Have you ever interacted with a chatbot? Chances are, you probably have without even realising it.
From virtual assistants on your phone to the (occasionally) helpful pop-up windows on websites, chatbot technology has revolutionised the way we communicate and receive information.
We're working on harnessing this technology to take it to the next level. We're exploring how chatbots can improve people's health and wellbeing.
But before we dive into our innovations, let's cover the basics of chatbots and how they're changing the game in healthcare.
Chatbots 101
Dr David Ireland is a senior research scientist at our Australian e-Health Research Centre (AEHRC). The chatbot developer community has described him as a ‘legend in chatbots for health’.
In our podcast, Everyday AI, David describes a chatbot as a computer program that imitates human conversation. Used in a health setting, chatbots can provide therapy, education or companionship.
Most modern chatbots use artificial intelligence (AI). This allows the bot to interact with people in an increasingly human-like manner.
Users can type, speak, draw, or press buttons on a screen to interact with AI chatbots. The 'brain' of the chatbot processes the information and determines an appropriate response based on the context. This includes the current topic, previous discussions and a profile of the user.
The 'brain' of AI chatbots consists of a combination of natural language processing (NLP) algorithms. These algorithms enable the chatbot to break down, interpret and produce language that we can understand.
But, as with human conversation, it's not just about what words are said. How the words are said is also important.
David and his team model the 'personalities' of their chatbots around health professionals who regularly engage with patients or clients.
Chatbots as champions in healthcare
Researchers at the AEHRC are busy working with speech therapists, occupational therapists, physiotherapists and other health professionals to develop technology that facilitates healthcare delivery.
They’ve created four virtual companions so far.
HARLIE
HARLIE was designed to assist people with Parkinson's disease who face language difficulties. The chatbot converses with people about everyday topics. In the background, the technology analyses aspects of their speech, such as articulation and word retrieval. Worsening of language difficulties can be a sign of disease progression.
HARLIE has also been used with people who are on the autism spectrum. Considerable anecdotal evidence suggests chatbots are good for people who have language difficulties because they offer a safe, judgement-free way to learn and use language.
Dolores
Dolores assists individuals who suffer from chronic pain. It asks the questions that a clinician normally would, such as "where is the pain?" and "how long have you had the pain?". It then provides support and education.
Dolores has two exceptional features.
It allows users to draw how they feel, rather than speaking or typing, which is particularly valuable for people who have difficulty expressing their pain. This includes children and people who are non-verbal.
Chronic pain can take a major toll on a person’s mental health. Unfortunately, chronic pain sufferers report suicidal behaviours more frequently than the general public. David and his team studied language patterns of people with suicidal thoughts. They then implemented algorithms to allow the chatbot to detect and respond to concerning patterns.
Edna
Edna (eDNA) helps patients decide whether to seek further genetic testing. It emulates a genetic counselling session and complements traditional in-person genetic counselling.
Quin
Quin is designed to help people quit smoking by offering evidence-based support, information and resources. The chatbot was built from Quitline calls, and was further refined after consultation with a multidisciplinary team of health experts.
Chatbots can be challenging
Like humans, chatbots will never be perfect.
That's why, when chatbots are used in healthcare settings, people remain involved in the process. Chatbots often flag risky situations and refer users to a human health professional.
Chatbots can’t, and won’t, replace humans. But in healthcare settings, AI can ease the load on medical professionals and give them more time to do what humans are best at: making connections with patients.
It's not a case of humans versus technology. It’s about working with technology to aid, innovate and inspire.