History Channel Chatbot
As a Storytelling Fellow at History Channel, I was tasked with exploring the future of premium storytelling. I investigated how we could bring history to life through new technologies like voice and chat interfaces. I created RAICH who is an artificially-intelligent (AI) chatbot who likes to chat about the history of AI. I first explored a Facebook Messenger chatbot but then fully developed the project as a Google Home application.
I conducted all of the background research, voice interface design, concept testing, and prototype development.
RAICH came out of an observation that while artificial intelligence is growing in popularity and getting common in conversations, there’s many people who don’t understand the history of its development and the basic nuts-and-bolts definitions of what artificial intelligence really covers.
Through the development of the project, I explored how much information is too much, how to prompt people to ask questions they didn’t know they needed to ask, and how to make somewhat technical information understandable to someone who has no background in the science. I was also invited to attend Google’s workshop about voice user interface best practices to better develop this prototype.
Research and User Testing
I found in my initial research that people were interested in AI as a topic, but generally knew little about the details. I noticed that people either knew what it was on a commercial level (i.e. self-driving cars are artificially intelligent) or on a scare-tactic level (i.e. AI is going to take over the world and technology will rule). It was in this moment that I thought about the gap in topical understanding of a trending topic and the ask at hand: make a premium storytelling experience. Raich (stands for Robust AI Chat) was the answer: an AI Chatbot who loves to chat about the history of AI development. This concept not only satisfied the needs of the brief but also aligned well with History Channel’s subject matter theme.
In user testing, I found that people often felt confused, overwhelmed with information, or didn’t know what to ask. From here, I started to shape the conversation to give users clues about other topics they could ask about. I also adjusted the type of language used. Instead of saying “Marvin Minsky was a cognitive neuroscientist” I would adjust the language to say “Marvin Minsky studied the brain.” This language shift made the topic and history much more accessible for anyone who wanted to learn.
A major part of the project was figuring out the persona of Raich and how she interacted with the user. Because I was essentially designing an interaction with a character, I had to start considering who this character was, how they spoke, and what their core drivers were. After a workshop at Google, I learned how to approach this type of design. I created lexicon guides that would show what phrases Raich would definitely say, what she’d never say, and what she might say. I developed her brand attributes and what design elements would support those and how those would impact her persona.
Given that I wouldn’t have been able to possibly collect all the facts about AI and stitch them together in a meaningful way, I had to think strategically about what content I collected and how I arranged the conversation to stay somewhat in the boundaries in a limited set of data.
I created spreadsheets of information and organized it by the following: decades and timelines, random facts, people-oriented information, definitions (or what is? questions) and quotes. From there, I created connections between each topic so that the chatbot would suggest inquiries that would lead the user to this main dataset.