Just as the brain is still the most elusive part of our body, artificial intelligence is still an undiscovered land. As human beings, we have come to accept that our brains control our bodies. That is not the case for the way in which AI is taking control of our society step by step. We would like to hold a few vigorous debates about this. In close cooperation with the Netherlands AI Coalition, Innovation Origins reveals in a series of reports and interviews what the average Dutch person notices of this all-defining social revolution. How do we, as human beings, keep our finger squarely on the pulse? The fears, the opportunities, the dilemmas. Today part 2: Chatbots in healthcare.
The customer service departments at banks and large corporations employ not just people, but robots as well. Chatbots look up data that is in the system and assist with frequently asked questions. This does not always run smoothly, although the technology is rapidly evolving. An algorithm that searches information at lightning speed and can help you within one second opens up opportunities in various disciplines. An assistant, a buddy, a professional, a chatbot is capable of performing a wide range of functions.
Let’s zoom in on healthcare for a moment. A sector where the work pressure has been rising steadily in recent years. Digitalization, automation, and consequently artificial intelligence can offer solutions. Not necessarily to replace the care provider directly, but instead to support them in their work. And to alleviate the workload, thereby reducing any risk of errors. One of the options out there is a chatbot. And not one that pops into view with a number of standard questions when you want to buy something from a webshop, but a well-trained algorithm that has up-to-date and correct information.
How does that work in practice? An example: The Maxima Medical Center in Eindhoven (the Netherlands) used a chatbot during the first corona wave to answer questions from employees. Flow.ai developed an algorithm that could answer the most frequently asked questions, e.g. about safety regulations and schedules. Employees could ask a question via Whatsapp. The chatbot analyzed the texts based on specific keywords. Using those keywords, it then searched a database with potential answers and sent a message with the most appropriate answer. If that didn’t suffice, employees could always contact their manager. “We were able to reduce the work pressure on managers by filtering questions,” says Sander Wubben from Flow.ai.
Read more about the chatbots from Flow.ai here.
This turned out to be a great solution for employees. But what about patients? General practitioners fulfill an important task in society. People with all kinds of questions and complaints turn to them first. But we often have to wait a while in the morning in order to make an appointment. What’s more, consultation hours are soon filled up. A chatbot could be a solution for this. “We’re already seeing a shift towards digital consultations with GPs,” says Yuan Lu, associate professor of Design for Healthy and Active Ageing’ at the Eindhoven University of Technology (TU/e). Patients can then send in a photo of, for example, a wound that they have some questions about. “The doctor looks at the photos and tells the patient if they should come by for a follow-up examination. If that isn’t necessary, the GP can give advice over the phone.”
A chatbot quickly provides patients with some initial advice and relieves the pressure on GPs.”
“A chatbot would be able to answer frequently asked questions, either via a chat message or over the phone,” she continues. Chatbots can be either text- or voice-driven. Even though, according to the researcher, the voice-driven algorithms are still not as advanced at the moment, developments are progressing rapidly. “If an algorithm can’t find a suitable answer, the question is immediately forwarded to the doctor,” Lu goes on to explain.
This makes a chatbot a kind of an assistant for the doctor’s assistant. “The system quickly provides patients with some initial advice and relieves the pressure on GPs” she says. Lambèr Royakkers, a robot ethicist at the TU/e, also agrees with this vision of the future: “I can imagine that this will be one of the fields of application for this technology. Although of course, we have to make sure that the chatbot only has access to the right kind of information so that it can’t say the wrong things”.
It all sounds like a relatively simple solution, but how do patients really feel about it? We asked a number of random Dutch people. “I see it as a kind of virtual helpdesk for asking all sorts of questions” is what is often mentioned. “Especially at times when the doctor is not readily available, for example in the evenings or on weekends.” The respondents would mainly turn to the chatbot for routine treatments and general questions, for instance about medicinal use or for blood or urine test results.
The fact that a chatbot is always available is seen as a major bonus. Apart from that, someone sees it as a useful replacement for Googling info themselves. “Nowadays, I often search the internet, but then I end up looking at some really horrible stuff.” A chatbot that has information that has been verified by health authorities can resolve this problem. Despite the positive aspects, people are also cautious. For example, hardly anyone wants to share personal data with a chatbot. “That’s too prone to abuse” is a key argument. People who would like to share data only pass on “strictly necessary data in order to get the best possible answer.”
According to research conducted by Yuan Lu, a chatbot can also play a role in acute care. “If the voice-driven chatbots are developed further, they can also improve things at the emergency call centers. An algorithm can quickly analyze what someone says when they call 112. This enables it to alert the emergency services very quickly and send them somewhere,” she adds.
Robotethicist Royakkers goes one step further: “There are even countries where chatbots are being developed that are capable of analyzing a person’s state of mind. Misuse of the emergency number can be prevented by quickly registering whether someone is panicking or not. This saves lives.”
As well as improving acute care, a chatbot can help with preventive care. Lu is researching how a chatbot can be used to help people improve their lifestyles. “Everyone knows that eating healthily is important, but it is not easy to stick to a diet,” says Lu. “Dieticians often ask people to fill in questionnaires and keep a diary. Many people eventually forget to do this or lose their motivation.” A chatbot can then help them out. “At a set point in time, the chatbot asks a number of brief questions and responds to the answers with tips and encouragement,” she goes on to say. The way the chatbot works and the way it responds to people will always be coordinated with professionals. Keeping a diary can become interactive that way and people would get the support and motivation they need right away.
We need to think carefully about how and where we want to utilize chatbots.”
These kinds of applications for a chatbot collect data about an individual’s personal life. “Suppose there really is a technology that can measure your state of mind, such as the one in the emergency call center. That would be an invasion of your mental privacy,” says Royakkers. “You can never easily say that things are going well when they aren’t. We have to think carefully about whether we want to do that. We also have to think about how we want to protect this privacy. There are other security issues surrounding the collected data as well. The TU/e has already given this some thought: “The data is stored in protected environments. In addition, we make sure that the data can never be traced back to a single individual,” Lu explains.
Chatbots can be used in a variety of ways across different sectors. They can provide convenience and service as a helpdesk for a webshop. In addition, there are plenty of applications in the healthcare sector that provide people with better care. This technology is still under development. Today’s chatbots are not perfect, but practice makes perfect. Royakkers: “It’s a rapidly developing technology that offers plenty of potential.”