8. December 2016
–They're coming: Chatbots for Facebook messenger. Maybe you've heard of Hi Poncho, JOBmehappy, Dinner Ideas and Nerdify Bot.
But what about two new healthcare chatbots, Izzy and Florence? I decided to test out these new bots and see what all the buzz is about. Full disclosure: I had never used a chatbot before testing Medilab's Izzy and David Hawig's Florence.
Izzy is a "cute little bird" that I never knew I needed to help me "take care" of my period. She did not waste any time getting personal.
Is it wrong that I decided to mess with her?
Sixty-three days, turns out is too long and now she was onto me.
After answering a few basic questions, she prompted me in the right direction; I am now set up to receive a message to remind me when my period will start. Lucky me.
I am not sold on this. If this is the extent of Izzy's capabilities, I don't see the point. I quizzed her. Cramps, STIs, STDs, family planning... any sexual health-related trigger word I could think of.
Her response?
The only way to redeem Izzy, I think, would be to create a chatbot that allows men and women the opportunity to ask private, sex-related questions and provides answers with information from a reliable source.
Wouldn't this be worthwhile? There are still many people and places in the world where such topics are difficult to discuss and often go unaddressed.
Access to information, moreover bringing information to people where they are, is exactly what David Hawig's chat Florence aims to do.
Here is what Florence offers:
First, I tested the Information option by searching the term Diabetes.
An essentially jargon-free definition. Pretty cool. What happens if I ask for more?
Each time I asked for more information, Florence provided two messages, each with increasingly specific information in a logical order: Causes, Diabetes types, treatment options, prevention, global costs associated with the disease and so on.
This function is digestible and useful, like Twitter meets WebMD. I could see myself using it. But it raised two questions: Does Florence somehow use the information I search for, like targeted advertisements? And what is the source Florence relies on; how reliable is this database of information?
So I asked Hawig.
"Unfortunately conversations with a chatbot in Messenger are, at the current time, not adequately secure," he said. But he went on to reassure me that "the data are not being collected at the moment and that will also not happen in the future."
I was relieved to hear this.
Hawig then told me about the APIs, or application program interfaces – a set of protocols that determine how software should interact – he uses. The information Florence provides comes from MedlinePlus, U.S. Food and Drug Administration, ApiMedic, Wikipedia, Genderize.io and Google places API, he explained. Though not all of them are active at the moment, he said.
Next, I tested the Symptoms option.
Florence lets chatters choose from the following options: Skin & general, Head & throat, Chest & back, Arms & shoulder, Abdomen & buttocks and Legs.
I selected Skin & general, which prompted me to choose from a "sub-location" and then directed me to 30 different symptom options, like flaking skin, blackheads, blue skin, scars and pallor.
I eagerly chose a symptom only for Florence to say:
We didn't make it any further.
After hitting this speed bump, I switched to the Medication reminder.
I noted the typo (scheduel, not schedule), which Hawig also warned me about, and realized the trouble I have with these chatbots. Is this chatting and am I benefiting?
One could, and I will, argue that the benefit of conversation, be it online or in person, is the exchange of ideas – new ways of framing and relaying information. Does that not require a sentient being who can relate to that sentiment? Other than the fact it is cheaper, is accessible 24/7 and requires zero man power, what are the advantages to chatting with a machine?
In the most limited sense, yes, this is a conversation, but really all I am doing is providing one word prompts that are simple enough for the artificial intelligence to understand and ultimately perform a desired task.
Maybe it boils down to expectations. What do I, what do we, expect from chatbots? What do we want from them, especially when it comes to health?
If I have a tummy ache, am I more likely to turn to Florence for possible causes, or will I message my family and friends?
I can personally say, I will turn to real conversations first, but that doesn't take away from the genius of integrating chatbots like Florence into social messaging platforms. Both Florence and my close friends are conveniently located in one platform, which means multitasking and sourcing is consolidated into one place.
I think the potential of chatbots lies in this sweet spot, this perfect overlap.