For a moment, think about all of the ways technology has become part of how humans keep fit, stay healthy, and improve their lifestyle. We go to the gym, and the treadmills there are smart enough to tell us how fast we are going, how much resistance we are facing, and how far we have travelled. Pedometers can be attached to our bodies to tally up our strides for the day. Fitbits track our fitness levels and act as our personalised trainers. Earphones provide a soundtrack to our exercises to help keep us focussed. We use a multitude of devices to look up nutrition tips, or ways to eat healthier. And hospitals rely on various pieces of technology to keep people alive.
I say ‘we’ and ‘our’ a lot, and I use these quite generally — not everyone has a fitness regime or a gym membership, or a devotion to certain kinds of health habits. No matter how into it you are, or aren’t, it’s clear that staying healthy often means interacting with some form of technology.
And we know that technology often evolves very quickly. We’ve come to the stage in chatbot development where great efforts are being made to familiarise the public with their different functions, styles, and uses. Whether or not this has been successful, is difficult to say. They certainly have a greater presence on social media than they did in the past. Before this year, I’d had some interactions with chatbots in various ways, but that was the limit of my involvement with them. Then, I came across SnatchBot, and made a student mental health bot in a few days. I also realised that platforms like SnatchBot are more common that you would think — though SnatchBot is entirely free, and as far as I know, it’s going to remain that way.
I built a health bot to test the waters of this technology. I found that apart from being very intuitive in its design — I just imagined myself in the user’s position — there are certain pitfalls that need to be avoided, and there are also certain elements that are crucial and need to be included.
Some advice to keep in mind when designing a health bot
1. Use accessible language
Always keep the language of the bot simple, conversational, with only one question per interaction. You need to design a chatbot with all users in mind, and this means ensuring that anyone who interacts with the bot will be able to understand it, and navigate it without any difficulties. As a health bot, there are a seemingly infinite number of possible issues that could arise with users. But that doesn’t mean that you need an infinite number of responses. One sentence structure can be applied to several possibilities. This is both a time saver, and it adds coherence to your bot. I give an example:
User: What are the symptoms of _____?
Health Bot: The common symptoms of _______ are ___ and ____. Less common symptoms are _____ and _____.
This sentence can be applied to many topics. Don’t feel that the bot needs to answer each question in a different way every time — it may sound more human, but sounding human is not the point. Providing accessible, and easy to understand, accurate information is.
2. Envisage your bot’s path
It won’t serve you to begin creating your chatbot interactions if you don’t know where you want these interactions to end up. When I designed the student mental health bot, I wanted the user to be directed to mental health services such as helplines and services, because a chatbot is no substitute for real human counselling. This was the essential destination, no matter how helpful the chatbot was as a point of contact or support. In the same way, before you begin you need to have a clear understanding of where you are going. A chatbot is not the same as a doctor, a fitness coach, or a nutritionist, and it is limited in what it can provide. Unless you’re creating a bot that is purely an information point on a topic, it would be worthwhile to link users to local specialists, or services that they can access for further information.
3. Avoid being prescriptive
This is very important. There is a difference between providing information, and providing solutions. The way to avoid this mistake is to use language carefully. For example, the difference between “You need to…’’, “You should…’’, and “It is recommended by [URL link to a source] that…’’ is quite a large one. That was something I was careful to avoid in the student mental health bot. Not only to avoid any legal issues that might arise from prescribing a course of action, but because in reality, being prescriptive in these situations can cause people to shut down and close themselves off from any help that might be offered. Very often, people just want to talk, and they don’t want to be seen as a problem that needs solving. An example of a mental health bot that listens and shares its thoughts without being prescriptive is WoeBot, on Messenger.
4. Figure out where your bot fits into the market
It’s wonderful that you’re thinking about designing a health bot. And the platforms are there ready for you to use, which is also wonderful. However, while your health bot may be super helpful and easy to use, there are also a bunch of other health bots out there which are also helpful and easy to use — though, definitely not as helpful as yours. If you want your bot to catch on, then a bit of market research is needed. One of the things that chatbots do better than humans is not sleeping. They can provide instant information 24/7. Chatbots are also portable and can be accessed just about anywhere. What kind of health information might people want to access quickly, on the go, or outside of normal working hours? How can your bot best deliver on this, and to whom? For example, a nutrition bot might give information about vitamins, healthy meals, and calorie intake. This could be useful for any supermarket shopper, and could be advertised as such. What, how, and to whom are important questions to ask.
Saving time and accessing information on the go has become a normal feature of our society. Photo credit: Zapp2Photo/Shuttershock.
5. Prioritise quality over quantity
Janis Joplin once sang “you only gotta do one well to make it in this world, babe’’. Because of the nature of chatbot technology, it might feel as if your health bot needs to be able to do lots of things for the user, and do all of these things equally well — perfectly, even. When it comes to health, quality is far more important than quantity. With a combination of the market research discussed earlier, and a clear idea of the bot’s purpose, you need to narrow its focus onto specific areas, or even just one area. Work on making sure that these are your bot’s areas of expertise. With the mental health bot, an area that I focussed on was knowledge about resources. It wasn’t enough to have just one or two helplines or websites in its database, because everybody has different needs.
If you want to create a health bot for women, for example, focus on a certain aspect of women’s health. And narrow that focus again. Women’s health could be specific to reproductive health, which could be again be narrowed down to information about contraceptive pills, which could again be narrowed down to what the approved medical advice is when you miss a pill on a certain day of your cycle. You’re not only making the task easier for yourself, but you’re also making sure that you deliver a truly helpful and valuable service.
6. Remember that a bot is a bot
This sounds obvious, but it’s important to make sure you integrate this knowledge into the design of your chatbot. It could be a small reminder in the bot, for example if the user says something like “But you’re just a machine’’, your bot can respond with “You’re right. I’m not human, but I was created by one who really does want to help’’. A bit of self-awareness on the chatbot’s part shows that it isn’t trying to be something else, to patronise or pretend. This may not be needed for every health bot, but it’s something to keep in mind nonetheless.