressing adequately in his/her treatment of a particular. A patient may then try other treatments if a particular drug or treatment regimen is not working as expected. Such information woudl greatly decrease recovery time in patients, and specifically in senior citizens.

As dicussed in the five examples below, machine learning offers unique opportunities to help improve the quality of life of senior citizens. I have no doubt that technology will continue to provide new solutions to problems. I also suspect it will help provide a new set of problems, but hopefully we will continue to have curious and hard working people who will seek even more answers.