The use of chatbots in some areas of healthcare (psychology, for example) seems to be a promising space for the early integration of AI-based systems, especially given the massive improvements in natural language processing over the past few years.
You may be thinking that we’ll be less likely to share personal information with a machine but it actually looks like the opposite is true, especially when it comes to personal and sensitive information.
And the reason is counterintuitive.
I thought that human beings would prefer interactions with other human beings, and that our shared humanity would lead to more honest and authentic conversations. But it’s actually our humanity that acts as a barrier to sharing certain types of information.
There are fairly strong social and cultural norms around what parts of our lives we believe are appropriate to share with others and there’s overlap between these taboo subjects and the kinds of information that healthcare providers are looking for. This is why we have to “build rapport” with clients; we have to work for their trust so that they’re comfortable sharing the sensitive information we want. A lifetime of conforming to social norms makes it hard to share personal information with others.
The reason we’re so sensitive to the information we share with people is because we worry about the social judgement that comes with the interaction. But this sensitivity seems to be dialed down when we know we’re interacting with a machine.
In fact, even when we know we’re interacting with a machine, if we humanise the machine-generated language by adding phatic cues (e.g. “umm”, “ah”, etc.), people are more likely to try and present themselves in a positive light. In other words, we’re so conditioned to respond “appropriately” to human beings that we only need to add minor language cues to a machine to make us more guarded and to filter what we say.
It’s early days but it seems like we’re more honest with machines than we are with people. This is probably going to be really important when we’re designing client-facing interfaces for chatbots and other AI-based systems in healthcare.
This is a good overview of some of the research describing how we interact with virtual humans and chatbots.