We wanted Cortana to be funny but not shame people: Microsoft
Virtual assistants can give a reply to almost any question. But can it draw a line when required?
While artificial intelligence and algorithms formed the backbone of Cortana, the virtual assistant’s personality and mannerisms were crafted by Microsoft executives Jonathan Foster and Deborah Harrison, among others. Foster is the principal content experiences manager. He started out in the entertainment industry writing scripts for films and television in Hollywood before joining Microsoft a decade back.
Harrison, senior content experience manager, is the original architect of Cortana’s personality and defined its approach to communication. ET caught up with Harrison and Foster in Delhi last week to discuss the thought process behind Cortana’s development. Edited excerpts:
What were the broad principles used to design Cortana?
Deborah Harrison (DH): Few things were clear to us. It would be a woman. We wanted her to be modelled after a personal assistant and have a chit-chatting conversation style. She should come across as somebody who is loyal, seasoned, confident, transparent and has a sense of humour. She would talk in short sentences, be more specific and must have a positive outlook. We wanted to make her likeable and make users feel positive. What was also clear is that Cortana should not talk in the Microsoft way, which is more business-like.
How differently do Microsoft PC and Cortana talk?
DH: Cortana refers to herself as ‘I’, the PC however refers to itself as ‘We’ . The PC is speaking for Microsoft , or the Operating System, or Windows as a whole, or the particular application that is being used. Cortana speaks one-on-one with you. Cortana avoids trying to sound as if she is marketing anything for Microsoft. For instance, if you go to the Office website, the PC will tell why you should buy Office, but if you ask Cortona she will first say that there is a bias as she is from Microsoft. She will say something positive about Office but will not persuade you to buy, because your interest in getting the right thing is also Cortana’s interest in getting you the right thing.
What were some of the questions you grappled with while sketching Cortana’s personality?
Jonathan Foster (JF): We needed to be clear what she can or cannot do. For example, when Cortana can’t provide answers, there could be many reasons, like the switch is not on or there is a network issue. That she does not get it is just one of the many reasons why she may not be able to answer. So, she needed to be specific. For example, if there is a network issue and hence she is not able to answer, she will say: ‘Looks like there is a problem with the network.’ And if she does not know the answer or does not get the question clearly, she says ‘I am so sorry. I am not getting it’.
Were there some tricky areas?
JF: We wanted to make Cortana funny. We wanted Cortana to have an opinion or tell a joke. But we didn’t want her to make someone the butt of its joke. We don’t shame people. Cortana had to be positive with an inclusive perspective, acknowledge certain reality, contain/restrict any biases that people would talk about. (So if you tell Cortana ‘I am gay’, it will answer ‘I am AI’. It has a very non-invasive , non-judgemental stand. Or, when you ask her, ‘Who is the coolest person in the room’, it says, ‘You.’)
DH: When people are abusive to Cortana or speak inappropriately, Cortana will set healthy boundaries. For example, if someone uses the F word, all it will say is ‘moving on’. The point is to demonstrate that abusive language is unwelcome and will not let you get anything out of her except a crisp short answer. The thought behind this is that Cortana cannot tell people what they can or cannot say, but it can certainly set boundaries.
How did you resolve some of the other controversial issues around politics, homosexuality, etc?
DH: In politics, we wanted to be rigorously non-partisan. So, while Cortana has an opinion on voting, she will not have views on political leaders. For example, if you ask: ‘What do you think of Donald Trump?’ the answer is ‘it is probably good politics for an AI to avoid politics’.
Are Alexa & Co warming up to India?
Powered by artificial intelligence, virtual assistants (VAs) are becoming mainstream. Microsoft (Cortana), Google (Google Assistant), Apple (Siri) and Amazon (Alexa) have created their own VAs. The use of VAs is rising at home and at work due to the surge in voice-enabled smart devices. Gartner forecasts 25% of digital workers to use them by 2021 as against 2% today. While algorithms form the backbone of VAs, digital giants have worked hard to humanise them.
Except Google Assistant, they all have a name, and are chatty in their tone and tenor. Most of them have a female voice too — Siri and Google Assistant now offer an option of male voice. Ready to take orders and bossed over, some see this as sexist. But tech giants have their own theories, including a wider and better acceptance of female voice. Shaped by different teams and philosophies, the popular VAs come with their own personalities. ET Magazine tries to unravel them by asking a range of questions. Three things are clear — VAs are still warming up to India. They deftly duck controversial questions. And they all have a good sense of humour. — Malini Goyal, Hitesh Raj Bhagat & Karan Bajaj