Jonathan Foster is the principal content experiences manager, Windows Content Intelligence at Microsoft. His work includes writing conversational interactions for products that include the Virtual Agent, a chatbot on support.microsoft.com. As a writer, Foster does not come from a technology background but started out in film and television writing screenplays. He was drawn away from Hollywood by the innovative spirit of the tech industry. Similarly, Deborah Harrison is a senior content experience manager, conversational AI and intelligence at Microsoft and is a writer on Microsoft's digital assistant Cortana's editorial team, which is the team tasked with crafting and creating Cortana’s written and verbal responses. In her role, Deborah leads a team of writers who are industry experts in personality design, dialogue authoring, and machine learning. She is one of the original architects of the personality for Cortana. Both these global experts spoke to THE WEEK on the sidelines of an event in New Delhi about the importance of conversation and emotions in chat bots and their future.
How important is Conversational AI in the field of Machine Learning and applying it in computing interactions? Why it is important to have emotions even in computing interactions?
There is compelling research showing that people become emotionally attached to their devices even if the device has no personality, People can describe the personalities of their cars, their vacuum cleaners, even their sewing machines, and they certainly do it with their computers and phones. Further research shows that people don’t trust personalities that seem ambiguous or wishy-washy. So when we set out to craft our writing guidelines for conversational agents like Cortana, we decided it was important to honor that, and create a personality that felt grounded in an emotional reality. We have been careful never to suggest that the personalities we design are actually human, but we look for human qualities like empathy and positivity to inform our writing. We figure that when someone asks an emotional question, we can choose not to answer, or answer coldly, but we believe that there is value in honouring someone’s emotional state even one half of the conversation comes from a non-human, as long as we are transparent about that fact.
There is another reason, as well. Conversation is a powerful means for people to communicate with their apps and devices, and the promise we make when we allow people to do is that they should be able to talk to their devices using the same kind of language they use to talk to other people. Ideally, there should be no learning curve. That means that our job is to make conversations feel as intuitive as possible, and emotions are part of human communication.
You mentioned about different types of computing conversations for different regions where conversational bots are present. How important is it to customise the computing interactions with regard to the different geographical regions?
We believe that it is very important to customise the interactions to each market to ensure cultural relevance and a natural, human fluency. Conversationally founded interactions require a lot more design attention than more traditional modes of interaction relying on keyboard and mouse. This is simply because these natural language-based modalities are much more familiar to people and resonate much closer to the human experience. So translation of what is being written in the US market will not suffice because it will feel considerably less authentic. Human communication is extremely nuanced. Language, and its foundational relationship to conversation, plays a huge role in that nuance. So you need dedicated resources that have a deep understanding of how any given language functions, and more importantly, feels.
What kind of people work in your team with regard to conversational AI and how challenging is it to bring in the element of emotions in the field?
We are passionate about bringing the right kind of writing and designing talent onto our teams. It means looking in new places for a new kind of writer and designer. Conversational AI requires a different perspective because we are really trying to get to the humanness of interaction, and in doing so, are really pressing in to the human experience. Core to that human experience is the emotional state we are always in, which is very powerful and, to many of us, something we are often unaware of. This requires a design approach that considers the kind of impact our products and experiences are having on people. This requires people who have focus on the human experience, writers and designers from humanities and arts. We have people with backgrounds in screenwriting, playwriting, fiction and non-fiction writing, as well as philosophy, psychology and English literature.
When we sit down to consider the emotional state of an individual in any given moment, we intent on accommodating that state. Products and experiences are more responsive than they have ever been, and that is a game changer. We need to respond appropriately in human-like interactions, while being completely transparent that the response is not real human emotion. This is a design responsibility, to acknowledge, respect and honor the emotional reality of the human experience in every moment. For example, if someone is very frustrated, we want the experience we are creating to respond with sensitivity. We have found that writers and designers with experience in the arts and humanities are very well suited for this work because they are inherently focused on what it means to be human.
Conversational Bots are active in different fields such as the real estate segment, hospitals etc? Can you give some real time examples as to how the element of conversational AI will make the experience of a customer or an individual much more enriching?
We believe that conversation is an extremely powerful way to help people solve problems and get information. One of the most valuable tools we can offer in conversation is the ability to help someone figure out how to even identify their issue, and then solve it. It gives you the power to not know everything. For example, imagine that someone has just bought a new headphones and wants to use them with their phone. Without conversation, they need to figure out on their own that their headphones use something called “Bluetooth,” and they need to do something called “pairing” to make it work, and they need to do something on their headphones and on their device. But with conversation AI, you can imagine a conversation where the person simply says to their device “I need to connect my headphones,” and the phone can ask a series of questions to figure out what you need to know to get things connected. Are they wireless? Are they turned on? How can you tell? Which of these items in the list is the right one? The person doesn’t have to know anything about Bluetooth and pairing. You can imagine something similar for a hospital, or in real estate. A conversation helps you think as you go. You are house hunting and you want to see only houses in your price range in a certain zip code. Ok, now with two bathrooms. Only with external garages. Never mind, show me the same thing but with townhouses. That’s all cool, but a good conversational agent could also ask you questions to understand better what you want, things you may not have thought to ask. Perhaps information about the school system, or bus routes, or tax levies, or the neighbourhood.