How human should a chatbot be? Transparency, personality, and on brand w/ Clara de Soto & John Keefe - a podcast by IBM thinkLeaders

How human should a chatbot be? In this episode of thinkPod, we are joined by chatbot experts Clara de Soto and John Keefe. Clara is a serial entrepreneur and co-founder of Reply.ai, and John is the technical director for bots and machine learning at Quartz. We talk to Clara and John about the types of personalities chatbots should have, bot transparency, why liberal arts majors may now have jobs as conversational designers, and delicate balance brands face when crafting a chatbot. We also hear about creating a Westworld chatbot for superfans, the danger of having a chatbot make suggestions for pregnant women, and whether our pumpkin spice latte should chat with us.

Connect with the guests:
John Keefe @jkeefe
Clara de Soto @Claradactyl

Some of the questions we deal with include:

Does it make sense to have chatbots and AI have names and their own personalities?
Do companies think through their brand identity when creating a bot?
When is it appropriate to use a bot versus when do people want to have that actual human conversation?
How will the rise of chatbots impact the job outlook for customer service agents?
How important is explainability?
Key quotes from the podcast include:

“You know, it’s almost like liberal arts majors everywhere can rejoice in having a career path as conversation designers.” – Clara de Soto

“Actually more challenging turns out to be, ‘how does this bot fit into the cannon of the storyline that exists already?’ And that actually turns out to be quite a challenge because suddenly the bot is speaking for the brand or the show and hardcore fans are going to be like, okay, what does this reveal about the existing characters? How does this fit in? And so that’s stuff that we have to work through.” – John Keefe

“In that case it had an average conversation length of three minutes [for the Kia Niro bot]. I will never forget this one dude in Texas talked to it for 42 minutes. A car.” – Clara de Soto

“It’s not just spitting out results onto the web page, right? It’s humans looking at it, reporters looking at it, questioning it, and then figuring it out. Being able to explain it is part of being able to be transparent and be truthful.” – John Keefe, discussing the explainability needed for bots

“I like to think that bots give those people [customer service agents] super powers. Because the bot will never be able to fully take over what a human being is capable of answering, especially when you get into more complexity. So it’s more just like making them more efficient and better.” – Clara de Soto

Full transcript: https://www.ibm.com/blogs/watson/2019/04/thinkleaders-podcast-how-human-should-a-chatbot-be/

Further episodes of IBM thinkLeaders

Further podcasts by IBM thinkLeaders

Website of IBM thinkLeaders