AI for Caring – Technologies needed to use AI in social activities

14-09-2020 |   |  By Sam Brown

Artificial intelligence is said to take over most jobs, and recent advances in the field may prove this to be true. What technologies do AI system need to work in the care industry, and what challenges does AI currently face?

What is AI currently used for?

Artificial Intelligence, or AI, is a technology that allows machines to learn behaviour patterns, and use those patterns to improve their performance. Two decades ago, practical AI systems were far and few between, with researchers only guessing at their possibilities, but now they are used in many applications around the world, and have helped to drive key technologies including self-driving systems, personal assistants, and improve industrial processes.

While AI provides many practical benefits, some are concerned with its introduction into different industries, with the greatest concern relating to automation. Automating any process generally requires some level of AI as the system needs to be able to make decisions based on unexpected events. For example, a CNC drill can follow g-code to know where to put drill holes, and their sizes, but if the process is to be fully automated, then AI may be needed to allow the system to determine if the wrong material has been fed or if the system may be close to failure and require repair. Optical inspection systems can train an AI to observe the work that it has done, and determine if it is up to standard, and thus be less dependent on human operators. What causes concern with AI and automation is the ability that AI has to replace human workers with robotic workers, and that modern automation consumes more jobs than it creates.

Can AI be used in social care?

Advances in technology have helped to improve living standards and increase life expectancy, while smaller family sizes have seen a reduction in global population growth. While this may seem positive, it, in fact, carries with it a very serious problem; an ageing population. Pension systems, which provide economic support for those above retirement age, are generally reliant on a larger, younger population that pays into the system. However, the gradual decline in birth rates in 1st world countries means that governments have to borrow more to pay for the current ageing population to continue providing pension payments. If population growth stagnates, then future generations may not be able to have a state pension. An ageing population also means that the proportion of young to old becomes smaller, and caring for a large number of older people may become too expensive. 

While there are many aspects to social care, one area that AI could potentially be used in is mental health. Humans are social creatures, and it is a well-known fact that older people in care homes can become lonely. Since AI has the ability to learn behavioural patterns as well as the ability to adapt, individual residents in a care home could have access to private AI systems that can provide interactive activities including conversation, games, and even therapy. But for an AI system to work well in such an environment, there are two main areas that they need to achieve; natural language capabilities and natural-looking appearances.

What is Natural Language Processing?

Natural Language Processing, or NLP for short, is an area of AI that specialises in the bridge between computers and humans. Computers are cold, unemotional machines that accept commands with clearly defined parameters, and this is something that most find unforgiving. NPL, however, allows people to talk to a machine as if they were themselves a person and extracts the important information that the computer can read.


Person - “Hi computer, I want to look at my new mail and then play some Skyrim.” 

Computer – Open Mail > Wait till mail closed > Open Skyrim

AI is important in NLP as NLP relies on deep learning methods to recognise keywords in sentences and what they mean. But it is important to understand that this is more than simply identifying keywords because language is fluid, and the meaning of a sentence is not always obvious. Thus, an NLP system needs to analyse large amounts of natural language to recognise hidden meanings.

On the flip side to natural language processing is natural language generation, and this is just as important. Natural language generation, or NLG, is the opposite of NLP and converts computer-generated text into a readable language. For example, a computer response with “OK” could be converted into “Just finished the task for you”. But, just like NLP, NLG requires AI algorithms to understand responses, hidden meanings, and how best to describe crude facts and figures with appealing language.

What is the uncanny valley?

The ability of an AI system to naturally interact with people is only one half of the task in social care; people often bond with physical entities instead of screens. This is where robotics become incredibly important, but creating friendly robotic systems must be done correctly; otherwise, they can be more creepy than friendly; this is where the uncanny valley comes in.

The uncanny valley describes the effect whereby objects that try to look human but can’t fully achieve it make humans feel uneasy, upset, or frightened. It’s called the uncanny valley because if a graph is plotted showing likeability on the Y-axis and the extent of mechanical features on the X-axis the result is a valley in the middle. This means that entirely robotic systems are more friendly and likeable than robotic faces which present human features, but do not quite have the realism needed to fool the brain. 

It can be seen here that the robot from “I Am Mother”, and the real face are both more appealing than the middle image of an artificial face.


For AI systems to apply to social care environments, multiple technologies need to be considered, including NLP, NLG, and friendly robotic interfaces. While researchers essentially have NLP and NLG covered, the next step will be to create emotive feedback systems that can respond to subtleties in human language that may help to identify a person’s mood. At the same time, developers need to understand that trying to make robotics look realistic is only worth doing if the resulting face is realistic; features that are slightly off will create AI systems that are more upsetting than helpful. More interestingly, a social carer that is fully mechanised with no human features may even prove to be just as effective as a person. At the end of the day, a quality conversation may be the most important factor.

Read More

By Sam Brown

Related articles